Abstract

   Colorectal cancer (CRC) is the third most common cancer diagnosed and
   the second leading cause of cancer-related deaths. Emerging evidence
   has indicated that long non-coding RNAs (lncRNAs) are involved in the
   progression of various types of cancer. In this study, we aimed to
   identify potential causal lncRNAs in CRC through comprehensive
   multilevel bioinformatics analyses, coupled with functional validation.
   Our bioinformatics analyses identified LINC02257 as being highly
   expressed in CRC, and associated with poor survival and advanced tumor
   stages among patients with CRC. Genome-wide association analysis
   revealed significant associations between variants near LINC02257 and
   CRC, suggesting a causal role for LINC02257 in CRC. Network analysis
   identified LINC02257 as playing a key role in the
   epithelial-mesenchymal transition pathway. Single-cell RNA sequencing
   showed that elevated expression of LINC02257 was associated with a
   reduced proportion of epithelial cells. In vitro experiments showed
   that LINC02257 positively regulated the metastatic and proliferative
   potential of CRC cells. Mechanistically, LINC02257 affected CRC
   malignancy by functioning as a competitive endogenous RNA of microRNAs
   and RNA-binding proteins. LINC02257 upregulated SERPINE1 by
   sequestering tumor suppressive miR-1273g-3p, thereby increasing
   metastatic and proliferative abilities of CRC cells. Additionally,
   LINC02257 directly interacted with YB1 and induced its phosphorylation,
   thereby facilitating YB1 nuclear translocation. The transcriptional
   activation of YB1 target genes was associated with the oncogenic
   functions of LINC02257. Taken together, our results demonstrate
   LINC02257 as a promising therapeutic target for CRC treatment.

   Subject terms: Oncogenes, Long non-coding RNAs

Introduction

   Colorectal cancer (CRC) is the most prevalent gastrointestinal cancer
   and the second leading cause of cancer-related mortality [[42]1].
   Advancements in diagnostics and therapeutics have dramatically
   increased the survival time of patients with early-stage CRC. However,
   the 5-year survival rate of patients with advanced CRC remains low
   [[43]2]. Despite its high mortality rate, an effective treatment for
   patients with CRC has not yet been developed. Therefore, it is
   essential to elucidate the molecular mechanisms underlying colon
   carcinogenesis and explore diagnostic and therapeutic biomarkers
   [[44]3].

   Research on the discovery of therapeutic and diagnostic biomarkers for
   CRC has shifted from protein-coding genes to non-coding RNAs (ncRNAs).
   LncRNAs are a class of ncRNAs that are more than 200 nucleotides in
   length. They are typically transcribed by RNA polymerase II, spliced,
   5’ capped, and polyadenylated [[45]4]. Because they are recognized as
   powerful regulators of gene expression which are responsible for the
   formation and malignant progression of cancers, including CRC, lncRNAs
   have attracted the attention of many oncologists interested in the
   molecular mechanism of carcinogenesis and metastasis. In CRC, a series
   of lncRNAs have been identified as oncogenes or tumor suppressors,
   closely related to the malignant phenotypes of CRC by regulating cell
   proliferation, cell cycle, epithelial-mesenchymal transition (EMT),
   drug resistance, and metastasis [[46]5].

   LncRNAs regulate gene expression in various ways. The role of lncRNAs
   as a competitive endogenous RNA (ceRNA) has recently garnered
   significant research interest [[47]6]. The regulatory function of ceRNA
   is largely dependent on the cis-acting elements present within its
   sequence, enabling it to competitively inhibit microRNAs (miRNAs) or
   RNA-binding proteins (RBPs) from binding to the target mRNA. Tens of
   lncRNAs function as ceRNAs and are closely associated with the
   malignant phenotypes and prognosis of patients with CRC [[48]7–[49]9].

   In this study, using The Cancer Genome Atlas (TCGA) and Genotype-Tissue
   Expression (GTEx) Project data [[50]10], we identified LINC02257 by
   combining multiple bioinformatic approaches, including gene expression
   analysis, survival analysis, expression quantitative trait loci (eQTL)
   analysis, and co-expression network analysis. Based on the results that
   LINC02257 is present in both nucleus and cytosol, we identified two
   mechanisms by which LINC02257 regulates metastatic and proliferative
   abilities in CRC. First, LINC02257 sequesters tumor-suppressive
   miR-1273g-3p, thereby upregulates SERPINE1. Second, LINC02257
   transactivates Y-box protein 1 (YB1) by inducing its phosphorylation
   and thereby enabling its nuclear translocation. Taken together, our
   results demonstrate LINC02257 as a promising target for the treatment
   of CRC.

Materials and methods

   Detailed analytical and experimental methods are described in
   Supplementary Methods.

Transcriptome data collection

   TCGA catalogs the genomic data of multiple cancers operated by the
   National Cancer Institute (NCI) and the National Human Genome Research
   Institute (NHGRI). We used the transcriptome and clinical data of CRC
   and normal-tissue samples, and excluded metastatic or recurrent tumor
   samples.

   The initial number of transcriptome data points was 478 and 41 for
   colon tumors and normal tissues, respectively, while the total number
   of clinical data points was 461 owing to redundant tumor data from the
   same patients. To eliminate redundant data, we specifically selected
   samples with the highest lexicographical number of TCGA barcodes, which
   served as the biospecimen identifier, except for formalin-fixed
   paraffin-embedded samples. The filtering criteria were provided by the
   GDAC Broad Institute
   ([51]https://gdac.broadinstitute.org/runs/gdc/report_2017_08_02/TCGA-CO
   AD_Replicate_Samples.html). If there are multiple aliquots associated
   with the chosen RNA analyte, we selected the aliquots that were
   preserved by freezing and had the later plate number. After excluding
   samples lacking clinical information on follow up days or tumor stage,
   482 samples (442 tumors and 40 control samples) remained (Supplementary
   Fig. [52]S1).

Normalization and differentially expressed gene (DEG) analysis

   For our gene expression analysis, we used read-count data from samples
   of both CRC and normal tissue (total number of genes = 60,483). Genes
   with read counts <10 were excluded, and the gene expression level
   values of the remaining genes were normalized across the samples and
   grouped into tumor and normal tissues. We used the maximum a posteriori
   (MAP) function to estimate gene-specific dispersion across samples and
   the log[2] fold change (LFC) of each gene between the two groups using
   the DESeq2 package (v 1.22.2). After fitting the LFC as a coefficient
   in the negative binomial generalized linear model, we tested the
   significance of these coefficients in each gene using the Wald test, as
   the default setting for DESeq2. For the identification of
   differentially expressed (DE) lncRNA, cases that met the criteria of
   |LFC | ≥ 1 and adjusted P < 0.05 were selected. In addition, an
   independent CRC data set, [53]GSE146009 from Gene Expression Omnibus
   (GEO) that provides only the numeric data, was imported into ‘edgeR’
   for DE analysis [[54]11]. Differential expression was assessed using an
   exact negative binomial test and further adjusted using the
   Benjamini-Hochberg (BH) procedure.

Prognostic analysis

   To investigate the prognostic effects of the identified lncRNAs on each
   patients with cancer, we performed Cox proportional hazard regression
   analysis using the ‘Survival’ R package (v 3.1-8). For each lncRNA,
   samples were classified into groups with high- or low-expression based
   on the median expression value and the hazards between the groups were
   compared. Spearman’s correlation and Cochran-Armitage trend tests were
   used to assess whether the expression levels of candidate lncRNAs
   increased as the tumor stage advanced. For the Cochran-Armitage trend
   test, the data were divided into two groups according to the median
   expression value [[55]12, [56]13]. To test the association between
   lncRNA expression and microsatellite instability (MSI), Spearman’s
   correlation was used after removal of samples whose MSI information was
   “not reported” or “indeterminate”. A significant association of lncRNAs
   with prognosis was examined using a BH-adjusted P-value of <0.05.

Cell culture and transfection

   CRC cell lines (DLD1, HCT116, RKO, and HT29) were cultured in
   Dulbecco’s modified Eagle’s medium (Gibco, Waltham, MA, USA). For
   transfection, Lipofectamine 2000 (Invitrogen, Carlsbad, CA, USA) was
   used. The siRNA sequences used in this study are shown in Supplementary
   Table [57]S1.

Western blot and RT-qPCR analyses

   Protein and mRNA expression levels were determined using western
   blotting and RT-qPCR, respectively. Information regarding the
   antibodies and primer sequences used is provided in Supplementary
   Tables [58]S2 and [59]S3, respectively.

Determination of malignant phenotypes

   Metastatic potential including invasive and migratory abilities was
   determined using the BD Biocoat™ Matrigel invasion chamber (BD
   Bioscience, San Jose, CA, USA) and Transwell® permeable supports
   (Corning, Rockville, MD, USA), respectively [[60]14]. Cell
   proliferation was assessed using a colony formation assay and cell
   counting.

Cellular fractionation

   The localization of LINC02257 was examined using a cellular
   fractionation assay [[61]15]. The levels of α-tubulin and lamin B were
   assessed to determine the composition of cytosolic and nuclear lysates,
   respectively.

Ribonucleoprotein immunoprecipitation (RIP)

   Argonaute 2 (AGO2) RIP assay was conducted to determine the binding of
   miR-1273g-3p to LINC02257 and SERPINE1 mRNA. The interaction between
   YB1 and LINC02257 was verified using RIP assay [[62]15].

Antisense oligonucleotide (ASO) pulldown assay

   ASOs that recognize LINC02257 were designed and used to identify
   LINC02257-associated miRNAs and RBPs. A schematic of ASO pulldown
   experiment is shown in Supplementary Fig. [63]S22.

Luciferase reporter assay

   The sequence-specific interaction of miR-1273g-3p was verified using
   luciferase vectors (pmirGLO dual-luciferase vectors; Promega, Madison,
   WI, USA) containing wild-type or mutated sequences of the miR-1273g-3p
   miRNA recognition element (MRE) (detail information in Supplementary
   Fig. [64]S6).

Results

Identification of candidate lncRNAs for CRC

   To identify candidate causal lncRNAs, we systematically identified
   lncRNAs that were differentially expressed (DE) between normal and
   cancerous tissues and exhibited association with survival and tumor
   stages in CRC using RNA-seq data from TCGA cohort (Fig. [65]1A and
   Supplementary Fig. [66]S1A). We initially identified 832 DE lncRNA
   ( | LFC | ≥ 1 and adjusted P < 0.05); 566 were over-expressed while 266
   were downregulated in tumor tissues (Fig. [67]1B, C and Supplementary
   Table [68]S4). Next, the association of these 832 lncRNAs with the
   survival of patients with CRC was investigated; the expression levels
   of five lncRNAs were associated with overall survival (OS) (Fig. [69]1D
   and Supplementary Fig. [70]S1B). Increased expression levels of four
   lncRNAs ([71]AL137145.2, [72]AL157400.2, LINC02257, and LINC01836) were
   associated with a worse prognosis, whereas higher expression of
   [73]AC079612.1 was associated with a good OS. Third, we explored the
   association between lncRNA expression and CRC stages. High expression
   levels of LINC02257 and LINC01836 were positively correlated with
   advanced tumor stages (adjusted P = 0.0022 and adjusted P = 0.0023,
   respectively, Cochran-Armitage trend test) (Fig. [74]1E and
   Supplementary Fig. [75]S1C). Finally, significant positive associations
   between expression of two lncRNAs ([76]AL137145.2 and LINC02257) and
   the MSI class were observed in patients with MSI data (adjusted
   P = 1.39 × 10^−10 and adjusted P = 1.39 × 10^−10, Spearman’s
   correlation) (Fig. [77]1F and Supplementary Fig. [78]S1D). In Cox’s
   proportional hazard regression analysis adjusted for sex, tumor stage,
   and MSI, high LINC02257 expression was associated with OS (hazard ratio
   [HR] = 2.00, 95% confidence interval [CI] = 1.29–3.10, P = 0.002) (Fig.
   [79]1G, H).

Fig. 1. Overview of the process of identifying candidate causal long
non-coding RNAs for colon cancer.

   [80]Fig. 1
   [81]Open in a new tab

   A The systematic analyses process to identify differentially expressed
   (DE) lncRNAs between colon cancer and normal tissue samples is
   illustrated. B Filtered genes are represented using volcano plot. Genes
   exhibiting |log2 fold change | ≥ 1, and p-value < 10^−6 are highlighted
   in red, those with p-value < 10^−6 (blue), and |log2 fold change | ≥ 1
   in green, and non-significant genes in grey. Significantly
   prognosis-associated 5 lncRNAs were annotated. C Clustered samples are
   represented using heatmap, revealing 832 DE lncRNAs in colon tumors
   compared with normal tissues. Sample types and tumor stages are
   indicated in the top bar, while gene types are denoted on the right.
   Five annotated lncRNAs ([82]AL157400.2, [83]AC079612.1, [84]AC137145.2,
   LINC02257, and LINC01836) exhibit higher expression and are associated
   with poor prognosis in colon cancer (false discovery rate < 0.05). D
   Kaplan-Meier curves of overall survival in patients with colon cancer
   were generated based on high and low expression of LINC02257. For
   individual lncRNAs, patient samples of colon cancer were divided into
   groups with either high or low expression levels of each lncRNAs based
   on the median expression value. E, F Expression levels of LINC02257 was
   increased according to (E) tumor stages and (F) MSI categories.
   (****P < 0.0001, ***P < 0.0005, **P < 0.005, *P < 0.05). Dots represent
   outlier samples whose expression levels are beyond the minimum and
   maximum values. The minimum value is the 25th percentile of the data
   minus 1.5 times the interquartile range (IQR) and the maximum value is
   the of 75th percentile of the data plus 1.5 times the IQR. G, H Forest
   plots describe the hazard ratio (HR) and 95% confidence interval (CI)
   of clinical and genetic factors. Univariate (G) and multivariate (H,
   removed 6 sample that have “not reported” in MSI status). Cox’s
   proportional hazard regression was used to assess association with
   overall survival.

   Next, we examined the associations between dysregulated LINC02257
   expression and various types of cancer using two independent datasets:
   TCGA RNA-seq data for various cancers, and GEO data for colon cancer
   (Supplementary Table [85]S5). LINC02257 was upregulated in tumor
   tissues across most organs (Fig. [86]2A). Notably, upregulation of
   LINC02257 was significantly associated with advanced tumor stages in
   rectum, kidney, head and neck, and bladder cancers, and correlated with
   poor OS in patients with pancreas and kidney cancers (P < 0.005,
   Spearman’s correlation; Cochran-Armitage trend test; survival
   analysis). In the association analyses of DE, OS, and tumor stages,
   [87]AC079612, [88]AL137145, and [89]AL157400 did not consistently
   exhibit associations with pan-cancer (Supplementary Fig. [90]S2).
   Although LINC01836 is upregulated in colon and rectal cancers, it was
   down-regulated in kidney, lung, and head and neck cancers.
   Additionally, high expression of LINC01836 was associated with good OS
   but not with advanced tumor stages in other organs (Supplementary Fig.
   [91]S2). Moreover, in an independent [92]GSE146009 colon cancer RNA-seq
   data [[93]16], compared with normal tissues, LINC02257 exhibited
   significantly higher expression levels in colon cancer tissue sample in
   both African American and Caucasian American populations (Fig. [94]2B).
   These results suggest that LINC02257 may be a key candidate lncRNA in
   pan-cancer carcinogenesis.

Fig. 2. Multi-validated analysis of long non-coding RNAs associated with
colon cancer.

   [95]Fig. 2
   [96]Open in a new tab

   A Bar plot of p-values calculated from differentially expressed (DE),
   overall survival (OS), and tumor stage association analysis for
   LINC02257 in pan-cancer data of the Cancer Genome Atlas (TCGA). The red
   dashed line indicates a p-value of 0.05. The log-rank test was used to
   calculate OS in survival analysis. Cochran-Armitage trend tests
   evaluated expression effects in relation to tumor stages. B LINC02257
   expression in colon cancer among African Americans (AA) and Caucasian
   Americans (CA) from the GEO146009 dataset. C, D Colorectal cancer risk
   variants (diamonds) were identified based on Genome-wide association
   studies (GWAS). rs12032598, within 1 Mb of LINC02257, was found to be
   significantly associated with colorectal cancer in the East Asian GWAS
   (C). rs12135286, also within 1 Mb of LINC02257, exhibited significant
   association with colorectal cancer in the European GWAS (D). E–G Violin
   plot of gene expression levels of LINC02257 according to single
   nucleotide polymorphisms (SNPs). Increased LINC02257 expression levels
   were consistently associated with the risk alleles of SNPs. rs12032598
   associated with lung (p = 2.3 × 10^−42) and nerve (p = 4.4 × 10^−18)
   tissues (E), rs12135286 associated with lung (P = 4.3 × 10^−31) and
   nerve (P = 6.4 × 10^−13) tissues (F), and rs6695584 associated with
   lung (P = 5.6 × 10^−38) and nerve (P = 2.5 × 10^−16) tissues (G) from
   the GTEx database. H The analysis of LINC02257 expression according to
   the genotypes of rs6695584 in TCGA. I p-value of eQTL associated with
   rs6695584 and genes within 1 Mb of LINC02257. Abbreviations: COAD colon
   adenocarcinoma, ESCA esophageal carcinoma, STAD stomach adenocarcinoma,
   LIHC Liver hepatocellular carcinoma, PAAD Pancreatic adenocarcinoma,
   READ Rectum adenocarcinoma, KIRC kidney renal clear cell carcinoma,
   KIRP kidney renal papillary cell carcinoma, BRCA breast invasive
   carcinoma, CESC cervical squamous cell carcinoma and endocervical
   adenocarcinoma, UCEC uterine corpus endometrial carcinoma, LUAD lung
   adenocarcinoma, LUSC lung squamous cell carcinoma, HNSC head and neck
   squamous cell carcinoma, BLCA bladder urothelial carcinoma. ^aThe
   American Joint Committee on Cancer (AJCC) stage was used (Kidney renal
   clear cell carcinoma, head and neck squamous cell carcinoma, and
   bladder urothelial carcinoma). ^bFor gynecological cancer, the
   International Federation of Gynecology and Obstetrics (FIGO) stage was
   used (Cervical squamous cell carcinoma and endocervical adenocarcinoma,
   and uterine corpus endometrial carcinoma).

Genetic variants associated with colon cancer and expression of LINC02257

   Subsequently, we investigated whether genetic variants in the LINC02257
   region were associated with colon cancer using summary statistics data
   from genome-wide association studies (GWAS) of colorectal cancer in
   East Asian and European populations. In the East Asian GWAS, a lead
   single-nucleotide polymorphism (SNP), rs12032598, exhibited a
   significant association with the risk of CRC (P = 6.16 × 10^−10, odds
   ratio [OR] = 1.155, 95% CI = 1.104–1.208) (Fig. [97]2C and
   Supplementary Fig. [98]S3A) [[99]17]. The European GWAS also
   demonstrated a significant association between rs12135286 and colon
   cancer in close proximity to LINC02257 (P = 9.50 × 10^−7, OR = 1.141,
   95% CI = 1.082–1.203) (Fig. [100]2D and Supplementary Fig. [101]S3B)
   [[102]18]. All variants in high linkage disequilibrium with rs12032598
   and rs12135286 were located near the start site of LINC02257,
   suggesting a potential association of these variants with LINC02257
   expression as a cis-expression quantitative trait (eQTL) (Fig. [103]2C,
   D). Additionally, we examined whether the expression of other genes
   within 1 mega-base pair (Mb) of LINC02257 was associated with colon
   cancer. Among 25 genes located near LINC02257 (Supplementary Table
   [104]S6), eight genes, C1orf140, DUSP10, LINC01655, RNU6-403P,
   LINC02474, LINC01705, TAF1A, and [105]AL592148.1, showed significantly
   higher DE levels in CRC than in normal tissues. Of these, only
   LINC02474 demonstrated significant associations with poor survival
   (adjusted P = 0.018) and advanced tumor stage (adjusted P = 0.023,
   Cochran-Armitage trend test) (Supplementary Table [106]S6). At this
   locus, only LINC02257 was significant in all analyses, including the
   GWAS and DE analysis, and was associated with a poor prognosis.

   To identify eQTL for the two GWAS variants (rs12032598 and rs12135286)
   located near LINC02257, we used genotype and transcriptome datasets
   from the GTEx database (v8). Expression levels of the three genes were
   associated with genotypes of variants in three tissues: rs12032598 was
   associated with LINC02257 in lung (P = 2.2 × 10^−42) and nerve
   (P = 4.4 × 10^−18) tissues (Fig. [107]2E), and with LINC01705 in
   cell-cultured fibroblasts (P = 5.7 × 10^−7). rs12135286 was associated
   with LINC02257 in lung (P = 4.3 × 10^−31) and nerve (P = 6.4 × 10^−13)
   tissues (Fig. [108]2F), and with LINC01705 in cell-cultured fibroblasts
   (P = 3.2 × 10^−7) and DUSP10 in nerve tissues (P = 1.2 × 10^−5). The
   eQTL showed that the expression of LINC02257 increased among carriers
   of the rs12032598 C (risk allele) and rs12135286 T (risk alleles),
   except in the esophageal mucosa (Supplementary Fig. [109]S4–[110]5).

   We further examined the association of GWAS lead SNPs with LINC02257
   expression using the TCGA data. As the GWAS variants (rs12032598 and
   rs12135286) were not available in the TCGA genotype data, we assessed
   rs6695584, rs6691195, and rs17011200, which exhibit strong linkage
   disequilibrium (LD) with lead SNPs (r2 > 0.8, from the 1000 Genomes
   Project) (Fig. [111]2G and Supplementary Fig. [112]S6). The expression
   of LINC02257 was higher among carriers of rs6698857 GG or GA genotypes
   compared to AA carriers (T-test, P = 0.0054, Fig. [113]2H) and cis-eQTL
   analysis revealed an association (P = 8.03 × 10^−4) between rs6695584
   and the covariate-corrected expression of LINC02257 (Fig. [114]2I).

Functional annotation and cell-type inference from the co-expression network
in colon cancer

   LncRNAs have emerged as key players in cancer pathophysiology, engaging
   in intricate interactions with various genetic elements such as DNA,
   mRNAs, miRNAs, and proteins [[115]19]. To detect genes that interacted
   with LINC02257 and understand their associated mechanisms, we performed
   network analysis based on the gene expression data (Fig. [116]3A and
   Supplementary Fig. [117]S7) [[118]20]. Of the total 69 identified
   co-expression modules, representing clusters of highly interconnected
   genes, we focused on the ‘Grey60’ that includes LINC02257 which was
   shown to be a promising candidate causal gene. The ‘Grey60’ module
   contained 179 genes, and the module membership of LINC02257 was 0.52 in
   the ‘Grey60’ module (Fig. [119]3B, C).

Fig. 3. Gene clustering via WGCNA and functional analysis.

   [120]Fig. 3
   [121]Open in a new tab

   A Clustering results were obtained from the colon cancer dataset using
   weight gene co-expression network analysis (WGCNA). Co-expressed genes
   are grouped into 69 modules, each denoted by a distinct color. The red
   line signifies the cutoff value used for gene clustering in the
   dendrogram. B Counts of genes per gene types in the ‘Grey60’ module,
   which includes LINC02257. Blue bars represent total genes, and orange
   bar indicate genes with p (rho) > 0.6 (Spearman’s correlation). C
   Network plot of the ‘Grey60’ module utilizing Cytoscape. Circles
   represent the nodes connected to LINC02257, and yellow circle
   represents LINC02257. D Pathway and Gene Ontology (GO) analysis for
   genes co-expressed with LINC02257 using WebGestalt. Pathway enrichment
   analysis based on the Kyoto Encyclopedia of Genes and Genomes and
   Panther pathways was performed for genes within the ‘Grey60’ module. E
   GO analysis results for biological processes, cellular components, and
   molecular functions. F Workflow of CIBERSORTx to examine the cell
   proportion of each donor in the Cancer Genome Atlas. G Comparison
   between the distribution of Spearman’s correlation coefficients of two
   gene sets. Red board line; ‘Grey60’ module, and black board line; genes
   into WGCNA.

   To elucidate the functions of genes interacting with LINC02257, we
   performed pathway enrichment analysis of the 179 genes in the ‘Grey60’
   module [[122]21]. The top five significantly enriched pathways included
   protein digestion and absorption (P = 1.45 × 10^−13), integrin
   signaling pathway (P = 6.62 × 10^−13), extracellular matrix (ECM)
   receptor interaction (P = 1.01 × 10^−5), AGE-RAGE signaling pathway
   during diabetic complications (P = 2.99 × 10^−5), and Alzheimer’s
   disease-presenilin pathway (P = 6.00 × 10^−5) (Fig. [123]3D). Among the
   significantly enriched pathways, integrin signaling of transmembrane
   receptors promoting cell-ECM adhesion, and ECM receptor interactions
   were functionally related to the prognosis and progression of colon
   cancer [[124]22, [125]23]. We also examined the gene ontology (GO) in
   three categories: biological processes (BP), cellular components (CC),
   and molecular functions (MF) [[126]24]. The most significant GO term in
   each category was ECM (GO:0031012, P = 1.34 × 10^−41) in CC, ECM
   organization (GO:0030198, P = 1.43 × 10^−42) in BP, and ECM structural
   constituents in MF (GO:0005201, P = 1.89 × 10^−26), respectively (Fig.
   [127]3E). Thirteen mRNAs within the ‘Grey60’ module exhibited strong
   correlation with LINC02257 (rho > 0.6, Spearman’s Correlation) and were
   associated with ECM signaling. However, the GWAS did not reveal
   significant SNPs in the genomic regions of the 13 mRNAs (Supplementary
   Fig. [128]S8–[129]9).

   As co-expression modules in bulk tissues may be determined by cell
   composition, we further investigated the estimated cell proportions in
   colorectal tissue to identify the cellular mechanism underlying the
   role of LINC02257 (Fig. [130]3F) [[131]25]. Our results indicated
   positive correlations between LINC02257 expression and the proportions
   of innate lymphoid cells (ILCs), myeloid cells, and malignant cells,
   whereas a negative correlation was observed with epithelial cells
   (Spearman’s correlation, P < 0.05) (Supplementary Fig. [132]S10).
   Considering that during EMT, epithelial cells lose stemness properties
   and reorganize their cytoskeleton for dynamic elongation and motility
   [[133]26, [134]27], our findings indicates that decreased epithelial
   cells might be associated with cancer metastasis, and LINC02257 might
   play an essential role in EMT in cancer progression. Additionally, a
   two-sided t-test was conducted to determine differences in the
   correlation coefficient distributions across each cell type between the
   12,141 genes included in the Weighted Gene Co-expression Network
   Analysis (WGNCA) and the 179 genes in the ‘Grey 60’ module. All cell
   types exhibited significantly different distributions (Fig. [135]3G).
   While the average correlation of the total genes with each cell
   proportion was close to zero, the genes in the ‘Grey60’ module showed
   significant correlations with cell proportions (Supplementary Fig.
   [136]S10).

LINC02257 positively regulates the malignant phenotypes of CRC cells

   The results of the bioinformatic analysis compelled us to investigate
   the function of LINC02257 in the malignant phenotypes of CRC.
   Initially, the expression levels of LINC02257 were assessed in normal
   colon epithelial cells (CCD841) and various CRC cells, revealing that
   LINC02257 was highly expressed in CRC cells compared to normal colon
   epithelial cells (CCD841) according to RT-qPCR results (Fig. [137]4A
   and Supplementary Fig. [138]S11A). The proliferation rate and invasive
   ability of CRC cells were measured to confirm the relationship between
   the expression of LINC02257 and the degree of malignancy of colon
   cancer cells. HCT116 and DLD1 cells exhibiting high expression levels
   of LINC02257 showed a faster growth rate and higher invasive ability
   than HT29 and RKO cells, where LINC02257 was expressed at relatively
   low levels (Supplementary Fig. [139]S11B). These results demonstrate
   that LINC02257 is closely associated with the malignant phenotypes of
   CRC cells. Prior to exploring the functional role of LINC02257 in CRC,
   we examined its cellular localization, which revealed that LINC02257
   was slightly more localized to the nucleus (Fig. [140]4B). Therefore,
   we focused on the cytosolic and nuclear functions of LINC02257.

Fig. 4. LINC02257 is closely associated with malignant phenotypes in DLD1
cells.

   [141]Fig. 4
   [142]Open in a new tab

   A The expression level of LINC02257 in various colorectal cancer cells
   was examined using RT-qPCR. B Cellular fractionation assay was
   performed to determine the localization of LINC02257 in DLD1 cells. C,
   G LINC02257 knockdown (C) or overexpression (G) experiments followed by
   a Transwell assay was performed to examine its role in metastatic
   potential. Invasive and migratory abilities were determined by
   quantifying the number cells that invaded or migrated. D, E, G, H After
   transfection of DLD1 cells as performed above, cell proliferation was
   examined by colony forming assay (D for knockdown; H for
   overexpression) and cell counting (E for knockdown; I for
   overexpression). F, J To investigate the oncogenic role of LINC02257 in
   mouse model, the subcutaneous injection experiments were conducted.
   Tumor size was measured using a caliper, and tumor volume was
   calculated using the following formula: (short length × long
   length × width)/2. Data are expressed as mean ± standard deviation and
   the statistical significance was represented as follows: *p < 0.05,
   **p < 0.01.

   To determine the effects of LINC02257 knockdown on the metastatic and
   proliferative abilities of CRC cells, three independent siRNAs were
   designed (Supplementary Fig. [143]S12A). All the designed siRNAs
   efficiently decreased the expression level of LINC02257 in DLD1 cells
   (Supplementary Fig. [144]S12B). Representative results using LINC02257
   siRNA #1 are shown in the main figure and the all siRNA results are
   provided in the Supplementary Figures. Knockdown of LINC02257 weakened
   the metastatic potential, including the invasive and migratory
   abilities of the DLD1 cells (Fig. [145]4C). All three siRNAs induced a
   similar inhibitory effects on the invasive and migratory abilities of
   DLD1 cells (Supplementary Fig. [146]S13A). In addition to the Transwell
   migration assay, the wound closure assay indicated that DLD1 cells with
   low expression of LINC02257 exhibited decreased migratory activity.
   Knockdown of LINC02257 also decreased the number of colonies (Fig.
   [147]4D and Supplementary Fig. [148]S13B) and reduced the proliferation
   rate (Fig. [149]4E and Supplementary Fig. [150]S13C). These results
   were verified by examining the effect of LINC02257 knockdown in HCT116
   cells. Consistent with the results observed in DLD1 cells, LINC02257
   knockdown inhibited the metastatic and proliferative abilities of the
   HCT116 cells (Supplementary Fig. [151]S14). To investigate the
   long-term effect of LINC02257, we further conducted the subcutaneous
   injection experiment. It showed that the tumor volume of the group
   injected with LINC02257-silenced DLD1 cells was significantly reduced
   compared to negative control group (Fig. [152]4F). There was no
   difference of body weight between two groups.

   Conversely, overexpression of LINC02257 potentiated the invasive and
   migratory abilities of DLD1 cells in a dose-dependent manner (Fig.
   [153]4G and Supplementary Fig. [154]S15). An enhanced metastatic
   potential was also observed in LINC02257-overexpressing HCT116 cells
   (Supplementary Fig. [155]S16A). Overexpression of LINC02257 increased
   the colony-forming ability and accelerated the proliferation rate in
   DLD1 (Fig. [156]4H, I) and HCT116 cells (Supplementary Fig. [157]S16B,
   C). Additionally, we also found that the invasive and migratory
   abilities of RKO cells expressing low levels of LINC02257 were
   augmented by its overexpression (Supplementary Fig. [158]S17). The
   effect of LINC02257 overexpression on tumor formation was examined
   using the subcutaneous injection mouse model. The tumor volume of the
   group injected with LINC02257-overexpressing DLD1 cells was
   significantly increased compared to blank control without a change of
   body weight (Fig. [159]4J). These results demonstrate that LINC02257 is
   responsible for malignant phenotypes such as high metastatic potential
   and rapid proliferation of CRC cells.

SERPINE1 is involved in the regulation of malignancy in CRC by LINC02257

   RNA-seq analysis revealed 343 genes were downregulated in
   LINC02257-silenced DLD1 cells, and 51 genes of them were generally
   upregulated in CRC (Fig. [160]5A and Supplementary Fig. [161]S18A).
   SERPINE1, which encodes plasminogen activator inhibitor-1 (PAI-1), was
   identified as the target gene of LINC02257 by comparing the expression
   levels of 47 coding genes (except for four non-coding genes) and
   analyzing their association with the prognosis of patients with CRC.
   Survival analysis indicated that the expression of SERPINE1 was closely
   associated with the poor prognosis of patients with CRC (Fig. [162]5B
   and Supplementary Fig. [163]S18B). To verify whether SERPINE1 is a
   downstream target of LINC02257, protein and mRNA levels of SERPINE1
   were determined using western blotting and RT-qPCR in
   LINC02257-silenced and overexpressed DLD1 cells. Knockdown of LINC02257
   decreased SERPINE1 protein and mRNA levels (Fig. [164]5C and
   Supplementary Fig. [165]S18C). Conversely, LINC02257-overexpressed DLD1
   cells exhibited increased SERPINE1 protein and mRNA levels (Fig.
   [166]5D). To examine whether SERPINE1 is involved in the control of the
   malignant phenotypes of CRC, the metastatic and proliferative abilities
   of SERPINE1-silenced DLD1 cells were assessed. The transfection of DLD1
   cells with SERPINE1-targeting siRNA significantly decreased SERPINE1
   expression (Fig. [167]5E). SERPINE1 knockdown inhibited the metastatic
   potential, including invasive and migratory abilities, of DLD1 cells
   (Fig. [168]5F), decreased the number of colonies (Fig. [169]5G), and
   slowed cell proliferation (Fig. [170]5H). These results indicate that
   LINC02257 regulates metastatic and proliferative abilities by
   regulating SERPINE1.

Fig. 5. SERPINE1, a target of LINC02257, is involved in malignant phenotypes
of DLD1 cells.

   [171]Fig. 5
   [172]Open in a new tab

   A SERPINE1 was selected as a target of LINC02257 by comparing the list
   of genes downregulated by LINC02257 knockdown and gene that are
   typically upregulated in colorectal cancer (CRC). B Kaplan-Meier
   overall survival (OS) plot of SERPINE1 expression in patients with CRC
   from the Cancer Genome Atlas (TCGA). C, D Knockdown of LINC02257 was
   performed by transfecting DLD1 cells with control or
   LINC02257-targeting siRNA (C). For overexpression of LINC02257, DLD1
   cells were transfected with blank or overexpressing vector (D).
   Expression levels of SERPINE1 protein and mRNA were determined using
   western blot and RT-qPCR analyses, respectively. E–H DLD1 cells were
   transfected with control or SERPINE1-targeting siRNA. Expression levels
   of SERPINE1 protein were determined by western blot (E). Metastatic
   potential including invasive and migratory abilities was determined by
   the Transwell assay (F). Proliferative ability was assessed by a colony
   forming assay (G) and by counting the number of viable cells (H). Data
   are expressed as mean ± standard deviation and the statistical
   significance was represented as follows: *p < 0.05, **p < 0.01.

LINC02257 upregulated SERPINE1 by sequestering miR-1273g-3p

   Given that competitive endogenous RNA (ceRNA) is believed to be one of
   the main action mechanisms of lincRNAs, we searched for
   LINC02257-decoying miRNAs by comparing the lists of downregulated
   miRNAs obtained from miRNA sequencing using total RNA from
   LINC02257-silenced DLD1 cells and SERPINE1-targeting miRNAs using
   TargetScan (Fig. [173]6A and Supplementary Fig. [174]S18D). Using RNA22
   prediction algorithm analysis, 7 of the 34 miRNAs were predicted to
   bind to LINC02257. Because it exhibited the most optimal interaction
   (lowest folding energy), miR-1273g-3p was identified as the decoying
   miRNA of LINC02257. The sequences of SERPINE1 mRNA and LINC02257
   contained two and one miRNA recognition element (MRE) for miR-1273g-3p,
   respectively (Supplementary Fig. [175]S18E). To verify whether SERPINE1
   is a target of miR-1273g-3p, the protein and mRNA levels of SERPINE1
   were assessed in DLD1 cells transfected with miRNA mimic. Western blot
   and RT-qPCR analyses revealed that miR-1273g-3p decreased SERPINE1
   expression (Fig. [176]6B). In addition, the enrichment of SERPINE1 mRNA
   in AGO2 IP was enhanced by miR-1273g-3p (Fig. [177]6C). We additionally
   examined the direct interaction between miR-1273g-3p and SERPINE1 mRNA
   using a luciferase assay. Luciferase vectors containing wild-type or
   mutated sequences of miR-1273g-3p MREs in the 3’UTR of SERPINE1 mRNA
   were constructed (Supplementary Fig. [178]S19A, B). As previously
   observed, two miR-1273g-3p MREs in the 3’UTR of SERPINE1 mRNA exist
   (Supplementary Fig. [179]S18E). Of the two MREs, only MRE#2 exhibited
   an efficient binding of miR-1273g-3p to the 3’UTR of SERPINE1 mRNA
   (Supplementary Fig. [180]S19C). Overexpression of miR-1273g-3p
   inhibited the expression of luciferase containing only wild-type MRE
   (Fig. [181]6D), indicating that the decrease in SERPINE1 by
   miR-1273g-3p resulted from its direct binding to the 3’UTR of SERPINE1
   mRNA.

Fig. 6. miR-1273g-3p, competitively interacting with LINC02257, suppresses
malignant phenotypes in DLD1 cells.

   [182]Fig. 6
   [183]Open in a new tab

   A miR-1273g-3p was identified as a LINC02257-associated miRNA by
   comparing the list of upregulated miRNAs by LINC02257 knockdown and
   SERPINE1-targeting miRNAs. B, C DLD1 cells were transfected with a
   control or miR-1273g-3p mimic. Expression levels of SERPINE1 protein
   and mRNA were determined by western blot and RT-qPCR analyses,
   respectively (B). The enrichment of SERPINE1 mRNA in
   miR-1273g-3p-loaded miRISC was determined using RT-qPCR (C). D The
   sequence-specific binding of miR-1273g-3p to the 3’UTR of SERPINE1 mRNA
   was examined using luciferase vectors containing wild-type and mutated
   sequences of miR-1273g-3p MRE. E, F After DLD1 cells were transfected
   with a control or miR-1273g-3p mimic, the expression level of LINC02257
   was determined using RT-qPCR (E). Direct interaction between LINC02257
   and miR-1273g-3p was assessed by AGO2 RIP. The enrichment of LINC02257
   in miR-1273g-3p-loaded miRISC was determined using RT-qPCR (F). G
   Direct interaction between miR-1273g-3p and LINC02257 was checked by
   luciferase assay using luciferase vectors containing wild-type and
   mutated sequences of miR-1273g-3p MRE in the sequence of LINC02257. H
   Antisense oligonucleotide (ASO) pulldown experiment was conducted to
   check the association of miR-1273g-3p with LINC02257. The level of
   miR-1273g-3p in the pulldown materials was determined using RT-qPCR.
   I–K To prove that LINC02257 interacts with the miR-1273g-3p, DLD1 cells
   were transfected with control or LINC02257-targeting siRNA. The
   expression level of miR-1273g-3p was assessed by RT-qPCR (I). The
   enrichment was determined by measuring the level of SERPINE1 mRNA in
   AGO2 IP (J). Luciferase assay was conducted using the vectors described
   in Fig. 6d (K). L–N DLD1 cells were transfected with a control or
   miR-1273g-3p mimic. Metastatic potential, including invasive and
   migratory abilities, was determined by the Transwell assay (L).
   Proliferative ability was assessed by colony forming assay (M) and by
   counting the number of viable cells (N). Data are expressed as
   mean ± standard deviation and the statistical significance was
   represented as follows: *p < 0.05, **p < 0.01.

   To examine whether LINC02257 interacts with miR-1273g-3p, the level of
   LINC02257 in DLD1 cells transfected with miRNA mimic was assessed using
   RT-qPCR. The overexpression of miR-1273g-3p decreased the level of
   LINC02257 (Fig. [184]6E). Moreover, the RIP assay revealed that
   LINC02257 was more enriched in AGO2 IP than in control IgG IP (Fig.
   [185]6F). Luciferase vectors containing the wild-type or mutated
   sequences of the miR-1273g-3p MREs in LINC02257 were generated
   (Supplementary Fig. [186]S19D). Overexpression of miR-1273g-3p
   inhibited the expression of luciferase (Fig. [187]6G) containing
   wild-type MRE, however, it did not affect luciferase activity in case
   of mutated vector. In addition, the direct interaction between
   LINC02257 and miR-1273g-3p was verified using ASO pulldown assay (Fig.
   [188]6H). The level of LINC02257 was higher in LINC02257-targeting ASO
   pulldown materials than in the control ASO. Next, we investigated
   whether LINC02257 increases the level of miR-1273g-3p, thereby
   increasing the levels of SERPINE1 mRNA in the AGO2 IP. Knockdown of
   LINC02257 elevated the expression levels of miR-1273g-3p (Fig. [189]6I)
   and increased the interaction between SERPINE1 mRNA and the
   miR-1273g-3p-mediated miRNA-induced silencing complex (miRISC) (Fig.
   [190]6J). Moreover, knockdown of LINC02257 suppressed luciferase
   expression, indicating that LINC02257 enhances the expression of
   SERPINE1 by interfering the interaction of miR-1273g-3p with the 3’UTR
   of SERPINE1 mRNA (Fig. [191]6K). To verify the role of LINC02257 as a
   ceRNA of miR-1273g-3p, we selected 138 genes by comparing the genes
   whose expression was downregulated upon LINC02257 knockdown with the
   target mRNAs of miR-1273g-3p (Supplementary Fig. [192]S20A). Among
   them, three genes (SERPINE1, DGKI, and MEF2B) closely related to the
   prognosis of patients with CRC were selected (Supplementary Fig.
   [193]S20B, C) and the effects of LINC02257 and miR-1273g-3p on their
   expression level was measured (Supplementary Fig. [194]S20D). The
   expression levels of the target mRNAs (DGKI and MEF2B) were also
   downregulated by LINC02257 knockdown and miR-1273g-3p overexpression.
   Conversely, an increase in LINC02257 levels resulted in increased
   expression of DGKI and MEF2B.

   Next, we examined whether miR-1273g-3p suppresses the malignant
   phenotypes of CRC. Overexpression of miR-1273g-3p reduced the number of
   invading and migrating DLD1 (Fig. [195]6L) and HCT116 cells
   (Supplementary Fig. [196]S21A). It also inhibited the colony-forming
   ability and reduced the proliferation rate of both DLD1 (Fig. [197]6M,
   N) and HCT116 cells (Supplementary Fig. [198]S21B, C). Based on these
   results, we concluded that LINC02257 positively regulates malignant
   phenotypes by competitively mitigating SERPINE1-targeting miR-1273g-3p.
   Additionally, we tested the correlation of LINC02257, SERPINE1, and
   miR-1273g-3p. RT-qPCR results using the tissues of CRC patients and
   TCGA analyses revealed that there is a positive correlation between
   LINC02257 and SERPINE1 (r = 0.9035, P < 0.001) (Supplementary Fig.
   [199]S22A, B). In stage 2 and 3 CRC tumor tissues, the expression
   levels of miR-1273g-3p were higher than those in normal tissues
   (Supplementary Fig. [200]S22C) and negatively correlated with the
   expression levels of LINC02257 (r = −0.8052, P < 0.001) (Supplementary
   Fig. [201]S22D).

LINC02257 transactivated YB1 by inducing its phosphorylation

   To search for LINC02257-associated RNA-binding proteins (RBPs),
   proteins bound to LINC02257 were isolated by ASO pulldown, separated by
   SDS-PAGE, and analyzed by Nanospray LC/MS/MS (Supplementary Fig.
   [202]S22). Three RBPs, Y-box binding protein 1 (YB1), YB3, and
   heterogeneous nuclear ribonucleoprotein C (HNRNPC), showed significant
   interactions with LINC02257. Among these, YB1, encoded by YBX1, was
   selected because it exhibited the highest binding score. To verify the
   interaction between LINC02257 and YB1, an ASO pulldown assay was
   performed. Prior to examining the binding of YB1 to LINC02257, the
   pulldown efficiency was examined by measuring the level of LINC02257 in
   ASO pulldown materials (Fig. [203]7A). RT-qPCR results showed that the
   ASO designed for LINC02257 was efficient. An ASO pulldown performed
   followed by western blot analysis, verified the direct interaction
   between LINC02257 and YB1. The association between YB1 and LINC02257
   was examined using YB1 RIP experiments (Fig. [204]7B). The level of
   LINC02257 was higher in YB1 IP materials than in control IP. Based on
   these results, we concluded that YB1 is a LINC02257-associate RBP.

Fig. 7. LINC02257 is responsible for the phosphorylation of YB1 to regulate
malignant phenotypes of CRC.

   [205]Fig. 7
   [206]Open in a new tab

   A The efficiency of antisense oligonucleotide (ASO) pulldown was
   assessed by determining the level of LINC02257 in the pulldown
   materials (left). The association of YB1 with LINC02257 was determined
   by western blot (right). B RNA immunoprecipitation assay was conducted
   using a YB1-specific antibody and the level of LINC02257 in IP
   materials was determined using RT-qPCR. C–E DLD1 cells were transfected
   with a control or LINC02257-targeting siRNA. The levels of phospho- and
   total YB1 were determined using western blot (C). The level of GAPDH
   was assessed as loading control. The activation of ERK was assessed by
   western blot using phospho-ERK antibody (D). The level of total ERK was
   checked for loading control. E Cellular localization of YB1 was
   examined by fractionation assay. Lamin B and GAPDH protein were
   determined using western blot to prove that fractionation was
   appropriately done. F The expression of YB1 target genes was determined
   by RT-qPCR in LINC02257- or YBX1-silenced DLD1 cells. G, H Following
   the overexpression of LINC02257, the level of phosphorylated and total
   YB1 was determined by Western blot (G) and the expression of YB1 target
   genes was determined by RT-qPCR (H). I The effects of knockdown of YB1
   target genes on oncogenic phenotypes of DD1 cells were examined by
   Transwell and colony forming assays. J The effect of miR-1273g-3p on
   the phosphorylation of YB1 was examined by Western blot. K–N DLD1 cells
   were transfected with control or YBX1-targeting siRNA. Expression
   levels of the YB1 protein were determined using western blot (K).
   Metastatic potential including invasive and migratory abilities was
   determined by the Transwell assay (L). Proliferative ability was
   assessed by colony forming assay (L) and by counting the number of
   viable cells (N). Data are expressed as mean ± standard deviation and
   the statistical significance was represented as follows: *p < 0.05,
   **p < 0.01.

   We assessed YB1 expression in LINC02257-silenced DLD1 cells. Notably,
   knockdown of LINC02257 affected the level of phosphorylated YB1 without
   altering the total YB1 (Fig. [207]7C). In addition, the activation of
   ERK, a major cellular signal for YB1 phosphorylation, was inhibited by
   knockdown of LINC02257 (Fig. [208]7D). These results indicated that
   LINC02257 is responsible for YB1 phosphorylation via ERK. Given that
   the phosphorylation of YB1 is closely implicated in its function and
   localization, a cellular fractionation experiment was conducted (Fig.
   [209]7E). Compared to the control, the cytosolic level of YB1 was
   higher in LINC02257-silened cells. In contrast, YB1 expression in the
   nucleus was reduced by knockdown of LINC02257. These results were also
   observed by immunofluorescence (IF) staining experiments (Supplementary
   Fig. [210]S24). To verify whether LINC02257 regulates the localization
   of YB1, the expression levels of YB1 target genes were examined (Fig.
   [211]7F). All three YB1 target genes tested (EGFR and MMP1) were
   downregulated by knockdown of LINC02257 or YBX1. In contrast,
   overexpression of LINC02257 increased the level of phosphorylated YB1
   (Fig. [212]7G) and the expression of its target genes (Fig. [213]7H).
   We tested whether YB1 targets including EGFR and MMP1 are involved in
   the control of proliferative and metastatic abilities in DLD1 cells.
   DLD1 cells transfected with EGFR- or MMP1-specific siRNA exhibited
   weakened metastatic potential including migratory and invasive
   abilities (Fig. [214]7I and Supplementary Fig. [215]S25A, B). In
   addition, knockdown of both YB1 targets induced retardation of cell
   growth and inhibited colony forming ability (Supplementary Fig.
   [216]S25C). These results demonstrate that YB1-mediated gene regulation
   is responsible for the oncogenic function of LINC02257. In addition,
   miR-1273g-3p hindered the phosphorylation of YB1, which resulted from
   the decreased LINC02257 expression (Fig. [217]7J).

   To determine whether YB1 is involved in the malignant phenotypes of
   CRC, the metastatic and proliferative abilities of YBX1-silenced cells
   were measured. The YBX1-specific siRNA used in this study efficiently
   suppressed YB1 expression (Fig. [218]7K). Knockdown of YB1 suppressed
   the metastatic potential, including invasive and migratory abilities
   (Fig. [219]7L), reduced the number of colonies (Fig. [220]7M), and
   slowed the proliferation rate (Fig. [221]7N) in DLD1 cells. Similar
   results were observed in the YBX1-silenced HCT116 cells (Supplementary
   Fig. [222]S26). These results indicate that LINC02257 regulates the
   localization of YB1 through its phosphorylation and that the
   transactivation of YB1 plays a significant role in regulating the
   malignancy of CRC by LINC02257.

Discussion

   Although lncRNAs play an important role in cancer progression, only a
   few can be used to treat patients with CRC. It is difficult to identify
   the causal lncRNAs because their functions are governed by complex
   interactions with other genes. Therefore, we systematically
   investigated the TCGA transcriptome data to identify lncRNAs associated
   with CRC malignancy. Among the DE lncRNAs discovered here,
   dysregulation of LINC02418 [[223]6], AFAP1-AS1 [[224]28], BLACAT1
   [[225]29], and MIR31HG [[226]30] has been linked to the growth and
   progression of CRC (Supplementary Table [227]S4). However, we
   prioritized LINC02257 as a colon cancer-associated lncRNA because it is
   over-expressed in primary colon tumors compared to normal tissues, and
   its expression levels correlate with poor survival and MSI in patients
   with CRC. MSI is caused by an abnormality in mismatch repair that
   occurs during DNA replication, resulting in the accumulation of
   mutations. Recently, the mutation burden has been shown to
   substantially increase in CRC compared with normal cells [[228]31], and
   was significantly associated with survival outcomes and response to
   immune therapy [[229]25, [230]32].

   Emerging evidence indicates that lncRNAs function as powerful
   regulators of gene expression. The mechanism by which lncRNAs regulate
   gene expression involves both transcriptional and post-transcriptional
   processes. Recently, the function of lncRNAs as ceRNAs has been
   revealed, attracting the attention of many scientists [[231]33].
   Generally, ceRNAs mitigate the inhibitory function of miRNAs by
   competitively binding to target mRNA. Imbalances in the ceRNA network
   have been identified in various cancers and are closely implicated in
   cancer malignancy [[232]34]. To data, numerous ceRNAs associated with
   tumorigenesis and metastasis of CRC have been reported in CRC.
   LINC00342 promotes the proliferation and metastasis of CRC by
   inhibiting the miR-19a-3p/NPEPL1 axis [[233]35]. LINC01296 enhances CRC
   metastasis by sponging miR-141-3p and accelerating EMT [[234]36].
   LINC00473 facilitates CRC cell proliferation and invasion by targeting
   miR-195 [[235]37]. LINC01569 promotes the proliferation and metastasis
   of CRC via the miR-381-3p/RAP2A axis [[236]38]. We also discovered that
   the tumor-suppressive miR-1273g-3p interacts with LINC02257 as a
   decoying miRNA, thereby regulating SERPINE1 expression. The inhibition
   of CRC proliferation by miR-1273g-3p results from the activation of
   AMPK signaling by suppressing MAGEA3/6 [[237]39]. Several lncRNAs
   function as ceRNAs for miR-1273g-3p and exhibit oncogenic effects in
   various cancers. In glioma and glioma-stem-like cells, lncRNA
   RP11-279C4.1 enhances malignant phenotypes by targeting the
   miR-1273g-3p/CBX3 axis [[238]40]. LINC02418 regulates MELK expression
   by absorbing miR-1273g-3p and serves as a diagnostic marker for CRC
   [[239]41]. SNHG3 lncRNA is responsible for bone metastasis in breast
   cancer through regulation of the miR-1273g-3p/BMP3 axis [[240]42].

   In addition to acting as miRNA sponges, lncRNAs influence the function
   of RBP by acting as ceRNAs. In this study, we identified YB1 as a
   LINC02257-associated RBP. YB1 encoded by YBX1, is a family of RBP
   containing an evolutionally conserved cold-shock domain [[241]43]. It
   is widely recognized as regulator of gene expression processes such as
   transcription, splicing, and translation. YB1 is primarily localized in
   the cytoplasm, but upon phosphorylation at S102, it can translocate to
   the nucleus [[242]44]. High YB1 expression is associated with local
   recurrence and a poor prognosis of patients with CRC [[243]45].
   Moreover, the inhibition of YB1 suppresses the growth and motility of
   CRC [[244]46]. Several lncRNAs serve as ceRNAs for YB1, which regulate
   its stability and localization. HIF1A-AS1 promotes the gemcitabine
   resistance of pancreatic cancer by translationally upregulating HIF1α
   and facilitating the interaction of YB1 to Akt, inducing the
   phosphorylation of YB1 [[245]47]. Myc-activated lncRNA MNX1-AS1
   promotes the CRC progression by directly binding to and stabilizing YB1
   [[246]48]. By interacting with YB1, MIR200HG enhances proliferation,
   invasion, and drug resistance [[247]49]. In breast cancer, the lncRNA
   [248]AC073352.1 increases metastasis and angiogenesis by interacting
   with YB1 [[249]50]. The lncRNA HULC modulates the phosphorylation of
   YB1 by serving as a ERK scaffold protein, thereby enabling the release
   of translational suppression of its bound mRNAs such as cyclin D1,
   cyclin E1, and matrix metalloproteinase 3 [[250]51]. Our results
   revealed that LINC02257 facilitates the phosphorylation of YB1 via ERK
   and induces the nuclear translocation of YB1, where it
   transcriptionally upregulates its target genes such as EGFR, CCNA1, and
   MMP1.

   Molecular analysis revealed that LINC02257 controls CRC malignancy via
   miR-1273g-3p/SERPINE1 and YB1. Furthermore, bioinformatic analyses
   indicated that LINC02257 plays an essential role in EMT during cancer
   progression. EMT is a process by which epithelial cells lose their
   polarity and adhesion, resulting in the acquisition of migratory and
   invasive abilities [[251]52]. The ceRNA network is a well-known
   regulator of the EMT in CRC [[252]53]. SERPINE1, a target gene of
   LINC02257/miR-1273g-3p, is associated with EMT and facilitates cell
   proliferation, migration, and invasion [[253]6, [254]52, [255]54]. It
   is also involved in remodeling of the tumor microenvironment during CRC
   progression [[256]55]. In addition to SERPINE1, YB1 is a downstream
   molecule responsible for the oncogenic function of LINC02257. The
   expression and phosphorylation of YB1 impact the EMT process through
   various mechanisms [[257]56, [258]57]. At the post-transcriptional
   level, YB1 regulates the EMT process by triggering cap-independent
   translation of EMT-promoting genes and suppressing cap-dependent
   translation of growth-promoting genes [[259]58]. Additionally, YB1
   upregulates ZEB1 expression by inhibiting the maturation of
   ZEB1-targeting miR-205/200b, which in turn facilitates EMT [[260]59].
   Natural low-molecular compounds exhibiting anti-proliferative and
   anti-metastatic effects are also being studied as promising targets for
   the treatments for colon cancer [[261]60–[262]62]. In addition, the gut
   microbiome has recently been discovered to play an important role in
   the carcinogenesis and malignancy of gastrointestinal cancer, emerging
   as a new paradigm for the treatment of colon cancer [[263]63].

   Our findings are summarized in Fig. [264]8. The comprehensive
   multilevel bioinformatics indicates LINC02257 is highly expressed in
   CRC and its higher expression is closely associated with a poor
   prognosis in patients with CRC, suggesting that LINC02257 is a
   potential causal lncRNA in CRC. Interestingly, the clinical
   significance of LINC02257 can be found in MSI (approximately 15% of all
   CRC) rather than MSS, which accounts for the majority of colon cancer.
   Compared to the expression level of LINC02257 in MSS CRC, it is
   specifically high in MSI CRC and is more closely related to poor
   prognosis in patients with MSI CRC. Because MSI CRC exhibits a good
   responsiveness to immunotherapy, LINC02257 as an MSI-specific lncRNA,
   might be a promising biomarker for the development of treatments and
   diagnostics for CRC immunotherapy. Knockdown of LINC02257 suppressed
   invasive and migratory abilities and delayed the proliferation of CRC
   cells. Conversely, its overexpression strengthens the metastatic
   potential and enhances proliferation. SERPINE1 was identified as a
   target of LINC02257. Mechanistically, LINC02257 upregulates SERPINE1 by
   sequestering miR-1273g-3p. In addition, LINC02257 regulates the nuclear
   translocation of YB1 by direct interacting and phosphorylating YB1,
   controlling the metastatic and proliferative abilities by LINC02257.
   Taken together, we demonstrate that LINC02257 is a promising
   therapeutic target for CRC treatment.

Fig. 8. Schematic illustration demonstrating the role of LINC02257 in
controlling malignant phenotypes of CRC by interacting with miR-1273g-3p and
YB1.

   [265]Fig. 8
   [266]Open in a new tab

   Details in the main text.

Supplementary information

   [267]Supplementary Methods^ (39.6KB, docx)
   [268]Supplementary Figures^ (12.5MB, pptx)
   [269]Supplementary Tables^ (162.7KB, xlsx)
   [270]Uncropped WB images^ (9MB, pptx)

Acknowledgements