Comprehensive analysis of RNA-5-methylcytosine modification in breast cancer brain metastasis

Abstract

Aim: To delineate the RNA-5-methylcytosine (m5C) modification of breast cancer brain metastasis (BCBM).

Methods: Methylated RNA immunoprecipitation next-generation sequencing (MeRIP-seq) was performed to obtain RNA-m5C patterns of BCBM.

Results: 1048 hypermethylation and 1866 hypomethylation m5C peaks were identified in BCBM compared with those in breast cancer. The most significant m5C hypermethylated genes included ENG, SHANK1, IGFN1, EVL and MMP9, whereas the most significant m5C hypomethylated genes included AREG, SAA2, TP53I11, KRT7 and LCN2. MeRIP-qPCR data were concordant with the corresponding MeRIP-seq results in terms of the observed m5C levels. Conjoint analysis identified 190 hyper-up genes characterized by concurrent m5C hypermethylation and up-regulation, alongside 284 hypo-down genes exhibiting both m5C hypomethylation and down-regulation.

Conclusion: This study presents the first comprehensive analysis of RNA-m5C modification in BCBM.

Plain language summary

Article highlights.

• Breast cancer (BC) is the second most common solid malignancy to involve the brain. However, the underlying molecular mechanism of breast cancer brain metastasis (BCBM) remains largely unknown. We attempted to delineate the RNA 5-methylcytosine (m5C) modification landscape of BCBM to elucidate its underlying molecular mechanism.

• We employed methylated RNA immunoprecipitation next-generation sequencing (MeRIP-seq) to obtain a comprehensive profile of m5C methylation patterns in the BCBM group (231-BR cells) in comparison to the BC group (MDA-MB-231 cells).

• A total of 1048 hypermethylation and 1866 hypomethylation m5C peaks were identified in the BCBM group compared with those in the BC group. The most significant m5C hypermethylated genes included ENG, SHANK1, IGFN1, EVL and MMP9, whereas the most significant m5C hypomethylated genes included AREG, SAA2, TP53I11, KRT7 and LCN2.

• MeRIP-qPCR results were consistent with the corresponding MeRIP-seq data regarding m5C methylation levels of the candidate molecules IGFN1, EVL, AREG and KRT7.

• Conjoint analysis identified 190 hyper-up genes characterized by concurrent m5C hypermethylation and up-regulation, alongside 284 hypo-down genes exhibiting both m5C hypomethylation and down-regulation.

• Protein-protein interaction (PPI) network analysis based on differentially m5C methylated and expressed genes identified MMP9, ICAM1, THBS1, LAMA5, PDGFRB, TGFB2, NOTCH1 and SAA2 as hub genes related to BCBM. The hub genes were associated with the positive regulation of cell migration, response to oxygen levels, regulation of epithelial cell migration and inflammatory response.

• An analysis of methylation regulators indicated that the methyltransferases NOP2, NSUN5, DNMT3A and DNMT3B and the demethylase TET2 may be implicated in the regulation of m5C RNA methylation related to BCBM.

• This study presents the first m5C modification profile of mRNAs in BCBM, providing a framework for a better understanding of the mechanisms that trigger brain metastasis and delineating innovative biomarkers and therapeutic targets for improving the diagnosis and treatment of BCBM.

1. Introduction

Brain metastasis represents a prominent source of morbidity and mortality in patients suffering from breast cancer (BC) [1]. BC ranks as the second most prevalent solid malignancy that metastasizes to the brain [2]. Approximately 15% of metastatic BC cases develop brain metastasis [3]. Breast cancer brain metastasis (BCBM) frequently impairs cognitive and sensory functions, resulting in severe neurological deficits and a dismal prognosis, characterized by an alarming mortality rate of approximately 80% within 1 year of diagnosis [4]. Therefore, the high incidence of BCBM and its impact on survival present a critical unmet need to identify the underlying mechanisms that regulate the key steps of BCBM occurrence and progression.

Cancer cells acquire metastasis ability through a complex interplay of genetic alterations and epigenetic modifications. A recent study has uncovered a significant upregulation of circKIF4A in triple-negative breast cancer (TNBC) cell lines and brain metastases, where it promotes brain metastasis through a competitive endogenous RNA (ceRNA) mechanism involving the circKIF4A-miR-637-STAT3 axis [5]. Furthermore, SOX2 has been shown to promote brain metastasis of BC by upregulating the expression of FSCN1 and HBEGF [6]. Epigenetic modifications, encompassing alterations to DNA and RNA, play crucial roles in various biological processes [7]. The most common DNA modification in mammalian genomes is methylation at the 5th carbon of cytosine (5mC), catalyzed by DNA methyltransferases (DNMTs) and predominantly found in symmetrical CpG dinucleotides [8]. Similar to DNA 5mC, an active methyl-group from the donor, typically S-adenosyl-methionine (SAM), is transferred to the carbon-5 position of cytosine in RNA, resulting in the formation of the 5-methylcytosine (m5C) modification. This modification is a ubiquitous RNA alteration found in messenger RNA (mRNA) as well as non-coding RNAs (ncRNAs), including transfer RNA (tRNA), ribosomal RNA (rRNA), long non-coding RNA (lncRNA), small nuclear RNA (snRNA), microRNA (miRNA) and enhancer RNA (eRNA) [9]. Accumulating evidence indicates that m5C RNA modification regulates gene expression by influencing RNA metabolism, nuclear export, translation, RNA fragmentation and ribosome composition [10,11]. Recent advances in methylated RNA immunoprecipitation sequencing (MeRIP-seq) technologies and computing platforms have made it possible to sequence the entire mRNA m5C mapping. In previous studies, MeRIP-seq was applied to characterize the unique m5C modification profiles in various biological contexts, including mRNA from hepatocellular carcinoma [12], osteoarthritis (OA) cartilage [13] and lncRNAs in influenza A virus (IAV)-infected A549 cells [14]. Mounting evidence indicates that RNA m5C modification affects a diverse array of genes implicated in various cancers, including esophageal squamous cell carcinoma, bladder cancer and ovarian cancer [15–17]. The m5C RNA modification exhibits dual effects on the development of BC. Recent studies have identified NOP2/Sun RNA methyltransferase 2 (NSUN2), Aly/REF export factor (ALYREF) and DNA methyltransferase 3 beta (DNMT3B) as risk factors and NSUN5, NSUN6, tet methylcytosine dioxygenase 2 (TET2) and DNMT2 as protective factors in BC [18–21]. However, few studies adopted the MeRIP-seq strategy, which enables precisely elucidating the global mRNA m5C profile, to delineate the m5C modification landscape in BCBM.

In this study, we harnessed m5C MeRIP-seq, RNA sequencing (RNA-seq) and bioinformatics analyzes to elucidate the role of m5C modification in BCBM pathogenesis. Our comprehensive analysis of MeRIP-seq and RNA-seq data revealed significant differences in the number and distribution of m5C peaks between human breast cancer brain metastasis cells (231-BR) and their parental cells (MDA-MB-231). Bioinformatics analysis further revealed that the differentially m5C methylated genes played critical roles in cell function and participated in critical pathways. Furthermore, we assessed the correlations between mRNA and m5C modification in BCBM. Our results provide novel insights into the underlying molecular mechanisms of BCBM, potentially paving the way for the future development of innovative diagnostic and therapeutic strategies.

2. Materials & methods

2.1. Cell culture

The culture conditions of human BCBM cell line 231-BR and its parental BC cell line MDA-MB-231 were as described previously [22–24]. Both of these cell lines were kindly provided by Patricia S. Steeg (National Cancer Institute, NIH, MD, USA).

2.2. RNA preparation

Total RNA was extracted from the cultured cells using TRIzol reagent (Invitrogen Corporation, Carlsbad, CA, USA). Ribosomal RNA was depleted using the Ribo-Zero rRNA Removal Kit (Illumina, Inc, CA, USA). RNA concentrations were measured using a NanoDrop ND-1000 spectrophotometer (Thermo Fisher Scientific, MA, USA). RNA integrity and gDNA contamination were determined using denatured agarose gel electrophoresis.

2.3. MeRIP library preparation & sequencing

MeRIP-seq was conducted according to a previously reported procedure [12]. Briefly, m5C methylated RNAs were immunoprecipitated using the GenSeqTM m5C RNA IP Kit (GenSeq Inc, Shanghai, China). For RNA-seq library generation, non-immunoprecipated input samples and m5C immunoprecipitated samples were processed using the NEBNext^® Ultra II Directional RNA Library Prep Kit (New England Biolabs, Inc, MA, USA). The quality of the libraries was assessed using the BioAnalyzer 2100 system (Agilent Technologies, Inc, CA, USA). The library was sequenced on an Illumina Hiseq sequencer, generating 150 bp paired-end reads for subsequent analysis.

2.4. RNA-seq

RNA-seq was performed as described previously [14]. RNA-seq libraries for the non-immunoprecipated input samples were constructed using the NEBNext^® Ultra II Directional RNA Library Prep Kit (New England Biolabs, Inc). Briefly, the library preparation involved the fragmentation of RNA into smaller fragments, adapter ligation and amplification of the library. The libraries were then sequenced on the Illumina NovaSeq 6000 platform (Illumina, Inc) using 150 bp paired-end reads as described previously [25].

2.5. Identification & analysis of m5C peaks

Paired-end reads with a Q30 quality score of over 80% were retained (Supplementary Table S1). After 3′ adaptor-trimming, low-quality reads were removed using Cutadapt version 1.9.3, and mRNA peaks were identified using the DCC software [26]. Clean reads of all libraries were then aligned in Hisat2 version 2.0.4 to the reference genome [27]. Methylation modifications were identified through model-based analysis of ChIP-Seq (MACS) [28]. Differentially methylated sites between the two groups were determined using diffReps [29]. Peaks indicating differential methylation were overlapped with mRNA exons and identified using homemade scripts in both MACS and diffReps.

2.6. Sequence alignment

Read abundance at specific genomic positions was visualized in the Integrative Genomics Viewer. The identified m5C peaks underwent motif enrichment analysis in DREME and then localized to a given mRNA sequence. Subsequently, high-quality reads were mapped to the human reference genome (GENCODE Human Release 32) using Hisat2 version 2.0.4 [27]. Additionally, mRNA expression profiles were obtained using HTSeq version 0.9.1 [30]. Data were normalized and differentially expressed mRNAs (fold change ≥2.0 and p < 0.0001) were screened using EdgeR version 3.16.5 [31]. LogCPM (counts per million reads) were then obtained for the differentially expressed mRNA. A scatterplot depicting the differential expression of methylated genes and their degree of methylation was generated.

2.7. Gene ontology & Kyoto Encyclopedia of Genes & Genomes pathway enrichment analyzes

Gene ontology (GO) analysis (http://www.geneontology.org) was conducted to elucidate the function of differentially methylated genes (DMGs) [32]. Kyoto Encyclopedia of Genes and Genomes (KEGG) pathway enrichment analysis was performed on the differentially expressed genes using the KOBAS software [33]. For both analyzes, p < 0.05 was considered significant enrichment.

2.8. m5C methylation level analysis

m5C methylated RNAs were immunoprecipitated using the GenSeq™ m5C RNA IP Kit (GenSeq Inc). Briefly, RNA was randomly fragmented to approximately 200 nt using RNA fragmentation reagents (GenSeq Inc). RNA fragments were then incubated with anti-m5C conjugated beads at 4°C for 1 h. The m5C methylated RNAs were eluted and purified via phenol/chloroform extraction. m5C methylation levels of the candidate genes were evaluated using RT-qPCR. The primer sequences utilized are detailed in Supplementary Table S2. The enrichment was calculated using the following formula:

$$\frac{[2^{\left(\text{Ct}\right(\text{Input})-\text{Ct}(\text{IP}\left)\right)}\times \text{dilution} \text{fold}]}{[2^{\left(\text{Ct}\right(\text{Input})-\text{Ct}(\text{IgG}\left)\right)}\times \text{dilution} \text{fold}]}$$

The relative m5C methylation levels (BCBM vs BC) were calculated by dividing the enrichment value of the BCBM group by that of the BC group.

2.9. RT-qPCR analysis

The mRNA was reverse transcribed to generate cDNA using the PrimeScript RT Master Mix kit (Takara, Dalian, China). RT-qPCR analysis was performed using the TB Green Premix Ex Taq II kit (Takara, Dalian, China). The amplification mix contained 0.8 μl of forward primer (10 μM), 0.8 μl of reverse primer (10 μM), 2 μl of cDNA (50 ng/μl), 10 μl of 2× master mix, 0.4 μl ROX II and 6 μl ddH2O. PCR was conducted on the 7500 Real-Time PCR System (Applied Biosystems, CA, USA). ACTB served as an internal reference gene.

2.10. Protein-protein interaction network

Protein-protein interaction (PPI) analysis of the genes with coefficient changes was performed using information from STRING database (https://cn.string-db.org/). The PPI network was constructed using the Cytoscape software. Utilizing the molecular complex detection (MCODE) algorithm, significant clustering modules were extracted from the PPI network.

2.11. TCGA data analyzes

Differential gene expression levels of the m5C RNA methylation regulators between BC and normal tissues were analyzed based on The Cancer Genome Atlas (TCGA) dataset, which was retrieved from the UCSC Genome Browser (http://genome.ucsc.edu/).

2.12. TIMER2.0 analysis

Correlations between the m5C RNA methylation regulators in BC (n = 1100) were analyzed using the Gene_Corr module based on TCGA data using TIMER2.0 (http://timer.comp-genomics.org/timer/) [34].

2.13. Statistical analysis

The quantitative data are expressed as mean ± standard error of the mean (SEM). Statistically significance between groups was determined using Student's t-test. Pearson correlation analysis was employed to assess the correlation between two groups. The p-value with the motif enrichment was calculated using Fisher's Exact Test [12]. The E-values for the motifs were calculated as described previously [35]. All statistical analyzes were conducted using SPSS26 software (SPSS Inc, IL, USA), with a p-value of <0.05 indicating statistical significance.

3. Results

3.1. General features of m5C modification of mRNAs in the BC & BCBM groups

To determine the m5C modification landscape of mRNAs in BC and BCBM, we performed MeRIP-seq using the 231-BR/MDA-MB-231 model. The human ‘brain-seeking’ BCBM cell line 231-BR was initially established from the TNBC cell line MDA-MB-231 [36]. 231-BR cells metastasize to the brain with 100% frequency and have been used as a preclinical research model of brain metastatic BC [37]. As shown in Supplementary Table S1, high-throughput sequencing yielded viable data of satisfactory quality. A total of 45,890 and 52,438 m5C methylation peaks were detected in the BC and BCBM groups, respectively, with 19,911 shared peaks (Figure 1A). We mapped 14,480 annotated genes in the BC group and 15,097 annotated genes in the BCBM group (Figure 1B). As shown in Figure 1C & D, the m5C methylation of mRNAs in both the BC and BCBM groups occurred mainly in mRNA coding sequences (CDS), especially in the region adjacent to the 5′ untranslated region (UTR). The proportions of m5C methylation of BC and BCBM in different regions, including the 5′UTR, startC, CDS, stopC and 3′UTR, are shown in Figure 1E & F.

Figure 1.

The characteristics and distributions of m5C RNA methylation peaks in the BC and BCBM groups. (A-B) Venn diagram of m5C methylation peaks (A) and annotated gene read counts (B) in the BC and BCBM groups. (C-D) Distributions of m5C methylation peaks in the BC (C) and BCBM (D) groups. (E-F) Proportion of m5C occurred at different regions in the BC (E) and BCBM (F) groups. The m5C methylation of mRNAs in both the BC and BCBM groups occurred mainly in CDS.

3.2. Comparison of m5C RNA modification between the BCBM & BC groups

We found that 51.5% (4726/9177) of genes in the BC group and 49.3% (4925/9992) of genes in the BCBM group exhibited a single m5C methylation peak (p = 0.002; Supplementary Figure S1A). The average log2 enrichment folds of m5C methylation peaks in the BC and BCBM groups were 3.257 and 3.135, respectively (p = 0.716; Supplementary Figure S1B). The m5C methylation peak length was significantly longer in the BC group than in the BCBM group (p = 0.023; Supplementary Figure S1C). The motif analysis revealed significant differences in m5C methylation motifs between BC and BCBM groups (Supplementary Figure S2).

3.3. Differential analysis of m5C RNA modification between the BCBM & BC groups

We identified 1,048 hypermethylation peaks and 1,866 hypomethylation peaks between the BCBM and BC groups (threshold: fold change ≥2.0 and p ≤0.00001). The distribution of differential m5C methylation sites throughout the whole genome is displayed in Supplementary Figure S3A. The proportion of BCBM vs BC hypermethylated peaks was found to be significantly lower than the hypomethylated peaks (p < 0.001; Supplementary Figure S3A). The log2 enrichment folds of BCBM vs BC hypermethylated peaks were significantly lower than those of hypomethylated peaks (p < 0.001; Supplementary Figure S3B). The peak length of m5C methylation showed no significant differences between BCBM vs BC hypermethylated and hypomethylated peaks (p = 0.851; Supplementary Figure S3C).

The top 10 differentially hypermethylated and hypomethylated peaks between the BCBM and BC groups are listed in Table 1. The most significant m5C hypermethylated genes included endoglin (ENG), SH3 and multiple ankyrin repeat domains 1 (SHANK1), immunoglobulin like and fibronectin type III domain containing 1 (IGFN1), Enah/Vasp-like (EVL) and matrix metallopeptidase 9 (MMP9), while the most significant m5C hypomethylated genes included amphiregulin (AREG), serum amyloid A2 (SAA2), tumour protein p53 inducible protein 11 (TP53I11), keratin 7 (KRT7) and lipocalin 2 (LCN2). Subsequently, we randomly selected a hypermethylated gene, IGFN1 and a hypomethylated gene, KRT7, to view their m5C methylation profiles. As shown in Supplementary Figure S3D, IGFN1 was hypermethylated, whereas KRT7 was hypomethylated in the BCBM group compared with those in the BC group.

Table 1.

Top 10 hypermethylated or hypomethylated peaks in BCBM vs BC samples.

Hypermethylated peaks					Hypomethylated peaks
Gene name	Chromosome	TxStart	TxEnd	Log2Fold change	Gene name	Chromosome	TxStart	TxEnd	Log2Fold change
ENG	9	130616661	130617052	12.585	AREG	4	75310814	75310960	-14.012
ENG	9	130577801	130578320	11.940	AREG	4	75480590	75480740	-13.864
SHANK1	19	51171521	51172040	11.716	SAA2	11	18270081	18270182	-12.907
IGFN1	1	201178561	201179120	11.302	TP53I11	11	44953901	44954360	-11.877
EVL	14	100589901	100589939	10.894	KRT7	12	52626953	52627160	-11.831
SHANK1	19	51170821	51171360	10.745	TNS4	17	38643441	38643712	-11.665
SHANK1	19	51190000	51190069	10.696	SORCS2	4	7194521	7194853	-11.519
ENG	9	130591965	130592106	10.590	LCN2	9	130912516	130912620	-11.315
ARHGAP30	1	161018281	161018880	10.400	CD74	5	149784647	149784743	-11.271
MMP9	20	44641893	44642020	10.256	LRP1	12	57574941	57575074	-11.160

3.4. m5C RNA methylation level analysis of the candidate genes

To further validate the MeRIP-seq results, we randomly selected two hypermethylated and two hypomethylated genes, shown in Table 1, to detect their relative m5C methylation levels in the BCBM group compared with those in the BC group using MeRIP-qPCR. As shown in Figure 2A-D, IGFN1 and EVL were significantly hypermethylated (p < 0.05), whereas AREG and KRT7 were significantly hypomethylated (p < 0.05) in the BCBM group compared with those in the BC group. MeRIP-qPCR results were consistent with m5C MeRIP-seq data regarding the relative m5C methylation levels of these four genes (p = 0.015), indicating that the differential m5C RNA methylation data were reliable (Figure 2E-F).

Figure 2.

Validation of m5C RNA methylation level of the candidate genes. (A–D) Relative m5C RNA methylation level of IGFN1, EVL, AREG and KRT7 in the BCBM group versus the BC group by MeRIP-qPCR. n = 3. (E) Comparison of the mean fold changes (log10 transformed) between MeRIP-qPCR and MeRIP-seq. (F) Correlation analysis of the mean fold changes (log10 transformed) between MeRIP-qPCR results and MeRIP-seq data regarding the relative m5C RNA methylation levels.

3.5. Functional analysis of the differentially m5C methylated genes

To predict the potential functions and enriched pathways of the differentially m5C methylated genes, GO and KEGG pathway analyzes were performed. GO analysis revealed that the m5C hypermethylated genes were primarily linked to the BP terms GO:0051124∼ synaptic growth at the neuromuscular junction and GO:1902287∼ semaphorin-plexin signaling pathway involved in axon guidance (Figure 3A), whereas the m5C hypomethylated genes were associated with the BP terms GO:0072017∼ distal tubule development and GO:0007256∼ activation of JNKK activity (Supplementary Figure S4A). Furthermore, the m5C hypermethylated genes were associated with the CC terms GO:1990907∼ beta-catenin-TCF complex, GO:0002116∼ semaphorin receptor complex, GO:0048188∼ Set1C/COMPASS complex and GO:0044666∼ MLL3/4 complex (Figure 3B), whereas the m5C hypomethylated genes were associated with GO:0030877∼ beta-catenin destruction complex, GO:0099092∼ postsynaptic density, intracellular component and GO:0035102∼ PRC1 complex (Supplementary Figure S4B). Finally, the m5C hypermethylated genes were associated with the MF terms GO:0005522∼ profilin binding, GO:0042800∼ histone methyltransferase activity (H3-K4 specific) and GO:0008525∼ phosphatidylcholine transporter activity (Figure 3C), whereas the m5C hypomethylated genes were linked to GO:0005161∼ platelet-derived growth factor receptor binding and GO:0015280∼ ligand-gated sodium channel activity (Supplementary Figure S4C).

Figure 3.

Functional enrichment analysis of the m5C hypermethylated mRNAs in the BCBM group versus the BC group. (A–C) GO analysis of the m5C hypermethylated mRNAs. (D) KEGG pathway analysis of the m5C hypermethylated mRNAs.

BP: Biological process; CC: Cellular component; MF: Molecular function.

KEGG pathway analysis showed that the hypermethylated genes were enriched in pathways associated with adherent junction, basal cell carcinoma and pathways involved in cancer (Figure 3D). In contrast, the hypomethylated genes were associated with pathways associated with cancer, focal adhesion and endocytosis (Supplementary Figure S4D).

3.6. Differential expression analysis of mRNAs between the BCBM & BC groups

We performed RNA-seq to identify the differentially expressed mRNAs between the BCBM and BC input samples. As shown in Figure 4A, a total of 2373 DEGs were identified, of which 1085 were up-regulated and 1288 were down-regulated in the BCBM group compared with those in the BC group (threshold: fold change ≥2.0, p < 0.05 and FPKM ≥0.1 in at least one group). The top 10 up-regulated and down-regulated mRNAs are listed in Supplementary Table S3. Of these, Rho GTPase activating protein 30 (ARHGAP30), scavenger receptor class F member 1 (SCARF1) and MAGUK p55 scaffold protein 4 (MPP4) were the most significantly up-regulated mRNAs in the BCBM group compared with those in the BC group, whereas SAA2, transglutaminase 2 (TGM2) and serpin family B member 9 (SERPINB9) were significantly down-regulated. The hierarchical clustering heatmap of mRNAs illustrates the different expression patterns of mRNAs transcript variants between the BCBM group and the BC group (Figure 4B). Based on the expression levels of the DEGs, principal component analysis (PCA) revealed that the genes in the BCBM group were distinct from those in the parental non-specific metastasis BC group (Figure 4C).

Figure 4.

Differentially expressed transcript variants of the input samples in the BCBM group versus the BC group by RNA-seq. (A) Volcano plot of the differentially expressed transcript variants in the BCBM group versus the BC group. A total of 1,085 up-regulated and 1,288 down-regulated transcript variants were found in the BCBM group versus the BC group. Horizontal dotted line, padj = 0.05 (-log10 scaled). (B) Hierarchical cluster analysis showing the different expression patterns of transcript variants between the BCBM group and the BC group. (C) Principal component analysis (PCA) using the differentially expressed transcript variants (DEGs). PCA shows that the variants in the BCBM group are distinct from those in the BC group.

PC1: Principal component 1; PC2: Principal component 2.

3.7. Conjoint analysis between m5C RNA methylation level & differential expression

To explore the relationship between m5C RNA epitranscriptomic modification and mRNA expression, a conjoint analysis was conducted. As shown in Figure 5A & B, conjoint analysis identified 190 hyper-up genes with m5C hypermethylation and up-regulation, 7 hyper-down genes with m5C hypermethylation and down-regulation, 31 hypo-up genes with m5C hypomethylation and up-regulation and 284 hypo-down genes with m5C hypomethylation and down-regulation (threshold: fold change ≥2.0, p < 0.05).

Figure 5.

Correlation analyzes between m5C RNA methylation and transcript variant differential expression. (A) Scatter plot graph for m5C RNA methylation and transcript variant expression. (B) Up-set graph for m5C RNA methylation and transcript variant expression. (C) Top two PPI networks in MCODE analysis based on the genes with coefficient changes. (D) Enrichment analysis of the top two MCODE genes.

We constructed a PPI network based on the differentially m5C methylated and expressed genes with coefficient changes. The network comprised 439 nodes and 1702 edges. Subsequently, using MCODE, we aggregated and extracted sub-network modules from the PPI network. Figure 5C highlights the top two sub-network modules. Module A, with a score of 8.0, contained 17 nodes and 64 edges, including MMP9, intercellular adhesion molecule 1 (ICAM1), thrombospondin 1 (THBS1) and laminin subunit alpha 5 (LAMA5). Module B, with a score of 6.7, contained 12 nodes and 37 edges, including platelet derived growth factor receptor beta (PDGFRB), transforming growth factor beta 2 (TGFB2), notch receptor 1 (NOTCH1) and SAA2.

To gain further insights into the biological functions of the hub genes, we conducted a GO enrichment analysis based on the genes of the two modules using Metascape. The top 20 GO terms are displayed in Figure 5D. The identified hub genes are involved in the positive regulation of cell migration, response to oxygen levels, regulation of epithelial cell migration, inflammatory response and humeral immune response.

3.8. Validation of the differentially expressed levels of hypermethylated & hypomethylated genes

After identifying the hypermethylated and hypomethylated genes in the BCBM and BC groups, we next sought to validate their differential expression levels using RT-qPCR. As shown in Figure 6A-D, IGFN1 and EVL were significantly up-regulated (p < 0.05), while AREG and KRT7 were significantly downregulated (p < 0.05) in the BCBM group compared with those in the BC group. RT-qPCR results were consistent with RNA-seq data regarding the expression levels of mRNAs (p = 0.039; Figure 6E). A four-quadrant scatterplot indicated that the m5C hypermethylated genes IGFN1 and EVL were up-regulated, whereas the m5C hypomethylated genes AREG and KRT7 were downregulated in the BCBM group compared with those in the BC group (Figure 6F).

Figure 6.

Validation of the differential expression levels of the candidate mRNAs in the BCBM group versus the group. (A-D) Relative mRNA levels of IGFN1, EVL, AREG, and KRT7 in the BCBM group versus the BC group by RT-qPCR. n = 3. (E) Correlation analysis of the mean fold changes (log10 transformed) between qPCR results and RNA-seq data regarding the relative expression levels. (F) Quadrantal diagram of m5C RNA methylation and mRNA expression based on the MeRIP-qPCR and RT-qPCR data.

3.9. Analysis of m5C RNA methylation regulators involved in BCBM

m5C RNA methylation regulators can be functionally categorized as ‘writers’ (methyltransferases), ‘erasers’ (demethylases) and ‘readers’ (binding proteins) [38]. Considering the dynamic regulatory effect of m5C regulators on mRNA methylation, we analyzed the key regulators involved in the m5C RNA methylation in BCBM. As shown in Figure 7A, RNA-seq data showed that the methyltransferases NOP2, NSUN5, NSUN7, DNMT3A and DNMT3B were significantly up-regulated, while the demethylase TET2 was down-regulated in the BCBM group compared with those in the BC group (p < 0.05). There was no significant difference in the expression levels of other m5C regulators, including NSUN2, NSUN3, NSUN4, NSUN6, DNMT1, ALYREF, Y-box binding protein 1 (YBX1) and TET3, between the BCBM and BC groups (p > 0.05; Figure 7A). Except for NSUN7, the methyltransferases NOP2, NSUN5, DNMT3A and DNMT3B were significantly up-regulated, while the demethylase TET2 was downregulated in the BC group compared with those in the normal group, as observed using TCGA samples (p < 0.05; Figure 7B-G). Further correlation analysis revealed a significantly negative correlation between the methyltransferase NSUN5 and the demethylase TET2 (p = 5.49e-25; Figure 7H) and a significantly positive correlation between DNMT3A and TET2 (p = 5.72e–36; Supplementary Figure S5A). No correlation was observed between TET2 and the other differentially expressed methyltransferases (Supplementary Figure S5B & C).

Figure 7.

Analysis of m5C RNA methylation regulators involved in BCBM. (A) Differential expression level of the m5C RNA methylation regulators. Methyltransferases: NOP2, NSUN1, NSUN2, NSUN3, NSUN4, NSUN5, NSUN6, NSUN7, DNMT1, DNMT3A and DNMT3B. Demethylases: TET2 and TET3. Binding protein: ALYREF and YBX1. (B-G) Validation of the differential expression level of the candidate regulators based on TCGA samples. (H) Correlation analysis between the methyltransferase NSUN5 and demethylase TET2.

4. Discussion

To the best of our knowledge, this study offers the first comprehensive insight into the unique patterns and characteristics of m5C RNA hypermethylation and hypomethylation specifically in BCBM. Accumulating evidence indicates that the RNA m5C modification, is an important posttranscriptional modification of RNA and involved in multiple cellular processes as well as tumorigenesis and metastasis [39]. In our comparative analysis of BCBM and BC groups, distinct patterns of m5C hypermethylation and hypomethylation were observed, highlighting key genes including ENG, SHANK1, IGFN1, EVL and MMP9 for hypermethylation, and AREG, SAA2, TP53I11, KRT7 and LCN2 for hypomethylation. The identification of these differential m5C RNA epitranscriptomic modifications may be involved in BCBM and serve as novel diagnostic and therapeutic targets.

The analysis of m5C RNA modification in BC and BCBM transcripts revealed a predominant occurrence in the CDS, specifically in the start codon. While the precise impact of m5C modification on start codons remains elusive, several studies have hinted at potential connections. The presence of an m5C peak close to the translational start codon may indicate that m5C exerts an influence on the initiation of translation. This potential influence could be mediated by either enhancing or impeding the efficiency of ribosome scanning and subsequent start codon recognition [40]. In vitro translation studies conducted with both eukaryotic and bacterial translation systems, where either all cytosines (Cs) were substituted with m5C or m5C was incorporated into specific codons, have indicated that m5C exerts a negative effect on translation [41,42]. Similarly, N(1)-methyladenosine (m1A) modification is enriched in proximity to start codons and could promote translation initiation by modulating the secondary or tertiary structure of the RNA or by enhancing the recognition of translation initiation sites by specific reader proteins [43,44]. Further research is needed to investigate the specific roles and mechanisms of m5C modification in translation initiation.

Our findings revealed that IGFN1 and EVL exhibited m5C hypermethylation coupled with elevated expression, while AREG and KRT7 exhibited m5C hypomethylation coupled with reduced expression in the BCBM group compared with those in the BC group. These findings suggest that m5C modification may play a regulatory role in the expression of these genes in BCBM. IGFN1, known to produce multiple proteins through alternative splicing, is predominantly expressed in the skeletal muscle [45]. Its role in BC has been of interest, as previous studies have found that IGFN1 is frequently mutated in metastatic BC compared with early-stage BC [46]. Our results indicate that IGFN1 may be positively regulated by m5C modification in BCBM, suggesting a potential role for this modification in IGFN1's involvement in cancer progression. EVL, a member of the Ena/VASP family, is a regulatory factor with multiple functions in actin cytoskeleton remodeling, actin polymerization and cell adhesion [47]. EVL has been found to be up-regulated in human BC and its expression correlates with clinical stages [48]. Our finding of EVL hypermethylation and elevated expression in BCBM suggests that m5C modification may contribute to EVL's role in cancer progression and metastasis. AREG, a ligand for the EGFR, plays crucial roles in inflammatory responses, tissue regeneration and immune system function [49]. In erbB2/HER2-positive BC cells, AREG regulates cell proliferation and migration [50]. Our results showing AREG hypomethylation and reduced expression in BCBM indicate that m5C modification may negatively regulate AREG's functions in this context. KRT7, a member of the keratin gene family, plays a crucial role in regulating cell growth, migration and apoptosis in various cancers [51]. Recently, KRT7 was identified as a key effector for N6-methyladenosine (m6A)-induced breast cancer lung metastasis [52]. However, the role of KRT7 and its potential regulation by m5C modification in BCBM remains unclear. Our results provide initial evidence that KRT7 may be negatively regulated by m5C modification in this setting. Overall, the specific biological functions and underlying mechanisms of m5C modification of these genes in BCBM remain to be further explored.

The methyltransferases (“writers”), demethylases (“erasers”) and binding proteins (“readers”) of m5C RNA modification have been well-documented [10]. Over 10 known RNA m5C methyltransferases, including the NOL1/NOP2/SUN domain (NSUN) family member 1–7 (NSUN1-7), DNMT1, DNMT3A and DNMT3B [53] and erasers or m5C demethylases, such as alpha-ketoglutarate-dependent dioxygenase ABH1 (ALKBH1) and TET, have been recognized [38]. We found that the methyltransferases NOP2, NSUN5, DNMT3A and DNMT3B and the demethylase TET2 may be implicated in the regulation of m5C RNA methylation related to BCBM. NOP2 (also termed NSUN1) methylates human 28S rRNA cytosine at position 4,447 (C4447) [54]. NSUN5 introduces m5C at C3782 in the human 28S rRNA [38]. These methyltransferases and demethylases may be candidate regulators involved in the regulation of differential m5C RNA modification between the BCBM and BC groups. NSUN5 and TET2 have distinct roles in the regulation of m5C methylation, with NSUN5 acting on rRNA and TET2 acting on mRNA. A significant positive correlation was observed between DNMT3A and TET2, suggesting that these may have synergistic effects on the regulation of m5c methylation modification in BC. It is worth noting that the regulator of m5C RNA modification may be different for specific genes. These results provide novel insights into the differential mechanisms between BCBM and BC, highlighting the need for further study on the involvement of candidate regulators.

The results of GO analysis reinforce the role of the involvement of m5C RNA methylated genes in numerous biological processes. Several pathways have valuable implications, such as hemidesmosome assembly and apoptosis regulation involved in morphogenesis; these data imply that m5C modification influences the apoptotic process [55]. The KEGG pathway analysis revealed a compelling dichotomy in the functional enrichment of m5C hypermethylated and hypomethylated genes. Notably, the hypermethylated genes exhibited significant enrichment in pathways linked to adherent junction and pathways involved in cancer. This finding suggests that RNA m5C hypermethylation may play a pivotal role in the dysregulation of cellular adhesion and tumorigenesis. Adherent junctions are critical for maintaining tissue integrity and cell polarity and their disruption can contribute to tumor invasion and metastasis [56,57]. In contrast, the hypomethylated genes displayed enrichment in pathways associated with focal adhesion and endocytosis. Focal adhesion complexes are involved in cell-matrix interactions and play a crucial role in migration, polarization and metastatic cancer formation [58,59]. Endocytosis regulates the internalization of extracellular molecules, including nutrient, growth factors and receptors, which can influence cellular signaling and fate [60]. These findings highlight the complexity of RNA m5C methylation in cancer biology and the need for further research to elucidate the precise roles of hypermethylation and hypomethylation in tumor initiation, progression and metastasis.

The identification of distinct patterns of m5C modification and expression in BCBM compared with BC highlights its clinical implications, particularly with regard to its potential roles in tumor aggressiveness and chemoresistance. A recent single-cell RNA sequencing study has elucidated the landscape of BCBM and identified ILF2 as a potential therapeutic target [61]. The hub genes identified in our PPI network analysis, including MMP9, ICAM1, THBS1 and NOTCH1, are involved in key biological processes such as cell migration, epithelial-mesenchymal transition and angiogenesis, which are crucial for tumor progression and metastasis. The m5C modification status of these genes in BCBM cells suggests that they may serve as potential therapeutic targets. By targeting these m5C-modified genes or the enzymes regulating their m5C methylation status, we may be able to disrupt the metastatic cascade and improve the treatment outcomes for BCBM patients. Moreover, chemoresistance remains a significant hurdle in the treatment of metastatic BC. Notably, several studies have provided evidence that the combined utilization of targeting RNA modifications has the potential to improve chemoresistance to certain drugs [62]. For example, YBX-1 has been documented to facilitate the stabilization of m5C-modified mRNAs [16], while its overexpression is implicated in drug resistance among several human tumors [63,64]. In epithelial ovarian cancer (EOC), the silencing of YBX-1 leads to a significant reduction in drug resistance of EOC cells [65]. Further research is needed to elucidate the underlying mechanisms and develop targeted therapeutic strategies based on RNA m5C modification.

5. Conclusion

In conclusion, our study offers the first comprehensive insight into the patterns and characteristics of m5C RNA hypermethylation and hypomethylation in BCBM. We have identified an extensive set of m5C modification sites in mRNA and uncovered m5C-associated key genes. Additionally, a variety of in-depth bioinformatic analyzes were performed to demonstrate the alterations and possible functions of m5C methylation in BCBM. The limitations of this study encompass the absence of an evaluation of the expression of candidate m5C hypermethylation and hypomethylation molecules in both human breast cancer and their paired brain metastasis specimens, as well as the lack of thorough functional investigation and mechanistic studies of these molecules of interest. Our study provides a framework to deepen our understanding of the molecular mechanisms underlying BCBM and highlights promising diagnostic and therapeutic targets of clinical significance.

Funding Statement

This work was supported by National Natural Science Foundation of China (Grant no. 81702884), Natural Science Foundation of Shandong Province (Grant no. ZR2022MH272 and ZR2023QH115), Medicine and Health Science and Technology Foundation of Shandong Province (Grant no. 202111000399 and 202202080721), Traditional Chinese Medicine Science and Technology Foundation of Shandong Province (Grant no. M-2022122) and Liaocheng Key R&D Project (Grant no. 2022YDSF31 and 2022YDSF35).

Supplemental material

Supplemental data for this article can be accessed at https://doi.org/10.1080/14796694.2024.2405459

Author contributions

B Fu, W Zhang and Y Liu conceived and designed the study; P Cai, J Li, M An, M Li, J Guo, J Li, X Li, S Chen, A Zhang, P Li and Y Liu performed experiments and analyzed data; P Cai and B Fu wrote the main manuscript text and prepared all of the tables and figures; All authors reviewed the manuscript before submission; All the authors approved the final version of the manuscript.

Financial disclosure

This work was supported by National Natural Science Foundation of China (Grant no. 81702884), Natural Science Foundation of Shandong Province (Grant no. ZR2022MH272 and ZR2023QH115), Medicine and Health Science and Technology Foundation of Shandong Province (Grant no. 202111000399 and 202202080721), Traditional Chinese Medicine Science and Technology Foundation of Shandong Province (Grant no. M-2022122) and Liaocheng Key R&D Project (Grant no. 2022YDSF31 and 2022YDSF35).

Competing interests disclosure

The authors have no relevant affiliations or financial involvement with any organization or entity with a financial interest in or financial conflict with the subject matter or materials discussed in the manuscript. This includes employment, consultancies, honoraria, stock ownership or options, expert testimony, grants or patents received or pending, or royalties.

Preprint

A preprint has previously been published: P Cai, B Jiang, A Zhang, et al., “5-methylcytosine profile of mRNA in breast cancer brain metastasis,” PREPRINT, Research Square [https://doi.org/10.21203/rs.3.rs-1481071/v1], 2022.

Data availability statement

The data used to support the findings of this study are available in the NCBI repository. The data is accessible via NCBI GEO submission ID: GSE246721. It can be viewed at https://www.ncbi.nlm.nih.gov/geo/query/acc.cgi?acc=GSE246721.