Joint analysis of lncRNA m⁶A methylome and lncRNA/mRNA expression profiles in gastric cancer

N⁶-methyladenosine (m⁶A) is the most predominant internal chemical modification of messenger RNAs (mRNAs) in eukaryotes [1, 2]. It has been implicated in all aspects of post-transcriptional RNA metabolism, including half-life, splicing, translational efficiency, nuclear export and RNA structure [3]. The widespread nature of m⁶A in human transcriptomes has attracted a huge interest owing to technological advances in sequencing. The exploration on methylation patterns in tissue and cells could not only reveal the specific distribution of m⁶A modification in numerous transcripts, but also uncover the differences in m⁶A status under physiological and pathophysiological conditions [4]. Therefore, in-depth knowledge of m⁶A methylome profile has great benefit for elucidating the development of various human diseases especially malignancy [5, 6].

In recent years, the regulatory roles of m⁶A methylation in gastric cancer (GC) has been paid increasing attention, which is the fifth most common cancer worldwide and the third leading cause among cancer-related deaths [7]. The m⁶A modification might be closely associated with GC genesis and progression [5, 8]. Previously, several studies have investigated some m⁶A or expression patterns of m⁶A-related genes in GC. Zhang et al. evaluated the m⁶A modification patterns of GC samples based on 21 m⁶A regulators and their correlation with tumor microenvironmental (TME) cell infiltration [9]. Another research conducted by Guan et al. analyzed the expression status and determinate prognostic values of m⁶A-related genes in GC [10]. Nevertheless, all these reports were supported by complete bioinformatics methods utilizing the public data of GC patients from the Cancer Genome Atlas (TCGA) and Gene Expression Omnibus (GEO) database. To date, a dataset established by high-throughput assay for m⁶A methylome and expression profiles has been still lacking. Moreover, nearly all studies for m⁶A and GC were focused on protein-coding genes. The field of GC-related m⁶A methylation based on non-coding RNAs (ncRNAs), however, remains relatively blank.

Long non-coding RNAs (lncRNAs), generally defined as transcripts longer than 200nt, comprise the majority of ncRNAs [11]. They play critical roles in chromatin organization, transcriptional and posttranscriptional regulation [12, 13]. Similar to mRNAs, lncRNAs are also modulated by m⁶A and the levels of m⁶A residues strongly depend on the cell line, tissue type and growth condition [1, 14, 15]. According to the methylated sites located on lncRNAs, m⁶A might affect their biosynthesis, secondary structure and thus biological function for tumorigenesis [16, 17]. Several lncRNAs involved in different types of cancer were shown to simultaneously acquire dynamic m⁶A modification within their structures, such as XIST, MALAT1 and HOTAIR [18]. It has been reported that ALKBH5, a demethylase mediating methylation reversal, could promote GC invasion and metastasis by demethylating the lncRNA NEAT1. In spite of this, the evidence for m⁶A-reglulated lncRNAs associated with GC is very limited. It remains unclear what the overall patterns of lncRNA m⁶A methylation in GC are like, whether they have differences with normal status, and how they influence downstream molecules and participate in gastric carcinogenesis and progression.

In the present study, the lncRNA m⁶A methylome was established based on tissue samples to identify the differentially m⁶A-methylated lncRNAs in GC. Meanwhile, the expression profiles of lncRNA/mRNA were also designed to further explore the potential function of m⁶A-regulated lncRNAs involved in GC initiation. The study aims to provide novel clues for the disclosure of epigenetic etiology and pathogenesis of GC related to lncRNA m⁶A methylation.

LncRNA modification is a hot emerging field in cancer epigenetics with rapidly expanding interest. Here, we presented a comprehensive identification of differentially m⁶A-methylated lncRNAs in GC via MeRIP-Seq. Combined with gene expression profiling, four dme-lncRNAs were discovered including RASAL2-AS1, LINC00910, SNHG7 and LINC01105. Co-expression analysis and gene enrichment analysis were subsequently performed to explore their potential target genes and related function. Finally, the targets were selected and validated by CCLE database and literature review. To the best of our knowledge, this study firstly established the lncRNA m⁶A methylome by means of high-throughput assay and reported four dme-lncRNAs in GC. It is also the first time to illustrate the regulatory roles of differential m⁶A in lncRNAs with further impacts on the biological behaviors of GC cells.

Currently, m⁶A-centred RNA methylation has been well accepted to have close relationship with tumorigenesis including GC. For instance, a vitro experiment proved that m⁶A suppression promoted GC cell proliferation and invasiveness through activating Wnt and PI3K-Akt signaling, while m⁶A elevation reversed these phenotypical and molecular changes [8]. Another research claimed that the level of m⁶A in peripheral blood RNA was a promising noninvasive diagnostic biomarker for GC patients [5]. Therefore, the exploration for m⁶A patterns would deepen our insights into RNA posttranscriptional regulatory network participating in the complex biological processes implicated in cancer. In contrast to mRNAs, m⁶A residues in lncRNAs are distributed along the whole body of transcripts and are more concentrated in the lncRNAs undergoing alternative splicing [15]. In 2014, Batista et al. mapped m⁶A methylome in mouse and human embryonic stem cells and thousands of lncRNAs showed conserved m⁶A modification, suggesting that m⁶A was a mark of transcriptome flexibility required for stem cells to differentiate to specific lineages [23]. Moreover, Xiao et al. generated 21 whole-transcriptome m⁶A methylomes across major fetal tissues and reported that m⁶A were enriched in enhancer long intergenic non-coding RNAs (lincRNAs) [15]. Interestingly, this outcome was also indicated in our methylome study. The lincRNAs and exon-derived lncRNAs were shown to enrich m⁶A in GC, and similar trend also occurred in the 191 differentially methylated lncRNAs. It has been revealed that tissue m⁶A regions may preferentially occupy genes with single nucleotide polymorphisms (SNPs) and CpG-rich promoters, and genetic or epigenetic variation at promoters was widely associated with cancer [15, 24]. Hence, it is worth further verification whether the m⁶A modifications of differentially methylated lncRNAs in GC are regulated by these factors.

Dynamic RNA modifications are often enriched for quantitative traits and complex traits including common diseases, and thus m⁶A is potentially correlated with gene expression homeostasis [25, 26]. For instance, METTL3-mediated m⁶A modification led to LINC00958 upregulation through stabilizing its RNA transcript, and LINC00968 sponged miR-3619-5p to upregulate HDGF expression thereby facilitating the lipogenesis and progression of hepatocellular carcinoma (HCC) [27]. Another investigation revealed that m⁶A demethylase ALKBH5 could suppress the degradation of lncRNA PVT1, and its overexpression promoted osteosarcoma cell proliferation in vitro and tumor growth in vivo [28]. Similar phenomenon could also be observed in GC-related lncRNAs with m⁶A [29]. In our research, four dme-lncRNAs were newly found to have both significant m⁶A hypermethylation and differential expression levels in GC compared with normal tissue. Among them, RASAL2-AS1, SNHG7 and LINC01105 have been preliminarily studied so far, except LINC00910. RASAL2-AS1 (RASAL2 antisense RNA 1), located in chromosome 1q25.3, is a natural antisense lncRNA with 2485nt length. It has only been referred to in a bioinformatics analysis based on TCGA database for the prognostic implications of aberrantly expressed methylation-driven genes in HCC [30]. The methylation degree of RASAL2-AS1 was included in the calculation of prognostic risk score for HCC, suggesting that it could be a functional m⁶A-regulated lncRNA in carcinoma. SNHG7 (small nucleolar RNA hostgene 7) is an intergenic lncRNA located in chromosome 9q34.3 with 3590nt length, which is a novel vital oncogenic lncRNA [31]. Accumulating studies have demonstrated the association of SNHG7 with multiple human cancers via complicated mechanisms. It was found that the relative expression of SNHG7 was up-regulated in GC tissues and cells, and partially contributed to GC development and progression through regulating the expression of p15 and p16 [32]. SNHG7 was also shown to accelerate cell migration and invasion through regulating miR-34A-Snail-EMT axis in GC [33]. However, few investigations have addressed the m⁶A modification in this lncRNA yet. As for LINC01105, also named ‘SILC1’, is an exon sense overlapping lncRNA located in chromosome 2p25.2 with 47531nt length. The study for LINC01105 has been limited in its role as an oncogene of neuroblastoma. It could influence the proliferation and apoptosis of neuroblastoma cells via HIF-1alpha and p53 pathways [34, 35]. Despite the lack of direct evidence for m⁶A regulation in the four dme-lncRNAs involved in GC, our methylome and expression profiles indicated the expression of them were very likely to be regulated by m⁶A. The specific mechanisms need to be clarified by further molecular experiments.

Given that lncRNAs usually exert their regulatory roles by making effects on the expression of protein-coding genes, we identified the potential target genes for all dme-lncRNAs by analyzing their expression correlation in GC. The subsequent prediction of biological function revealed their possible association with mitosis and cell cycle. A hint could be obtained that these dme-lncRNAs might be retained and functioned in the nucleus. Nuclear lncRNAs may regulate gene expression by modulating the activity of regulatory protein complexes, chromosomal conformations and more generally, nuclear organization [36]. Our results showed the differentially co-expressed genes were mainly enriched in the following items: nucleosome, kinetochore and centrosome of CC; DNA binding of MF; chromosome segregation of BP; and DNA replication, mitotic phase and prometaphase of pathways. The cell cycle of mitosis is comprised of interphase (G1 phase, S phase and G2 phase) and mitotic phase (prophase, prometaphase, metaphase, anaphase and telophase) [37]. DNA replication occurs in S phase. After mitotic period begins, the chromatin is condensed into two chromatids connected by kinetochore and two centrosomes move towards the cell poles forming a spindle. The kinetochores are linked to centrosomes in prometaphase and then the chromatids are separated. Consequently, the potential function of those differentially co-expressed genes almost covered the biological activities in the whole stages of mitotic cell cycle, suggesting that they might be tightly associated with GC cell proliferation. Based on the above-mentioned findings, a reasonable access for lncRNA m⁶A methylation to GC could be inferred that some exogenous or endogenous factors elevated the m⁶A levels of several nuclear lncRNAs in normal cells, caused the up- or down-regulated expression of them, and then changed the expression levels of their target genes. As a result, the cell cycle and mitotic processes in which these genes may participate were affected, leading to aberrant cell proliferation and ultimate gastric carcinogenesis.

To verify the feasibility of our assumed mechanisms, the regulatory roles of screened targets on cellular biological behaviors were further explored by the aid of CCLE database and available publications. We chose 29 differentially co-expressed genes with expression correlation in all the four dme-lncRNAs as representatives comprised of 16 up-regulated and 13 down-regulated genes. Their background expression status in GC cell lines demonstrated that the up-regulated genes generally had higher expression levels in low differentiated/undifferentiated GC cell lines than high/middle-differentiated GC cell lines (RAD51D gene with significance). Meanwhile, an opposite trend was manifested in the down-regulated genes (PGA3 gene with significance). As is known to all, the tumor cells with poor differentiation usually had more malignant features like faster proliferation, stronger invasion and migration when compared with well-differentiated types. In other words, those up-regulated oncogenes might facilitate the malignant transformation of GC cells, while the down-regulated ‘tumor suppressor genes’ could inhibit or reverse their malignant phenotypes. Similar findings could also be found in published original studies. Briefly, most up-regulated genes were associated with the promotion of tumor growth, invasiveness and metastasis, while the down-regulated genes tended to suppress tumor growth and induce apoptosis. Considering the two aspects of validation results, we believed that these targets were very likely to affect the biological behaviors of GC cells such as cell proliferation, and thus indeed potential pathways mediating the m⁶A in dme-lncRNAs to exert regulatory function. Even so, the detail mechanisms for each step need exact confirmation.

It should be acknowledged that our study had a few limitations. Firstly, the sample size for detection needed to be enlarged for more accurate results. Secondly, only association study and bioinformatics analysis were focused on this topic. All the hypotheses and relevant mechanisms need to be verified by further investigations with molecular experiments. In spite of these defects, as the first report, the joint analysis of lncRNA m⁶A methylome and lncRNA/mRNA expression profiles in GC provided valuable reference for the researches in this field and also theoretical basis for future experiments.

In summary, a joint analysis of lncRNA m⁶A methylome and lncRNA/mRNA expression profiles in GC was conducted to explore the regulatory roles of m⁶A modification in GC-related lncRNAs. We newly found four lncRNAs that might be modulated by m⁶A with differential expression in GC. Their potential target genes were suggested to influence the cellular processes and biological behaviors involved in mitosis and cell cycle. The potential impacts of these targets on GC cells were also supported by CCLE database and literature review. Therefore, the m⁶A levels in dme-lncRNAs might promote GC cell proliferation by regulating the expression of lncRNAs and associated genes. The present study would provide clues for the lncRNA m⁶A methylation-based research on GC epigenetic etiology and pathogenesis.