Association between lncRNA GAS5, MEG3, and PCAT-1 Polymorphisms and Cancer Risk: A Meta-Analysis.

PURPOSE: Long noncoding RNAs (lncRNAs) have been widely studied, and single nucleotide polymorphisms (SNPs) in lncRNAs are considered to be genetic factors that influence cancer susceptibility. The lncRNA GAS5, MEG3, and PCAT-1 polymorphisms are shown to be possibly associated with cancer risk. The aim of this meta-analysis was to systematically evaluate this association.
METHODS: Studies were selected from PubMed, Web of Science, Embase, Google Scholar, Cochrane Library, the Chinese National Knowledge Infrastructure (CNKI), and the Chinese Biomedical Literature Database (CBM) through inclusion and exclusion criteria. Odds ratios (ORs) and 95% confidence intervals (CIs) were calculated using the random-effects model or fixed-effects model to assess the association between lncRNA polymorphisms and cancer susceptibility. Metaregression and publication bias analyses were also conducted. All analyses were performed using the Stata 12.0 software.
RESULTS: Sixteen articles (covering 13750 cases and 17194 controls) were included in this meta-analysis. A significant association between SNP rs145204276 and gastric cancer risk was observed (del vs. ins: OR = 0.79, 95%CI = 0.72-0.86; del/del vs. ins/ins+del/ins: OR = 0.74, 95%CI = 0.59-0.91; del/ins vs. ins/ins: OR = 0.84, 95%CI = 0.67-1.05). For rs16901904, a decreased cancer risk was observed in three genetic models (C vs. T: OR = 0.79, 95%CI = 0.70-0.90; CC vs. CT+TT: OR = 0.49, 95%CI = 0.37-0.65; CC vs. TT: OR = 0.49, 95%CI = 0.37-0.66). No statistical significance was found in the metaregression analysis. For all of the included SNPs, no publication bias was found in all genotype models.
CONCLUSIONS: The rs145204276 SNP in lncRNA GAS5 is likely to be associated with gastric cancer risk, whereas the rs16901904 SNP in lncRNA PCAT-1 bears association with a decreased cancer risk.

1. Introduction

Cancer has become a leading cause of death worldwide not only in high-income but also in middle-income countries [1, 2]. Due to the growth of the aging population, the cancer burden is expanding in many countries; there were about 18.1 million new cancer cases and 9.6 million cancer deaths in 2018 [3]. Many investigators have devoted much effort to reducing cancer mortality and morbidity. For example, in clinical therapy, surgical resection is an effective treatment for localized tumors, but the disease could still exhibit potential regional or distant metastasis, or high resistance toward conventional chemotherapy and radiotherapy. Therefore, researchers have focused on the potential molecular biological mechanisms of cancer development, such as genetic biomarkers which can be used as prognostic factors for cancer patients [4].

High-throughput sequencing technologies have identified a great number of noncoding RNAs in the genome, which could be classified into small (18-200 nts) and long (200 nts to >100 kb) noncoding RNAs (lncRNAs) [5]. Although lncRNAs have once been viewed as “transcriptional noise” or “dark matter” in the genome, an abnormal expression of lncRNAs virtually plays a vital role in cancer pathogenesis, such as in cancer initiation, progression, and metastasis [6]. Studies suggested that genetic variation also played important roles in the development of many types of cancer. Single nucleotide polymorphisms (SNPs), a common type of genetic variation, present in lncRNAs have been found to be able to affect the structure, expression level, and function of lncRNAs by interfering with the corresponding target mRNAs [7]. Evidence has demonstrated that SNPs in lncRNA genes were related to the occurrence of various diseases, particularly cancer [8]. Therefore, lncRNA polymorphisms are considered to be potentially associated with the mechanism underlying cancer susceptibility.

Recently, research showed that GAS5 overexpression could inhibit the invasive ability of hepatocarcinoma cells affecting the epithelial-mesenchymal transition (EMT) process, which is very important in early events of the metastatic spread of tumor cells and can make cells more active and invasive [9]. Meanwhile, colorectal cancer (CRC) cell growth and colony formation were also inhibited by the induction of cell cycle G0/G1 arrest and apoptosis [10], increasing casp9 mRNA and pho-Casp9 protein and decreasing extracellular-regulated protein kinases (ERK), Casp3 mRNA, p-Akt, p-ERK, and pho-Casp3 proteins [11]. However, the downregulation of GAS5 expression accelerates depletion of the YBX1 protein and decreases the expression of p21, thereby eliminating G1 arrest to control the proliferation of gastric cancer [12]. A functional 5-base pair (AGGCA/-) insertion/deletion (indel) polymorphism (rs145204276) that exists in the promoter region of lncRNA growth arrest-specific 5 (lncRNA GAS5) has been investigated in multiple cancer types [13-23]. A deletion (del) allele appears to enhance gene transcription activity when compared to an insertion (ins) allele [15]. Zheng et al. performed a two-stage, case-control study to investigate the association between lncRNA GAS5 polymorphisms (rs145204276) and CRC risk, and the results indicated that the del allele of rs145204276 was significantly associated with a 21% decreased risk of CRC [21]. However, logistic regression analysis showed that the deletion allele of rs145204276 significantly increased the risk of hepatocellular carcinoma (HCC) in two independent case control sets (1034 HCCs and 1054 controls) [18].

Furthermore, lncRNAs such as GAS5, H19, MEG3, and TUSC7 play oncogenic or tumor suppressor roles in correlation with tumor suppressor protein P53 or oncoprotein c-Myc, respectively [24]. Balci et al. discovered that PCAT-1, ANRIL, and H19 could inhibit glioblastoma (GBM) cell proliferation, invasion, and migration. Conversely, primary tumors in GBMs displaying tumor progression were characterized by increased MEG3 and HOTAIR expression levels [25]. SNPs in lncRNA prostate cancer-associated lncRNA transcript 1 (PCAT-1), such as rs1902432, rs16901904, rs4871771, and rs710886, have also been investigated in cancer development [26-28]. Yuan et al. investigated the association between lncRNA PCAT-1 rs1902432, rs16901904, rs4871771, and rs710886 and prostate cancer risk in the Chinese population, but only rs1902432 was found to be associated with an increased risk of prostate cancer [28]. In contrast, there was no association between lncRNA PCAT-1 rs1902432 and bladder cancer in the two-stage, case-control study conducted by Lin et al. [27]. Similarly, the published results on the relation between lncRNA maternally expressed gene 3 (lncRNA MEG3) polymorphisms and cancer risk were also diverse [29-32]. Therefore, we conducted a meta-analysis in order to summarize all eligible studies and evaluate the overall relation between cancer risk and lncRNA (GAS5, PCAT-1, and MEG3) polymorphisms.

2. Material and Methods

2.1. Search Strategy

Two investigators independently conducted an electronic literature search (published before April 20, 2019) using PubMed, Web of Science, Embase, Google Scholar, Cochrane Library, the Chinese National Knowledge Infrastructure (CNKI), and the Chinese Biomedical Literature Database (CBM). The restriction in publication languages was selected as English and Chinese. During our search process, the following keywords were used: “lncRNA GAS5,” “lncRNA MEG3,” “lncRNA PCAT-1,” “polymorphism,” “Single nucleotide polymorphism,” “genetic variant,” “cancer,” “tumor,” “malignancy,” “carcinoma,” and “neoplasm.” All clearly irrelevant studies such as case reports and review articles were excluded. The relevant studies cited in the references of review articles were also manually searched for additional eligible studies.

2.2. Study Selection

Two researchers independently evaluated the titles and abstracts of the identified articles to decide whether they met the study criteria. Differences were solved by consensus. Both cross-sectional and longitudinal studies on the relationship between SNPs in lncRNAs GAS5, MEG3, and PCAT-1 and cancer risk were included. Studies were included in this meta-analysis only when they met the following criteria: (1) the study evaluated the association of SNPs in lncRNAs GAS5, MEG3, and PCAT-1 with cancer risk; (2) the study was published in Chinese or English language; (3) the study had an original case-control design or was a cohort-designed study in humans; (4) the study provided sufficient genotyping data to estimate ORs and 95% CIs; (5) genotype frequencies of subjects in controls were in accordance with the Hardy-Weinberg equilibrium (HWE); (6) all cancer cases were diagnosed at any stage by pathology, but only primary tumor could be involved. If there was a parallel publication, we selected the study with a larger sample size.

Studies were excluded if (1) they were not case-control studies; (2) they were case reports, comments, meta-analyses, or review articles; (3) they were a previous study that was duplicated; (4) they were the control groups that did not conform to HWE; (5) there were not enough genotyping data to calculate the ORs and 95% CIs; and (6) they were studies on cell lines and gene expression. If there was only one study investigating one particular lncRNA gene which was not suitable for meta-analysis, it was also excluded. Additionally, when researchers did not report detailed information regarding the genotype distribution in each group, the corresponding authors of the study were contacted for unpublished data.

2.3. Data Extraction

Data extraction was performed independently by two researchers, including the following information from each study: first author, year of publication, ethnicity, cancer type, source of controls, number of cases and controls, genotype distribution of cases and controls, genotyping method, and P value of HWE in controls. Any inconsistencies were resolved by a third researcher until a consensus was reached.

2.4. Statistical Analysis

The present meta-analysis investigates the relationship between lncRNA GAS5, lncRNA MEG3, and lncRNA PCAT-1 SNPs and various types of cancers with ORs and 95% CIs. For each SNP, five genetic models of ORs and 95% CIs (additive model, dominant model, and recessive model as well as homozygous and heterozygous models) were estimated. Subgroup analysis was conducted according to ethnicity, source of controls, and cancer types.

Chi-square test (P > 0.05) was applied to calculate the P value of HWE in control groups. The associations between each SNP and cancer susceptibility were estimated by pooled odds ratios (ORs) and 95% confidence intervals (CIs) under five different genetic models. The heterogeneity of results throughout the studies was assessed using the I2 statistic, which describes the percentage of total variation across the studies that was attributable to heterogeneity rather than to chance. The I2 values of 25% and 75% corresponded to cutoff points for low and high degrees of heterogeneity. The pooled effect was calculated using the random-effects model when the I2 value was >75%. Otherwise, a fixed-effects model was used in case of significant heterogeneity across studies. A sensitivity analysis was performed to evaluate the influence of each individual study on overall estimates. Begg and Egger's tests were used to assess potential publication bias. All statistical analyses were conducted by Stata software version 12.0 (Stata Corp LP, TX, USA). A P value < 0.05 was considered statistically significant.

3. Results

3.1. Characteristics of the Included Studies and Subjects

A total of 365 related studies were preliminarily retrieved after a systematic publication search. After excluding duplicate literatures, we independently read the article abstracts and their references to assess their eligibility for the meta-analysis, as well as the published meta-analyses of the relevant genes. Subsequently, 25 studies were potentially included in the present analysis for further evaluation. After considering the full texts of these articles, nine studies were excluded for the following reasons: two articles did not present control groups; three articles did not have sufficient data; three articles did not conform to HWE; there was only one study about rs55829688 and rs1951625. Finally, 16 articles (13750 cases and 17194 controls) were included in the quantitative synthesis. A flowchart of the study selection process is shown in Figure 1.

Figure 1

A flowchart of the selection process for the included studies.

The characteristics of the studies included in this meta-analysis are presented in Table 1. Among these articles, 11 involved rs145204276 in lncRNA GAS5 [13-23]. Within these 11 articles, 10 investigated Asian and one investigated Caucasian (Iran) ethnicity, and the source of controls was mainly population-based (9 studies). Three studies regarded gastric cancer, and three were involved in CRC. Moreover, Zheng et al. [21] investigated rs145204276 in two independent stages, so we treated the study as two independent investigations. Four articles considered not only rs7158663 but also rs4081134 in lncRNA MEG3 [29-32]. Two articles involved rs1902432, rs16901904, rs4871771, and rs710886 in lncRNA PCAT-1 [27, 28]. The majority of these studies were case-control studies. A variety of genotyping methods, such as polymerase chain reaction-restriction fragment length polymorphism (PCR-RFLP), real-time PCR, TaqMan assay, and MALDI-TOF mass spectrometry, were used in these studies.

Table 1

Characteristics of the studies included in this meta-analysis.

lncRNA	First author	Year	Ethnicity	Source of control	SNPs	Genotyping method	Cancer type	Case	Control	Control HWEP value
GAS5	Jupeng Yuan	2018	Asian	HB	rs145204276	MassArray system	Glioma	404	820	0.14
	Keyvan Aminian	2018	Caucasian	PB	rs145204276	PCR	Gastric cancer	130	230	0.27
	Qianjun Li a	2018	Asian	PB	rs145204276	RT-PCR	Gastric cancer	853	954	0.47
	Qianjun Li b	2018	Asian	PB	rs145204276	RT-PCR	Gastric cancer	1253	1354	0.37
	Ruiyang Tao	2015	Asian	PB	rs145204276	RT-PCR	Hepatocellular carcinoma	1034	1054	0.07
	Yiyin Tang	2018	Asian	HB	rs145204276	PCR	Breast cancer	575	602	0.93
	Yongbin Zheng I	2016	Asian	PB	rs145204276	PCR	Colorectal cancer	600	600	0.76
	Yongbin ZhengII	2016	Asian	PB	rs145204276	PCR	Colorectal cancer	800	800	>0.05
	Weihao Li	2017	Asian	PB	rs145204276	RT-PCR	Lung cancer	600	600	0.07
	Leilei Xu	2018	Asian	PB	rs145204276	RT-PCR	Osteosarcoma	132	1270	0.56
	Zhansheng Zhu a	2016	Asian	PB	rs145204276	RT-PCR	Colorectal cancer	813	926	0.11
	Zhansheng Zhu b	2017	Asian	PB	rs145204276	RT-PCR	Cervical squamous cell carcinoma	920	1018	0.17
PCAT1	Qinbo Yuan	2018	Asian	HB	rs1902432 T>C	TaqMan	Prostate cancer	850	860	0.75
					rs16901904 T>C					0.01
					rs4871771 T>A					0.32
					rs710886 A>G					0.49
	Yadi Lin I	2017	Asian	HB	rs1902432 T>C	RT-PCR	Bladder cancer	578	1006	1.00
					rs16901904 T>C					0.60
					rs4871771 T>A					0.02
					rs710886 A>G					0.12
	Yadi Lin II	2017	Asian	HB	rs710886 A>G	RT-PCR	Bladder cancer	1028	1381	0.77
MEG3	Zhenjian Zhuo	2018	Asian	PB	rs7158663 G>A	RT-PCR	Neuroblastoma	392	783	0.72
					rs4081134 G>A					0.76
	Zitai Yang	2018	Asian	PB	rs7158663 G>A	RT-PCR	Lung cancer	526	526	0.71
					rs4081134 G>A					0.59
	Qi Zhang	2018	Asian	PB	rs7158663 G>A	TaqMan	Gastric cancer	172	224	0.91
					rs4081134 G>A					0.17
	Xiangming Cao	2016	Asian	HB	rs7158663 G>A	MALDI-TOF mass spectrometry	Colorectal cancer	518	517	0.81

Notes: PB: population based; HB: hospital based.

3.2. Meta-Analysis Results

3.2.1. The rs145204276 SNP in lncRNA GAS5 and Cancer Susceptibility

The relationship between the rs145204276 SNP in the lncRNA GAS5 gene and the risk for all types of cancers is shown in Table 2. In the overall analysis, we could not obtain an association between the rs145204276 polymorphism and cancer risk in none of the five genetic models. However, the stratified analysis by cancer type revealed that rs145204276 was associated with a decreased risk of gastric cancer in the del vs. ins genetic model (del vs. ins: OR = 0.79, 95%CI = 0.72‐0.86; del/del vs. ins/ins+del/ins: OR = 0.74, 95%CI = 0.59‐0.91; del/ins vs. ins/ins: OR = 0.84, 95%CI = 0.77‐0.98) (Table 2, Figure 2). A similar result was obtained using the del/del vs. del/ins+ins/ins genetic model and the del/ins vs. ins/ins genetic model (Table 2). Although results showed no significant association in any genetic model between rs145204276 and CRC risk, a decreased risk trend could still be seen in del vs. ins and del/del+del/ins vs. ins/ins genetic models (del vs. ins: OR = 0.96, 95%CI = 0.65‐2.48; del/del+del/ins vs. ins/ins: OR = 0.94), as shown in Table 2 and Figure 2(a). Nevertheless, we did not detect an association between the rs145204276 polymorphism and risks of lung cancer, cervical squamous cell carcinoma, glioma, hepatocellular carcinoma, breast cancer, or osteosarcoma. When the analysis was performed by ethnicity, no significant correlation was found in Asian and Caucasian ethnicities in all genetic models (Table 2, supplementary material). Moreover, we further performed subgroup analysis by the source of controls, and the result indicated that there was also no association between rs145204276 and cancer risk in either the PB or the HB subgroups. However, the function of rs145204276 polymorphism was diametrically different in cancer Stage I+II and Stage III+IV. In detail, an increased risk trend could be seen in all genetic models in cancer Stage I+II, and a decreased risk trend could be seen in all genetic models in cancer Stage III+IV (Table 2, supplementary material).

Table 2

Meta-analysis results of the association between rs145204276 and cancer risk.

Variables	Number of studies	del vs. ins		del/del vs. ins/ins		del/del+del/ins vs. ins/ins		del/del vs. ins/ins+del/ins		del/ins vs. ins/ins
		OR (95% CI)	P	OR (95% CI)	P	OR (95% CI)	P	OR (95% CI)	P	OR (95% CI)	P
Overall	12	0.95 (0.80, 1.13)r	0.54	1.09 (0.75, 1.57)r	0.67	0.97 (0.79, 1.19)b	0.78	1.02 (0.76, 1.38)r	0.88	0.94 (0.80,1.100)r	0.43
Cancer type
Gastric cancer	3	0.79 (0.72, 0.86)r	0.00	0.90 (0.35, 2.31)r	0.82	0.88 (0.53, 1.44)r	0.60	0.74 (0.59, 0.91)a	0.005	0.84 (0.77, 0.98)r	0.03
Colorectal cancer	3	0.96 (0.65, 1.40)r	0.82	0.96 (0.43, 2.16)r	0.93	0.94 (0.60, 1.45)r	0.77	0.99 (0.52, 1.87)r	0.98	0.93 (0.65, 1.34)r	0.71
Others	6	1.06 (0.82, 1.36)r	0.68	1.24 (0.72, 2.14)r	0.93	1.03 (0.76, 1.40)r	0.84	1.25 (0.81, 1.91)r	0.31	1.07 (0.97, 1.17)r	0.16
Ethnicity
Chinese	11	0.98 (0.82, 1.17)r	0.79	1.15 (0.79, 1.68)r	0.47	1.01 (0.82, 1.24)r	0.94	1.06 (0.79, 1.44)r	0.69	0.96 (0.82, 1.13)r	0.63
Iran	1	0.61 (0.42, 0.89)r	—	0.43 (0.17, 1.11)r	—	0.58 (0.37, 3.91)r	—	0.51 (0.20, 1.30)r	—	0.61 (0.38, 0.99)r	—
Source of controls
PB	10	0.92 (0.76, 1.11)r	0.38	1.05 (0.70, 1.57)r	0.80	0.95 (0.76, 1.17)r	0.62	0.99 (0.71, 1.36)r	0.93	0.91 (0.78, 1.08)r	0.28
HB	2	1.10 (0.57, 2.15)r	0.77	1.30 (0.34, 4.97)r	0.71	1.12 (0.49, 2.53)r	0.79	1.23 (0.45, 3.36)r	0.68	1.09 (0.54, 2.20)r	0.82
Cancer stage
Stage I+II	7	1.05 (0.86, 1.28)r	0.62	1.16 (0.80, 1.66)r	0.44	1.06 (0.81,1.37)r	0.68	1.18 (0.89, 1.56)r	0.24	1.01 (0.75 1.36)vr	0.94
Stage III+IV	7	0.75 (0.50, 1.11)r	0.15	0.76 (0.35, 1.61)r	0.47	0.70 (0.44, 1.13)r	0.14	0.94 (0.54, 1.63)r	0.82	0.68 (0.43,1.08)r	0.10

Notes: r: calculating based on a random-effects model.

Figure 2

Forest plots of the association between rs145204276 and cancer risk in subgroup analysis by cancer types under three genetic models. (a) del vs. ins genetic model; (b) del/del vs. del/ins+ins/ins genetic model; (c) del/ins vs. ins/ins genetic model; (d) Begg's funnel plot of rs145204276 in del vs. ins genetic model. OR: odds ratios; 95% CI: 95% confidence intervals.

3.2.2. rs7158663 and rs4081134 SNPs in the lncRNA MEG3 Gene and Cancer Susceptibility

Four of the included studies were evaluated to determine the association between rs7158663 and cancer risk. However, the results showed no significant association in any genetic model between rs7158663 and cancer risk by random-effects models (Figure 3(b)). Three studies were used to detect the association between rs4081134 and cancer risk. Results showed that no significant association was found in any genetic model between rs4081134 and cancer risk by random-effects models. However, a reduced risk trend could be verified (A vs. G: OR = 0.97, 95%CI = 0.71‐1.31; AA vs. GG: OR = 0.86, 95%CI 0.46‐1.63; AG vs. GG: OR = 0.98, 95%CI = 0.73‐1.33; AA+AG vs. GG: OR = 0.97, 95%CI = 0.69‐1.36; AA vs. AG+GG: OR = 0.87, 95%CI = 0.51‐1.47) (Figure 3(a)).

Figure 3

Forest plots of the associations between rs7158663 and rs4081134 and cancer risk in five genetic models. (a) Association between rs4081134 and cancer risk; (b) association between rs7158663 and cancer risk. OR: odds ratios; 95% CI: 95% confidence intervals.

3.2.3. rs1902432, rs16901904, rs4871771, and rs710886 Variations of the lncRNA PCAT-1 Gene and Cancer Susceptibility

For lncRNA PCAT-1 polymorphisms, we mainly focused on investigating the effects of four lncRNA polymorphisms (rs16901904, rs710886, rs4871771, and rs1902432). As shown in Figure 4(a), for rs16901904, decreased cancer risks were observed in three genetic models by fixed-effects models (C vs. T: OR = 0.79, 95%CI = 0.70‐0.90; CC vs. CT+TT: OR = 0.49, 95%CI = 0.37‐0.65; CC vs. TT: OR = 0.49, 95%CI = 0.37‐0.66). Similar associations were found in rs710886, although not statistically significant (GG vs. AA: OR = 0.99, 95% CI = 0.66-1.47; GA vs. AA: OR = 0.94, 95% CI = 0.82-1.07; GG + AG vs. AA: OR = 0.96, 95% CI = 0.79-1.15), shown as Figure 4(b). However, there was no significant association between rs4871771 or rs1902432 and cancer risk in any genetic models by fixed-effects or random-effects models (shown in supplementary material ()).

Figure 4

Forest plots of the association between rs16901904 and rs710886 and cancer risk in five genetic models. (a) Association between rs16901904 and cancer risk; (b) association between rs710886 and cancer risk. OR: odds ratios; 95% CI: 95% confidence intervals.

3.3. Metaregression Analysis

We observed a significant heterogeneity among studies in the meta-analysis on the rs145204276 polymorphism. We hence performed a metaregression analysis to assess the source of heterogeneity in the del vs. ins genetic model (Table 3). The cancer type, source of controls, ethnicity, and genotyping method were detected, but no statistical significance was found. Therefore, when obvious heterogeneity (I2 value > 75%) was observed in the overall or subgroup analyses, the random-effects model was used to make stable confidence intervals. Otherwise, the fixed-effects model was selected to conduct data analysis.

Table 3

Metaregression analysis of the GAS5 rs145204276 on cancer risk (del vs. ins genetic model).

SNP	Variables	Coefficient	Standard error	95% CI	P
GAS5rs145204276	Source of controls	-0.10	0.34	-0.86-0.67	0.79
	Ethnicity	0.55	0.48	-0.49-1.59	0.27
	Cancer type	0.14	0.15	-0.20-0.47	0.38
	Genotyping method	0.11	0.20	-0.32-0.55	0.57

3.4. Publication Bias

Begg's funnel plot and Egger's test were performed to examine publication bias in this meta-analysis. For all included SNPs, no evidence of publication bias was found in all genotype models. One of the publication bias test results is shown as Figure 2(d); in the del vs. ins genetic model, the P value of Begg's funnel plot test was 0.78.

4. Discussion

In recent years, lncRNAs have been confirmed to play important regulatory roles in gene expression, and their aberrant expression has been recognized as a hallmark feature of cancer [33]. SNPs have a wide distribution and can be found in any gene or mRNA regions, and researchers are increasingly paying attention to the important role of lncRNA SNPs in cancer [6]. Thus, it is necessary to identify the association between lncRNA polymorphisms and cancer risk. In this article, we selected all published articles in PubMed, Web of Science, Embase, Google Scholar, Cochrane Library, CNKI, and CBM and included polymorphisms of the lncRNAs GAS5, MEG3, and PCAT-1 in our meta-analysis. The results indicated that two of these lncRNA polymorphisms (GAS5 rs145204276 and PCAT1 rs16901904) may contribute to a decreased cancer risk.

Recently, lncRNA GAS5 has attracted attention as a new type of lncRNA which plays a key role in cancer development. Moreover, the 5 bp indel polymorphism (rs145204276) in the lncRNA GAS5 promoter region has also been studied in detail in different types of tumors. However, we could not obtain an association between the rs145204276 polymorphism and cancer risk in none of the five genetic models in the overall analysis. The results of the subgroup analysis by cancer type revealed that the rs145204276 del allele was associated with a decreased risk of gastric cancer (del vs. ins: OR = 0.79, 95%CI = 0.72‐0.86; del/del vs. ins/ins+del/ins: OR = 0.74, 95%CI = 0.59‐0.91; del/ins vs. ins/ins: OR = 0.84, 95%CI = 0.77‐0.98). Nevertheless, we did not detect an association of the rs145204276 polymorphism with risks for other cancers. Interestingly, the mechanism of action of the rs145204276 polymorphism in cancer cells has also not yet been completely elucidated. On the one hand, researchers observed that the del polymorphism rs145204276 may influence GAS5 transcriptional activity by affecting the methylation status of a CpG island in the promoter region of GAS5, which affects its tumor-suppressing function [34]. The rs145204276 del allele could also induce promoter activity by binding to SP1 and enhancing the expression level of lncRNA GAS5, which results in a higher risk factor for the development of breast cancer [17]. On the other hand, further studies showed that the overexpression of GAS5 induced by indel polymorphism rs145204276 can inhibit the expression of miR-221 and miR-182-5p, thereby reducing the proliferation, migration, and invasion properties of cancer [19]. Altogether, these conflicting results may be explained by the fact that the susceptibility loci are different in different cancer types. The mechanism of action of GAS5 in cancer is still poorly understood, which also requires further studies.

lncRNA MEG3 has been demonstrated to be abnormally expressed in various human cancers, such as bladder cancer, glioma, and gastric cancer [31]. Previous researches revealed that MEG3 markedly inhibited cell growth via the induction of G2/M cell cycle arrest, cell apoptosis, and the reduction of miR-21-5p content in cervical cancer [35]. Braconi et al. found that abnormal expression of MEG3 induced apoptosis in hepatocellular cancer PRC/PRF/5 cells [36]. Another study result showed that the downregulation of lncRNA MEG3 induced nickel malignant transformation of human bronchial epithelial cells through PHLPP1 transcription and HIF-1α translation [37]. Moreover, the SNPs in MEG3 also participate in the development of different types of cancer. For example, Cao et al. discovered that rs7158663 in MEG3 had a strong association with an increased risk of CRC [29]. Studies also demonstrated that rs4081134 in MEG3 was associated with lung cancer susceptibility in a hospital-based case-control study [30]. However, associations between the MEG3 rs7158663 polymorphism and lung cancer susceptibility have not been reported [30]. Neither rs7158663 nor rs4081134 significantly modifies the neuroblastoma risk [31]. As a result, only a reduced risk trend was observed between rs4081134 and cancer in our meta-analysis.

lncRNA PCAT-1 is a long intergenic noncoding RNA transcript from the 8q24.21 region [38]. It has also been suggested as related to various carcinomas, such as gastric bladder [38, 39]. Previous studies confirmed that lncRNA PCAT-1 could promote cell proliferation through binding to polycomb repressive complex 2 (PCR2) as a transcriptional repressor [40]. We also verified that the posttranscriptional silencing of PCAT-1 by miR-215 or PCAT-1 siRNAs significantly inhibited HCC cell proliferation and, conversely, that the inhibition of endogenous miR-215 upregulated PCAT-1 expression and promoted cell viability [41]. Additionally, the restoration of miR-145-5p attenuated the induction effects of PCAT1 on prostate cancer progression [42]. However, SNPs in lncRNA PCAT-1 are varied and also influence susceptibility to different types of cancer. As a result of our analysis, an association with a decreased cancer risk of rs16901904 was observed in three genetic models (C vs. T, CC vs. CT+TT, and CC vs. TT). Similar associations were observed in rs710886, although not statistically significant. This finding still needs to be further studied.

Although significant associations between rs145204276 and rs16901904 and cancer risk were observed in this meta-analysis, our current work still presents several limitations. Firstly, there is a high degree of heterogeneity among the included studies on rs145204276. Although we stratified our analysis by cancer type, ethnicity, and source of controls, between-study heterogeneity was still obvious. Metaregression also revealed that the source of heterogeneity was not the ethnicity, genotyping method, cancer type, or source of controls. Secondly, due to lack of original data of the included studies, the evaluation of associations between age, gender, family history, or environmental factors and cancer risk was not conducted. Thirdly, there was only one study in Caucasian population, and the results of this meta-analysis could only be suitable for the Chinese population. Finally, some sample sizes of the included studies were small, which may influence the results' stability.

5. Conclusions

Our results indicate that rs145204276 in lncRNA GAS5 is a protective factor in the development of gastric cancer, and the role of rs145204276 was also distinct in different cancer stages. In addition, rs16901904 in lncRNA PCAT-1 also plays a protective role against cancer development in Chinese populations. However, several limitations still exist in our analysis, so results should be regarded with caution. Larger and multiple ethnicity studies should be included in our further studies.