Comprehensive analysis of the long noncoding RNA expression profile and construction of the lncRNA-mRNA co-expression network in colorectal cancer

ABSTRACT

Long noncoding RNAs (lncRNAs) have been shown to play important roles in various tumors including colorectal cancer (CRC). Here, we obtained data from RNA-sequencing analysis using 3 paired of CRC tissues and corresponding normal tissues. Through Gene Ontology (GO) and Kyoto Encyclopedia of Genes and Genomes (KEGG) pathway analyses, the biological functions of these dysregulated genes were identified. Moreover, we analyzed the expression levels of lncRNA PGM5-AS1 and B3GALT5-AS1 by quantitative real-time PCR (qRT-PCR) assay. To evaluate the accuracy of the lncRNA-mRNA co-expression network we built, we also detected PGM5 expression and analyzed the relationship between PGM5-AS1 and PGM5 in CRC. In addition, we explored the potential function of PGM5-AS1 in vitro and in vivo. In conclusion, we identified dysregulated lncRNAs and constructed the lncRNA-mRNA co-expression network in CRC. Then, we showed that the expression levels of PGM5-AS1, B3GALT5-AS1 and PGM5 were significantly downregulated in CRC tissues compared with corresponding normal tissues. Besides, PGM5-AS1 expression was positively associated with PGM5 expression. These findings were consistent with our RNA-sequencing data. Functionally, overexpression of PGM5-AS1 could induce cell apoptosis and cell cycle arrest in CRC. Animal study indicated that PGM5-AS1 overexpression inhibited CRC growth in vivo. This work provides dysregulated lncRNAs as candidates for further study in CRC. The lncRNA-mRNA co-expression network brings novel insights into further function research. More importantly, PGM5-AS1 is a critical tumor suppressor in CRC.

Introduction

Colorectal cancer (CRC) is one of the most common malignancies and a leading type of cancer-related deaths worldwide. Over 1.8 million new colorectal cancer cases and 881,000 deaths are estimated to occur in 2018.^1,2 The majority of CRC patients with early stage can be cured by comprehensive therapy including chemotherapy, surgery, radiotherapy and molecular target treatment. However, patients who are diagnosed at advanced-stage usually suffer from treatment failure or even death. In fact, CRC develops by the successive accumulation of oncogenic mutations. Multiple molecules and genes are involved in the process.^3,4 “The Cancer Genome Atlas” (TCGA) database has well provided the genetic makeup of colorectal cancer. It has brought great progression in selecting suitable biomarkers to optimize CRC therapy. Since biomarker-guided therapy and diagnosis have prolonged the overall survival (OS), identifying novel biomarkers to screen CRC patients with early stage is in urgent need. Thus, a more comprehensive understanding of underlying mechanisms of CRC is necessary.

Recently, long noncoding RNA (lncRNA) has been a focus of study in regulating gene expression. LncRNAs, more than 200 nucleotides in length, is a distinct type of noncoding RNAs. With the development of bioinformatics analysis and high-throughput sequencing, a growing number of lncRNAs were reported to act as oncogenes or tumor suppressors.^5–7 For example, CCAT1 and CCAT2, displaying as oncogenes, were found to be independent prognostic factors in CRC.^8,9 Yang et al. reported that HAND2-AS1 was downregulated in endometrioid endometrial carcinoma and inhibited cancer metastasis.¹⁰ Luo et al. showed that the downregulation of lncRNA GAS5 functioned as a tumor suppressor in prostate cancer development and progression.¹¹ Besides, more and more studies have found that dysregulated lncRNAs were involved in tumor proliferation and metastasis, such as PVT1, NORAD, MIR31HG.^12–15

Here, we identified numerous differentially expressed (DE) lncRNAs and mRNAs by RNA-sequencing analysis. Then, we analyzed the co-expressed DEmRNAs to predict the functional roles of DElncRNAs. Gene Ontology (GO) and Kyoto Encyclopedia of Genes and Genomes (KEGG) analyses were also performed for a better understanding of the biological functions of dysregulated genes. To verify our RNA-sequencing results, we detected PGM5-AS1 and B3GALT5-AS1 expression by qRT-PCR. Meanwhile, we investigated the biological function of lncRNA PGM5-AS1 in vivo and in vitro. Our study provides novel candidates for further research and the first evidence that PGM5-AS1 is a critical tumor suppressor in CRC.

Results

Identification of differentially expressed lncRNAs and mRNAs

Based on the results of RNA-sequencing data, we presented all of dysregulated transcripts among different samples in Figure 1(a). We classified lncRNAs into four subtypes, including lincRNA, antisense lncRNA, sense-overlapping lncRNA and sense-intronic lncRNA. A total of 203 lincRNAs, 225 antisense lncRNAs, 615 sense-overlapping lncRNAs and 269 sense-intronic lncRNAs were identified (Figure 1(b)). It is reported that dysregulated expression of genes or transcripts may function as oncogenes or tumor suppressors. Thus, we screened DElncRNAs and mRNAs based on our RNA-sequencing data. A total of 346 lncRNAs and 2593 mRNAs were identified to be differentially expressed in CRC tissues compared with normal tissues (Figure 1(c-d)). Of them, 200 lncRNAs and 1561 mRNAs were upregulated while 146 lncRNAs and 1032mRNAs were downregulated. We respectively listed top 20 dysregulated lncRNAs and mRNAs in Tables 1 and 2. GO analysis revealed that dysregulated mRNAs were significantly enriched in several biological processes, such as receptor mediated endocytosis, aging. Besides, extracellular exosome, cytoplasm and cytosol were linked with cellular component. Protein binding and antigen binding were mainly involved in molecular function. KEGG pathway analysis showed that DEmRNAs were mainly enriched in cell cycle, RNA transport, pathways in cancer (Figure 2 a-b), and most of them were cancer-related.

Figure 1.

Heatmap of the expression profiles of differentially expressed lncRNAs and mRNAs in colorectal cancer. (a): Heatmap of differentially expressed transcripts among different samples. (b): Bar chart showed types and counts of different types of lncRNA. (c): Heatmap of differentially expressed lncRNAs in colorectal cancer. Red color represents up-regulated lncRNAs and blue color represents down-regulated lncRNAs. (d): Heatmap of differentially expressed mRNAs in colorectal cancer. Red color represents up-regulated mRNAs and blue color represents down-regulated mRNAs.

Table 1.

Top 10 upregulated and 10 downregulated lncRNAs in CRC.

Probe Set ID	FC	Regulation	P.Value	Gene_name
ENST00000575404	296.36553	Up	0.0091936	CTD-2529O21.2
ENST00000453149	47.700878	Up	0.0150231	AC147651.5
ENST00000433126	43.712304	Up	0.0339292	SLCO4A1-AS1
ENST00000527021	31.353856	Up	0.0021348	AP006621.9
ENST00000428786	30.62494	Up	0.003194	RP4-669P10.16
ENST00000443493	28.458912	Up	0.0223863	PTGES2-AS1
ENST00000634344	27.888382	Up	0.0030436	AP006222.2
ENST00000444770	26.908377	Up	0.0322923	RP11-170M17.1
ENST00000513626	25.237712	Up	0.0011569	LUCAT1
ENST00000585559	25.075242	Up	0.0141291	AC010525.4
ENST00000427517	−28.06888	Down	0.034264	LINC00892
ENST00000563715	−28.01853	Down	0.046712	RP3-388M5.9
ENST00000604533	−32.94586	Down	0.0110199	RP11-35P15.1
ENST00000613309	−36.00789	Down	0.000145	PGM5-AS1
ENST00000620769	−49.02283	Down	0.003512	RP11-94C24.13
ENST00000622757	−51.54305	Down	9.17E-05	RP5-984P4.6
ENST00000433764	−62.05879	Down	0.0134461	MMP24-AS1
ENST00000585798	−83.48574	Down	0.0001375	RP11-13K12.5
ENST00000489821	−87.71553	Down	0.0042345	B3GALT5-AS1
ENST00000450728	−351.7911	Down	0.0264957	SATB2-AS1

CRC: colorectal cancer; FC: fold change; lncRNA: long noncoding RNA

Table 2.

Top 10 upregulated and 10 downregulated mRNAs in CRC.

Probe Set ID	FC	Regulation	P.Value	Gene_name
ENST00000618833	88687.444	Up	0.0014375	ROBO1
ENST00000431239	36630.173	Up	0.037359	CHD1L
ENST00000221504	13387.137	Up	0.0085691	TRMT1
ENST00000534791	8174.6972	Up	0.0159975	GRINA
ENST00000541507	5893.7224	Up	0.0311456	M6PR
ENST00000423252	5512.662	Up	0.0223786	BANP
ENST00000596378	4444.7875	Up	0.000637	SRRM1
ENST00000581116	4290.2688	Up	0.0478964	THOC1
ENST00000274150	2380.5709	Up	0.0018565	NKD2
ENST00000414542	2095.423	Up	0.0192429	BLCAP
ENST00000394142	−251.9245	Down	0.0110252	CNPY3
ENST00000435420	−287.5533	Down	0.0289862	LDAH
ENST00000405957	−373.0552	Down	0.0146455	LSP1
ENST00000399899	−515.4227	Down	0.0004604	CLDN8
ENST00000396581	−528.5875	Down	0.0477333	TRIM40
ENST00000343070	−541.4281	Down	0.0271033	SCNN1B
ENST00000418842	−631.0441	Down	0.0178075	GCG
ENST00000400269	−955.7149	Down	0.0027612	C20orf96
ENST00000360085	−1454.225	Down	0.0182816	PYY
ENST00000615111	−2125.085	Down	0.0194485	GHR

CRC: colorectal cancer; FC: fold change.

Figure 2.

KEGG pathway analysis of differentially expressed mRNAs. (a): KEGG pathway analysis was used to assess down-regulated mRNAs. (b): KEGG pathway analysis was used to assess up-regulated mRNAs.

Construction of co-expression network and functional analysis of DElncRNAs and DEmRNAs

Recent studies have revealed that lncRNAs might be involved in the regulation of gene expression at the transcriptional, epigenetic, and posttranscriptional levels. In this study, 333 DElncRNAs were confirmed to be co-expressed with 1153 DEmRNAs (Pearson’s correlation coefficient analysis>0.99 or<-0.99). Detailed information was presented in Table S1. We selected several most downregulated lncRNAs to construct a simple lncRNA-mRNA network for further research (Figure 3). Moreover, to obtain an in-depth understanding of the interactions between differentially expressed genes, we also built a lncRNA target pathway network and global signal transduction network (Fig. S1-2).

Figure 3.

Several most downregulated lncRNA-mediated a lncRNA-mRNA co-expression network. Pearson correlation coefficient was >0.99 or<-0.09 with p < .001. blue round rectangles, downregulated lncRNAs; red circles, upregulated mRNAs; blue circles, downregulated mRNAs.

qRT-PCR validation of DElncRNAs in CRC

We chose top 20 downregulated lncRNAs from TCGA cohort and our RNA-sequencing cohort and found that lncRNA PGM5-AS1 and B3GALT5-AS1 were the intersecting lncRNAs between them (Figure 4(a)). TCGA data showed the expression levels of PGM5-AS1 and B3GALT5-AS1 were significantly decreased in CRC tissues compared to normal tissues (Figure 4(b-c)). Similarly, B3GALT5-AS1 and PGM5-AS1 expression were significantly downregulated in CRC tissues compared with normal tissues in our cohort (Figure 4(d-f)). Meanwhile, we analyzed the relationship between clinical pathological parameters and PGM5-AS1or B3GALT5-AS1 expression. We divided the expression levels of PGM5-AS1or B3GALT5-AS1 into two groups based on the median expression level. Our data showed that the group with higher PGM5-AS1 expression showed well differentiation. However, there were no relationships between PGM5-AS1 expression and other parameters (Table 3). No significant relationships were observed between B3GALT5-AS1 expression and clinical pathological parameters (Table S2). These data indicate that PGM5-AS1 may be involved in CRC development.

Figure 4.

Verification for the expression levels of DElncRNAs by qRT-PCR. (a): Intersecting downregulated lncRNAs in RNA-sequencing data and TCGA data. (b): The relative expression levels of B3GALT5-AS1 in colorectal cancer tissues and normal tissues in TCGA cohort. (c): The relative expression levels of PGM5-AS1 in colorectal cancer tissues and normal tissues in TCGA cohort. (d): Relative expression of B3GALT5-AS1 was examined in 25 paired colorectal cancer tissues and corresponding normal tissues by qRT-PCR. (e–f): Relative expression of PGM5-AS1 was detected in colorectal cancer tissues and corresponding normal tissues by qRT-PCR(n = 30).

Table 3.

Clinicopathologic characteristics of PGM5-AS1 expression in CRC patients.

	PGM5-AS1
Characteristics	Low	High	P value
Ages(years)
<60	5	3	0.409
≥60	10	12
Gender
Male	8	6	0.464
Female	7	9
Tumor size(cm)
≤4	6	6	1.0
>4	9	9
Location
Colon	8	7	0.715
Rectum	7	8
Differentiation
Well	0	4	0.032
Moderately and poorly	15	11
Tumor stage
I+ II	8	11	0.256
III-IV	7	4
Lymph metastasis
Yes	7	4	0.256
No	8	11

CRC: colorectal cancer

Validation of the co-expression relationship between PGM5-AS1 and PGM5 in CRC

As mentioned above, we filtered differentially expressed genes and constructed a lncRNA-mRNA co-expression network in CRC by bioinformation analysis. In TCGA cohort, PGM5-AS1 expression was positively associated with PGM5 (Figure 5(a)). Thus, we selected PGM5-AS1 and PGM5 to verify the lncRNA-mRNA co-expression network we built. As expected, PGM5 expression was markedly reduced in CRC tissues compared to corresponding normal tissues in both TCGA and our cohorts (Figure 5(b-d)). We also found a positive correlation between PGM5-AS1 expression and PGM5 expression in our cohort (Figure 5(e-f)). These findings confirm that our lncRNA-mRNA co-expression network is a reliable method to study the functions of dysregulated lncRNAs.

Figure 5.

Validation of PGM-AS1 and PGM5 co-expression relationship in colorectal cancer. (a): Correlation between the expression of PGM5-AS1 and the expression of PGM5 in the colorectal cancer in TCGA cohort. (b): Relative expression of PGM5 in colorectal cancer tissues and normal tissues in TCGA cohort. (c–d): qRT-PCR analysis of PGM5 expression in our cohort. The results are presented as the fold change in tumor tissues relative to the matched adjacent normal tissues. Red bars represent T/N > 1. (e–f): Correlation between PGM5-AS1 and PGM5 expression in our cohort (n = 30).

Overexpression of PGM5-AS1 inhibits CRC cell proliferation in vitro

In order to investigate the biological function of PGM5-AS1 in CRC, we examined PGM5-AS1 expression in two human colorectal cancer cells (HCT116, HT29) and one normal colonic epithelial cell (NCM460) by qRT-PCR. The expression levels of PGM5-AS1 were reduced in HT29 and HCT116 cell lines compared to NCM460 cell line (Figure 6(a). pEX-3-PGM5-AS1 or empty vectors were transfected into HCT116 and HT29 cells, and qRT-PCR was used to demonstrate PGM5-AS1 overexpression (Figure 6(b-c)). CCK8 and colony formation assays indicated that overexpression of PGM5-AS1 inhibited cell growth in HCT116 and HT29 cells (Figure 6(d-e)). However, transwell assay showed no significant differences of the number of migration cells between pEX-3-PGM5-AS1 groups and empty vector groups (Figure 7(a-b)). Similarly, western bolt assay showed no significant differences of E-cadherin and N-cadherin expression between pEX-3-PGM5-AS1 groups and empty vector groups (Figure 7(c)). These findings suggest that PGM5-AS1 inhibits the capacity of cell proliferation in CRC.

Figure 6.

PGM5-AS1 inhibits cell proliferation in colorectal cancer. (a): PGM5-AS1 expression in two CRC cell lines compared with the normal colon epithelium cell line NCM460 by qRT-PCR. (b–c): Overexpression of PGM5-AS1 in colorectal cancer cell lines (HCT116 and HT29) was confirmed by qRT-PCR. (d): The colony formation assay was used to test the effect of PGM5-AS1 overexpression on cell proliferation in colorectal cancer. (e): PGM5-AS1 overexpression inhibited cell proliferation by CCK8 assay.

Figure 7.

PGM5-AS1 has no significant influence on cell migration and invasion in colorectal cancer. (a): The transewell assay was used to evaluate the capacity of cell migration and invasion after overexpression of PGM5-AS1 in HT29 cells. (b): The transewell assay was used to evaluate the capacity of cell migration and invasion after overexpression of PGM5-AS1 in HCT116 cells. (c): Western bolt assay showed the expression levels of E-cadherin and N-cadherin in PGM5-AS1 overexpressing cells.

Overexpression of PGM5-AS1 induces cell apoptosis and affects cell cycle in CRC

Flow cytometric analysis showed apoptotic cells were increased in pEX-3-PGM5-AS1 groups compared to empty vector groups (Figure 8(a)). Additionally, the expression levels of apoptosis-related proteins (Bax and Cleaved-caspase3) were increased while Bcl-2 expression was decreased in pEX-3-PGM5-AS1 group (Figure 8(b)). These data demonstrate that PGM5-AS1 induces cell apoptosis in HT29 and HCT116 cells. To uncover the potential mechanism of PGM5-AS1 on CRC cell proliferation, we detected the number of cells in G0-1, S and M phase by flow cytometry analysis. Our results showed that more cells were in G1/G0 phase and fewer cells were in S phase of the cell cycle in cells transfected with pEX-3-PGM5-AS1 (Figure 8(d)). Thus, we speculated that PGM5-AS1 affected cell proliferation by modulating cell cycle in CRC. Besides, PGM5-AS1 overexpression reduced the protein expression levels of CDK4 and cyclinD1(Figure 8(c)), which indicated that PGM5-AS1 affected cell cycle by inducing G0/1 phase arrest.

Figure 8.

PGM5-AS1 overexpression induces cell apoptosis and cell cycle arrest. (a): The apoptotic rates of cells were detected by flow cytometry analysis. (b–c): Western blot analysis of apoptosis-related and cell cycle-related proteins after empty vector or pEX-3-PGM5-AS1 transfection in HT29 and HCT116 cells. (d): Flow cytometry showed the number of cells in G0-1, S, G2-M phases in PGM5-AS1 overexpressing cells.

PGM5-AS1 suppresses CRC cell growth in vivo

To further evaluate the role of PGM5-AS1 in vivo, we subcutaneously injected HT29 cells with PGM5-AS1 stable overexpression or negative controls. Consistent with the in vitro results, mice in PGM5-AS1 overexpression groups showed a significant decrease in tumor volume and tumor weight compared with negative control groups (Figure 9(a-c)). Tunel assay showed more apoptotic cells in PGM5-AS1 overexpression groups than negative controls. Furthermore, IHC assay confirmed that ki67 expression was decreased in PGM5-AS1 overexpression tumors than control tumors (Figure 9(d)). Taken together, these data reveal that PGM5-AS1 suppresses tumor growth as a tumor suppressor.

Figure 9.

PGM5-AS1 overexpression reduces CRC growth in vivo. (a): HT29 cells with stable PGM5-AS1 overexpression and control cells were subcutaneously into nude mice. (b–c): PGM5-AS1 overexpression markedly inhibited tumor growth in nude mice, and both the tumor weight and tumor volume were significantly reduced in the pEX-3-PGM5-AS1 group compared to the empty vector group. (d): PGM5-AS1 overexpression decreased ki67 expression and increased the number of apoptotic cells in vivo.

Discussion

Noncoding RNAs including lncRNAs, miRNAs were reported to play crucial roles in regulating gene expression. LncRNAs may modulate gene expression by acting as enhancer RNA,¹⁶ sponging microRNA,^17,18 or binding to target genes to recruit chromatin-modifying enzymes.¹⁹ Thus, lncRNAs affect cell proliferation, cell migration or genomic stability.^20–23 Recently, a growing number of lncRNAs have been demonstrated to be involved in tumor progression.^24,25 Damas ND et al. reported that SNHG5 promotes colorectal cancer cell survival by counteracting STAU1-mediated mRNA destabilization.²⁶ Wang P et al. demonstrated that the STAT3-binding Long noncoding RNA lnc-DC controls human dendritic cell differentiation.²⁷

Although a growing body of evidence has showed that dysregulated lncRNAs played key roles in CRC, the detailed functions and underlying mechanisms remain still largely unknown. In the present study, we utilized GEO, KEGG, TCGA databases for bioinformatics analysis. We identified DElncRNAs and mRNAs, and some of them were novel transcripts. Among the top 20 dysregulated lncRNAs, only three of them have been reported to be involved in tumor progression. B3GALT5-AS1 is an antisense lncRNA, located on chromosome 21q22.2. In this study, we found that B3GALT5-AS1 expression was downregulated in colorectal cancer tissues compared with corresponding normal tissues. Similarly, Wang L et al. reported that lncRNA B3GALT5-AS1 was reduced in primary colorectal cancer compared with normal colonic epithelium,²⁸ in accordance with our results.

PGM5-AS1 is also an antisense lncRNA, which is located on chromosome 9q21.11. Zhu hu et al. reported that PGM5-AS1 was significantly upregulated in stage III CRC compared to stage II CRC samples.²⁹ However, Pang et al. demonstrated that PGM5-AS1 expression was higher in patients with I/II stage than III/IV stage and high expression of PGM5-AS1 was associated with favorable OS in the pan-cancer data from TCGA.³⁰ We systematically analyzed PGM5-AS1 expression in our cohort and demonstrated that PGM5-AS1 expression was significantly decreased in CRC tissues. However, no significant differences of PGM5-AS1 expression were observed between CRCs with stage I/II and CRCs with stage III/IV. Interestingly, we showed that PGM5-AS1 expression was significantly higher in well differentiation group than moderately or poorly differentiation group. Taken together, a larger sample size is needed to validate the clinical significance of PGM5-AS1.

Our lncRNA-mRNA co-expression network is an important method to predict the function of lncRNAs. To obtain a more reliable conclusion, we chose DElncRNAs and DEmRNAs with the correlation coefficient >0.99 or <-0.99. Moreover, we selected PGM5-AS1 and PGM5 to further validate the co-expression network we built. Our qRT-PCR results indicated that there was a markedly positive relationship between PGM5-AS1 and PGM5. Uzozie AC et al. reported that PGM5 is downregulated in adenomas and/or adenocarcinomas, as compared with normal mucosa.³¹ Moreover, they showed that PGM5 was an indicator to screen colorectal adenoma. Therefore, we speculated that PGM5-AS1 might be a novel predictive marker in CRC. Thus, comprehensive consideration of PGM5-AS1 and PGM5 expression may be a better choice to early screening of CRC.

Another major finding of the present study is the first to report PGM5-AS1 inhibited tumor proliferation in colorectal cancer. As PGM5-AS1 expression was decreased in colorectal cancer compared to corresponding normal tissues, we explored the potential biological function of PGM5-AS1. In vivo and in vitro experiments showed that PGM5-AS1 overexpression inhibited tumor growth in CRC. PGM5-AS1 exerted functions by inducing cell apoptosis and G0/1 phase arrest in CRC, indicating that PGM5-AS1 might be a tumor suppressor in CRC. However, several limitations in this study require further improvements, such as lack of detailed mechanisms of PGM5-AS1. Further clinical studies are required to validate the clinical value of PGM5-AS1 in early diagnosis of CRC.

Taken together, our findings provide a great number of DElncRNAs and DEmRNAs as novel candidates for further study. The lncRNA-mRNA co-expression network provides a better understanding of potential functions of differentially expressed genes. More importantly, we demonstrated that PGM5-AS1 is a novel tumor suppressor factor in CRC.

Materials and methods

Clinical specimens

A total of 33 patients with CRC who underwent surgery at Nanjing Drum hospital affiliated with Medical school of Nanjing university were recruited in this study. None of them received chemotherapy or radiotherapy. Written informed consent was obtained from all patients. 3 pairs of CRC tissues and corresponding normal tissues were used to RNA-sequencing analysis. Additional 30 pairs of tissue samples (CRC tissues and corresponding normal tissues) were performed for qRT-PCR assay.

qRT-PCR and RNA-sequencing analysis

Trizol reagent (Invitrogen, Carlsbad, CA) was used to total RNA extraction according to the manufacturer’s instructions. Total RNA was reverse transcribed into cDNA using Thermo Scientific™ EP0733 1st Strand Synthesis Kit. GAPDH was used as an internal control. The PCR primers were as follows:

forward5ʹ TCATGCAGTTCCCATTTTACAG3ʹ (PGM5-AS1),

reverse5ʹ CCAGCAGGTTTCAACAGACG 3ʹ (PGM5-AS1).

forward5ʹGCTGGCTTCTGTGCTTCGTCT3ʹ(B3GALT5-AS1),

reverse5ʹGATGGGCCATGAACCAGGG3ʹ (B3GALT5-AS1).

forward5ʹACCCCAGCTGGATGGAGATT3ʹ(PGM5),

reverse5ʹTCCGGGCAGCAATAATGGAG3ʹ (PGM5).

forward5ʹTGGGTGTGAACCATGAGAAGT3ʹ(GAPDH),

reverse5ʹTGAGTCCTTCCACGATACCAA3ʹ (GAPDH).

qRT-PCR assays were performed on ABI7900. Each sample was analyzed in triplicate and comparative cycle threshold (CT) (2− ΔΔCT) method was conducted to analyze the data.

RNA-sequencing analysis was performed in the laboratory of Sangon Biotechnology Company (Shanghai, China). Briefly, RNA labeling and hybridization array were conducted according to the manufacturer’s protocol. Libraries were controlled for quality and quantified using Hiseq 2500 system (Illumina, USA). The read counts were expressed as FPKM. DElncRNAs and DEmRNAs were identified based on the student’s t-test for comparison of the two groups. A lncRNA or mRNA with a fold change (FC) > 2.0 and a P value < .05 was considered as a DElncRNA or DEmRNA

GO and KEGG analysis

GO analysis was used to construct gene annotations and analyze the primary function of the DEmRNAs, which can uncover the gene regulatory network based on biological process and molecular function. Besides, we put DEmRNAs into KEGG database to find out the significant pathways of the differentially expressed genes. P < .05 was considered as significant.

Co-expression network of lncRNAs with mRNAs

According to the normalized signal intensity of specific expression in mRNAs and lncRNAs, a lncRNA-mRNA co-expression network was built to identify the relationships between lncRNAs and mRNAs. Genes with a fold change>2 and P < .05 were considered to be differentially expressed. We chose DElncRNAs and DEmRNAs to build the network by the Pearson correlation coefficient>0.99 or<-0.99 with P < .001

LncRNA target pathway network and global signal transduction network

lncRNA target pathway network was built based on the correlations of significant pathways and lncRNAs, as well as the relationships among lncRNAs and pathways. We evaluated the regulatory status of lncRNAs and pathways with the degrees of lncRNAs and pathways in the network as the evaluation criteria. Key lncRNAs and pathways in the network had the biggest degrees. Global signal transduction network was constructed to identify genes which played crucial roles in this network. Signal transduction network (Signal-net) was used to uncover comprehensive interactions between the differentially expressed genes. Genes with higher degrees represent they have a more crucial position in the network.

Cell lines and cell culture

Two CRC cell lines (HT29, HCT116) and normal colon epithelial cell line NCM460 were purchased from Type Culture Collection Center of Chinese Academy of Science (Shanghai, China). All cell lines were cultured according to the manufacturer’s instructions.

Plasmid construction and cell transfection

Plasmid vectors with pEX-3-PGM5-AS1 were obtained from GenePharma (Shanghai, China) for PGM5-AS1 over-expression and empty vectors were used as negative controls. HCT116 and HT29 cells were transfected with plasmid vectors (pEX-3-PGM5-AS1 or empty vectors) by Lipofectamine 3000 (Invitrogen) according to the manufacturer’s instructions. Cells were harvested for RT-PCR or western blot assay at 48h post transfection.

CCK8 assay

3 × 10³ cells were collected and seeded in a 96-well plate at 37°C. Then,10μL CCK-8 solution was added to each well and cells were counted daily for 4 consecutive days. Plates were incubated at 37°C for 2h. Finally, the spectrophotometric absorbance at 450 nm was measured for each sample.

Colony formation assay

1 × 10³ HT29 or HCT116 cells were harvested and cultured in 6-well plates in a humidified atmosphere containing 5% CO2 at 37°C for 2 weeks. Cell colonies were washed with normal saline (NS), fixed with methanol, and stained with 0.1% crystal violet (1mg/mL). Colonies containing more than 50 cells were counted and the mean colony numbers were calculated.

Flow cytometric analysis

Apoptosis and cell cycle assays were detected according to the instructions. Briefly, cells were harvested and washed by NS. Then, cells were resuspended in 100μL of annexin V–Fluos labeling solution per sample, and incubated in the dark for 5 minutes at room temperature. For cell cycle analysis, cells were washed and resuspended in 500μL of propidium iodide-staining solution per sample, and incubated in the dark at room temperature for 30 minutes. The samples were analyzed by flow cytometry. Cells in G0–G1, S, and G2–M phases were counted and compared.

Migration and invasion assays

For cell migration assay, migration chambers (8-mm pore) (Corning, Corning, NY. USA) placed into a 24-well plate were used in the assays. 2 × 10⁵ cells were seeded on the upper insert while medium with 20% FBS was added in the lower chamber as chemoattractant. After incubated for 24 h, the insert was fixed, stained and counted. For cell invasion assay, matrigel (Corning) was added to the transwell membranes, the rest of steps were the same as cell migration assay.

Immunohistochemistry (IHC) assay

Tissue sections were deparaffinized in xylene and rehydrated with ethanol. Then, tissue sections were incubated with the primary antibody targeting ki67 (Abcam, 1:500) at 4°C overnight, and then were incubated with corresponding secondary antibody.

Western blotting analysis

Cells were lysed with RIPA extraction reagent (Beyotime). Extracted proteins were separated by 10% sodium dodecyl sulfate polyacrylamide gel electrophoresis (SDSP-AGE), and transferred to polyvinylidene fluoride (PVDF) membranes (Millipore). Then, PVDF membranes were blocked with 5% nonfat milk and incubated with primary antibodies at 4°C overnight. The following primary antibodies were used in this study: β-actin (CST,1:1000), CDK1 (CST,1:1000), CDK4 (CST,1:1000), CyclinD1 (CST,1:1000), Bax (CST,1:1000), Bcl-2 (CST,1:1000),Cleaved- caspase3(CST,1:1000), E-cadherin(CST,1:1000), N-cadherin(CST,1:1000). Then, membranes were incubated with secondary antibody for 1h at room temperature. Signals were visualized by ECL reagent.

Animal model

All animal experiments were performed under the experimental animal according to guidelines of the National Institutes of Health. Male BABL/c athymic nude mice (aged 5–6 weeks) purchased from the Experimental Animal Center of Nanjing Drum Tower Hospital. Lentivirus carrying LV5-PGM5-AS1or empty vectors were purchased from GenePharma (Shanghai, China). 1 × 10⁶ HT29 cells were stably transfected with LV5-PGM5-AS1 or empty vectors and selected by puromycin (Sigma, MI, USA), then they were inoculated subcutaneously into mice. The tumor volumes and weights were measured every 5 days. Mice were sacrificed after 20 days and tumor tissues were examined by H&E staining, IHC and Tunel tests. Tumor volumes were calculated by the following equation: V = 0.5 × D× d²(D: the longest diameter, d: diameter perpendicular to the longest diameter).

Statistical analysis

Data were analyzed using SPSS 17.0 software. In RNA-sequencing analysis, genes with FC>2 and P < .05 were considered significant. The significance of differences between groups was estimated by the Student t-test, χ2 test. Pearson’s correlation coefficient was used to evaluate the relationship between the expression levels of PGM5-AS1 and PGM5. Differences were considered significant if p < .05: *p < .05; **p < .01; ***p < .001.

Funding Statement

This work was supported by grants from Provincial Natural Science Foundation of Jiangsu [BK20161107]; Nanjing Science and Technology Development Project [201501013]; Jiangsu Health and Family Planning Commission development program [no H2017042] and Nanjing Health and Family Planning Commission key program [no ZKX17012]. Our research was supported by the Innovation Capability Development Project of Jiangsu Province [No. BM2015004], and National Human Genetic Resources Sharing Service Platform [2005DKA21300]; Nanjing Health and Family Planning Commission medical science technology innovation platform project [ZDX16006].

Disclosure of Potential Conflicts of Interest

No potential conflicts of interest were disclosed.

Supplementary material

Supplemental data for this article can be accessed on the publisher’s website.