Circular RNA expression profiles in human bronchial epithelial cells treated with beryllium sulfate

Abstract

Circular RNAs (circRNAs), is a novel type of endogenous non-coding RNAs (ncRNAs) participated in the pathogenesis of many diseases. Beryllium is one of the carcinogenesis elements. However, the mechanism and function of circRNAs in human bronchial epithelial cells (16HBE) induced by beryllium sulfate (BeSO₄) was rarely reported. Therefore, the high-throughput RNA sequencing analysis was performed to detect the circRNA profiles between control groups and BeSO₄-induced groups. Furthermore, circRNA-miRNA-mRNA network, Gene Ontology (GO), Kyoto Encyclopedia of Genes and Genomes (KEGG) pathway analysis, and PPI network analysis were used for bioinformatics analysis. CircRNA sequencing analysis revealed that 36 circRNAs were up-regulated and 35 circRNAs were down-regulated in the BeSO₄-exposed groups. The selected circRNAs were verified by real-time fluorescent quantitative PCR (qRT-PCR). Hsa_circ_0004214 and hsa_circ_0003586 were validated to be up-regulated, hsa_circ_0047958, hsa_circ_0001944, and hsa_circ_0008982 were down-regulated. The circRNA-miRNA-mRNA network annotated the key signaling pathway including cellular senescence, TNF signaling pathway, NF-kappa B signaling pathway, HIF-1 signaling pathway, and Hippo signaling pathway. The PPI network indicated the most circRNAs might participate in the BeSO₄ toxicity by acting as a sponge for the miR-663b through JAK–STAT signaling pathway. In summary, our study suggests that circRNAs may play roles in the mechanism of beryllium toxicity.

Introduction

Beryllium is widely used in the industries of rockets, missiles, satellites, aerospace, electronics, and metallurgical owing to its high melting point, electrical conductivity, and many other excellent physical and chemical properties [1]. Beryllium and their compounds are mainly inhaled through the respiratory tract by dust and smoke and mainly deposited in the lungs, causing pneumonia and other serious damages to multiple organs and systems [2, 3]. With the continuous innovation of production technologies and enhancement of personal protection measures, the occurrence of acute beryllium disease has decreased and the chronic beryllium disease induced by long-term exposure of low-dose beryllium is still an occupational health problem [4]. Our previous studies have shown when human bronchial epithelial cells (16HBE) were treated with a series concentration of beryllium sulfate (BeSO₄) for 48 hours. It showed some toxicity effects, such as oxidative stress, inflammation, autophagy, and apoptosis [5, 6]. However, the molecular toxicity mechanisms of beryllium have not been fully elucidated.

Circular RNAs (circRNAs) are one types of circular non-coding RNAs (ncRNAs) characterized by the presence of a covalently binding at the 3′ end and 5′ end produced by back-splicing [7]. It widely spreads and has the characteristics of structural stability, sequence conservation, cell or tissue-specific expression and so on [8, 9]. MicroRNAs (miRNAs) are almost 18–22 nucleotides long that are involved in many biological process including cell proliferation and differentiation, autophagy, apoptosis, and cancer [10, 11].

Current research indicates that circRNAs act as a natural miRNA sponge to participate in its expression regulation and also regulate the transcription, cell cycle or aging through interaction with proteins [12, 13]. In addition, circRNAs are closely related to the regulation process of diseases such as cancer and play a certain role in the process of development, immune, and other important physiological processes [14, 15]. Although it was demonstrated that miRNAs were involved in beryllium induced cell injuries [6], the role of circRNAs in the toxic mechanism of beryllium has not been determined.

Hence, in this study, we investigated the differential expression profiles of circRNAs in 16HBE induced by BeSO₄ and found the potential role of circRNA-miRNA interaction, which will reveal an underlying toxic mechanism of circRNAs in the BeSO₄-induced 16HBE cells.

Materials and Methods

Cell culture and beryllium sulfate exposure

Human bronchial epithelial cells (16HBE) were obtained from Peking University Health Science Center. Cells were cultured in Dulbecco’s modified Eagle’s medium (Solarbio, China) containing 10% (v/v) fetal bovine serum (FBS) (Biological Industries, Israel) and antibiotics (100 U/mL penicillin, 100 μg/mL streptomycin) (Biosharp, China) in a humidified incubator with 5% CO₂ at 37 °C. BeSO₄·4H₂O used for cells treatment was from Aladdin (Shanghai, China). Our previous study demonstrated that the cell viability of 75% was at the BeSO₄ concentration of almost 150 μM, which can induce inflammation, autophagy and apoptosis [5, 6]. According to the studies of beryllium sulfate [16, 17]and our previous study [6], we chose the concentration of 150 μM to investigate the mechanism of low-dose effect of beryllium sulfate. Then, cells were exposed to 150 μM BeSO₄ for 48 h and then collected for RNA sequencing.

RNA extraction, library construction and sequencing

Total RNA was extracted using the mirVana miRNA Isolation Kit (Ambion) following the manufacturer’ s protocol. RNA integrity was evaluated using the Agilent 2100 Bioanalyzer (Agilent Technologies, Santa Clara, CA, USA). The samples with RNA Integrity Number (RIN) ≥ 7 were subjected to the subsequent library construction. After rRNA depleted and linear RNA digested by Ribonuclease R (Epicentre, Madison, WI, USA), 1 μg of total RNA per sample was used to constructed the libraries through 15 cycles using TruSeq Stranded Total RNA with Ribo-Zero Gold according to the manufacturer’ s instructions. Then these libraries were sequenced on the Illumina sequencing platform (HiSeq 2500) and 150 bp paired-end reads were generated. The circRNA sequencing and analysis were conducted by OE Biotech Co., Ltd (Shanghai, China).

Bioinformatic analysis of circular RNA sequences

Raw data of fastq format were firstly processed using the Trimmomatic software [18]. Clean data (clean reads) were obtained by removing reads containing adapter, reads containing ploy-N and low-quality reads from raw data. Clean reads were aligned to the reference genome utilizing the MEM algorithm of Burrows-Wheeler aligner (BWA, version .7.5a) [19]. Based on the junction reads and GT-AG splicing signals, circRNAs were verified using CIRI2 software [20]. Combined with annotation information in protein database, circRNAs were annotated for further analysis. The expression level of circRNAs was measured by RPM (Mapped back-splicing junction reads per million mapped reads). To identify differentially expressed circRNAs, statistical comparison between two different groups was determined by the DESeq (2012) R package [21]. False discovery rate (FDR) was used as the threshold of p-value in multiple test to judge the significance of gene expression difference. The screening criteria of differential expressed circRNA was │log₂fold change│ > 1, P < .05 and FDR < .1 [22–24]. The sequencing raw data have been submitted to the NCBI Sequence Read Archive (accession: PRJNA668881).

Real-time fluorescent quantitative PCR (qRT-PCR) verification of differentially expressed circRNAs

circRNAs that predicted to act as competing endogenous RNA were selected for further validation by real-time quantitative RT-PCR. Total RNA was extracted using TRIzol reagent (Invitrogen, Carlsbad, CA, USA) and subjected to reverse transcription (RT) according to the manufacturer’s protocol. cDNA synthesis was performed by using HiScriptIIQ RT SuperMix for qPCR kit (Vazyme, Nanjing, China). ChamQ SYBR qPCR Master Mix kit (Vazyme, Nanjing, China) and LightCycler® 480 II Real-time PCR Instrument (Roche, Swiss) were used to perform real-time quantitative RT-PCR to validate the expression level. Each sample was run in triplicate for analysis. The primer sequences (Table 1) were designed in the laboratory and synthesized by Generay Biotech (Generay, PRC). The expression levels of circRNAs were normalized to ACTB and were calculated using the 2^-ΔΔCt method.

Table 1.

Primers of circRNA sequences

Name	Primer sequence
hsa_circ_0004214	F:5’-ACGAGATGGTCAAGCCCTA-3’
	R:5’-GGGAGTGGAAGTTACAAAGAG-3’
hsa_circ_0001206	F:5’-CACAGAATGGACCTGTCTTTG-3’
	R:5’-AGTCCGTACATATTCCAGGTTA-3’
hsa_circ_0003586	F:5’-GGAACAAGCAACAAAGCTAATC-3’
	R:5’-CCGCAACCACTTGAGAAT-3’
hsa_circ_0047958	F:5’-TGGAGATGCCAGAAACCA-3’
	R:5’-GTGCTGACGATGCTTCTAT-3’
hsa_circ_0001944	F:5’-AAATGGGAAGACTTGGTTGT-3’
	R:5’-GGAGCAGAGCAGCTTGAA-3’
hsa_circ_0008982	F:5’-CCGGAACAAGCATGAGAT-3’
	R:5’-TTGTTGTTGAGAGACACACG-3’
ACTB(Human)	F:5’-CATTCCAAATATGAGATGCGTT-3’
	R:5’-TACACGAAAGCAATGCTATCAC-3’

Gene function annotation

Gene ontology (GO) and Kyoto Encyclopedia of Genes and Genomes (KEGG) pathway analysis based on the databases of GO and KEGG were used to reveal the biological functions of the circRNA source transcripts and the target mRNA in the network. According to the annotation of these databases, the method of hypergeometric distribution test was used to calculate the significance of differential circRNA and the target mRNA in the network enrichment in each GO or pathway entry. Gene function annotation was performed using OECloud tools (https://cloud.oebiotech.cn.)

Annotation for circRNA-miRNA interaction and the predicted target of miRNA

The circRNA-miRNA interactions were predicted with Circular RNA Interactome database (https://circinteractome.nia.nih.gov/) [25]. The differentially expressed circRNAs were annotated with the circRNA-miRNA interaction and then the circRNA-related miRNA intersected with the differentially expressed miRNA profiles. The targets of them were predicted by using Targetsan 7.2 [26], miRDB [27]and miRWalk (http://mirwalk.umm.uni-heidelberg.de/search_mirnas/) and intersected with the differentially expressed mRNAs as the final target genes.

Construction of circRNA-miRNA-mRNA network

The significantly differentially expressed circRNA, the circRNA-related miRNA and the target mRNA were superimposed onto the circRNA-miRNA-mRNA network. The network was visualized using Cytoscape software (version 3.6.0).

Protein–protein interaction network analysis

STRING (https://string-db.org) was used to analyze the interaction of target genes and construct the protein–protein interaction (PPI) network. The construction of PPI network was under the condition of the highest confidence (the interaction score > .9) and no more than 10 interactions.

Statistical analysis

Validated data are expressed as the mean ± standard deviation. Differences between the groups were assessed by statistical software SPSS 18.0 using student’s t test. P < .05 was considered indicating a statistically significant difference.

Results

Profiling of differentially expressed circRNAs in the BeSO₄-treated group

There were 9564 human circRNAs predicted by CIRI software. Compared to the database of circBase, we obtained 5855 known circRNAs and 3709 newly predicted circRNAs. Hierarchical clustering and volcano plot visualization showed that the expression level of circRNAs were variable and distinguishable among the samples (Fig. 1). A total of 71 circRNAs were differentially expressed (│log₂fold change│ > 1, P < .05 and FDR < .1) between the BeSO₄-treated group and the control group. There were 36 circRNAs found to be up-regulated (fold change >2.0) and 35 circRNAs to be down-regulated (fold change <.5) in the BeSO₄-treated group. The top 10 up-regulated and top 10 down-regulated circRNAs in 16HBE induced by BeSO₄ is shown in Table 2. These data suggest that the circRNA expression profiles of 16HBE altered by the treatment of BeSO₄.

Figure 1.

CircRNAs were differentially expressed in BeSO₄-treated group. Hierarchical clustering (A) and the volcano plot (B) indicates differences in circRNA expression profiling between the BeSO₄ (D1, D2 and D3) and control groups(C1, C2 and C3). The C1, C2 and C3 represent the control groups and the D1, D2 and D3 represent the BeSO₄-treated groups. The differential expression criteria are│log₂fold change│ > 1, P-value <.05 and FDR < .1. The red represents the up-regulated circRNAs, while the green represents the down-regulated ones.

Table 2.

Top 10 up-regulated and top 10 down-regulated circRNAs in 16HBE cells induced by BeSO₄

circRNA	Fold Change	P-value	FDR	Regulation	Gene symbol
hsa_circ_0004214	9.63	0.01	0.02	Up	AMOTL1
hsa_circ_0081311	9.44	0.01	0.04	Up	TRRAP
hsa_circ_0004811	8.64	0.01	0.02	Up	CHEK2
hsa_circ_0008053	5.25	0.01	0.06	Up	PAK1IP1
hsa_circ_0061395	5.07	0.02	0.04	Up	BACH1
hsa_circ_0002111	4.84	0.02	0.07	Up	PSD3
hsa_circ_0000607	4.28	0.02	0.03	Up	VPS13C
hsa_circ_0006421	3.72	0.03	0.09	Up	PTK2
hsa_circ_0000741	3.34	0.01	0.01	Up	POLR2A
hsa_circ_0008720	3.33	< .01	0.01	Up	ERCC6L2
hsa_circ_0008982	0.11	0.01	0.05	Down	DIAPH1
hsa_circ_0007209	0.15	0.02	0.07	Down	ZFAT
hsa_circ_0002762	0.17	0.03	0.06	Down	CDK17
hsa_circ_0001785	0.20	0.02	0.08	Down	ELP3
hsa_circ_0002211	0.21	< .01	< .01	Down	DDX17
hsa_circ_0000842	0.28	0.03	0.05	Down	RPRD1A
hsa_circ_0006062	0.30	0.03	0.04	Down	DYRK1A
hsa_circ_0003441	0.33	< .01	0.01	Down	TDRD3
hsa_circ_0046941	0.35	0.04	0.08	Down	GNAL
hsa_circ_0004458	0.36	0.02	0.07	Down	PSD3

Validation of differentially expressed circRNA by qRT-PCR

Three up-regulated and three down-regulated circRNAs that predicted to function as competing endogenous RNA were selected for further validation. The primers used in the validation were in Table 1. As shown in Fig. 2, five of the six validated circRNAs were consistent with those of RNA sequence; hsa_circ_0004214 and hsa_circ_0003586 were validated to up-regulate, hsa_circ_0047958, hsa_circ_0001944, and hsa_circ_0008982 were down-regulated.

Figure 2.

Validation of selected circRNA using real-time quantitative RT-PCR. The presented results are expressed as means ± SD. ^**P < .01,^*P < .05.

GO and KEGG enrichment analysis of the circRNA source transcripts

The role of these circRNAs was further investigated by GO and KEGG analysis.

GO enrichment analysis comprises biological process (BP), cellular component (CC) and molecular function (MF). In GO analysis, all the results were ranked by enrichment score (−log₁₀(P-value)) and top 10 of every category were displayed in Fig. 3A. The top three enriched GO terms were negative regulation of sodium ion transmembrane transporter activity, cellular response to reactive oxygen species and RNA processing in BP; histone acetyltransferase complex, nucleoplasm, and cytosol in CC; ATP binding, protein kinase binding, and phosphatidylinositol-3,4,5-trisphosphate binding in MF.

Figure 3.

The biological function analysis of the significantly differentially expressed circRNAs. (A) GO analysis of the source transcripts of dysregulated circRNAs. Royal bars represent biological process terms, red bars represent cellular component terms, and cyan bars represent molecular function terms. (B) The KEGG pathway enrichment analysis of the source transcripts of dysregulated circRNAs.

In KEGG analysis, the top 20 KEGG pathways were shown in Fig. 3B. The top five enriched pathways of the differentially expressed circRNAs included shigellosis, focal adhesion, human immunodeficiency virus 1 infection, lysine degradation, and bacterial invasion of epithelial cells.

Construction and biological function analysis of the circRNA-miRNA-mRNA regulation network

Differentially expressed circRNAs contain corresponding miRNA binding sites. To facilitate the investigation, the interaction between miRNAs and circRNAs were predicted by circular RNA interactome database. The differentially expressed circRNAs were selected for further analysis. circRNA-related miRNAs predicted by circular RNA interactome database and differentially expressed miRNAs were intersected to obtain the selected miRNAs. The target genes of these miRNAs were predicted by using Targetscan, miRDB, and miRWalk. A circRNA-miRNA-mRNA network was constructed to reveal the interactions among circRNA, miRNA, and mRNA. The predicted network was shown in Fig. 4A.

Figure 4.

The circRNA-miRNA-mRNA network analysis in the database of circular RNA interactome. (A) The circRNA-miRNA-mRNA network in circular RNA interactome. Blue ellipses represent circRNA, the yellow “V” represents miRNA, and red triangles represent mRNA. (B) The GO analysis of the target genes in the network of the circular RNA interactome. Royal bars represent biological process terms, red bars represent cellular component terms, and cyan bars represent molecular function terms.(C) The KEGG pathway enrichment analysis of the target genes in the network of the circular RNA interactome.

To further investigate the biological function of circRNAs in 16HBE cells induced by BeSO₄, the function analysis were also conducted in the genes of the circRNA-miRNA-mRNA network. In Go analysis, all the results were ranked by the enrichment score(−log₁₀(p-value)) and top 10 GO terms in the BP, CC, and MF categories were displayed in Fig. 4B. The top three enriched GO terms were smooth muscle tissue development, megakaryocyte development, and heart development in BP; postsynaptic density, caveola, and transcriptional repressor complex in CC; palmitoyltransferase activity, protein-cysteine S-palmitoyltransferase activity, and phosphatase binding in MF. In KEGG analysis, the top 20 KEGG pathways were shown in Fig. 4C. The top three enriched pathways were cellular senescence, human T-cell leukemia virus 1 infection, and microRNAs in cancer.

P‌PI network analysis of the target genes of hsa-miR-663b related circRNAs

As displayed in Fig. 5A, it shows that hsa-miR-663b has the most enriched circRNAs in the constructed circRNA-miRNA-mRNA network, so we selected hsa-miR-663b related circRNAs and mRNAs to establish a new network (Fig. 5A). STRING was used to predict the protein interaction among the target genes of hsa-miR-663b binding circRNAs network. We set the condition of the highest confidence (the interaction score > .9) and no more than 10 interactions to construct the PPI network. We found ten hub proteins (JAK1, JAK2, STAT3, LIF, LIFR, IGF1, HBEGF, TYK2, IL6ST, IL10, CTF1) in the PPI network (Fig. 5B). Among these proteins, some of them were found to engage in the JAK–STAT signaling pathway (Fig. 6), which indicated these hub proteins and JAK–STAT signaling pathway may play a crucial role in the mechanisms of beryllium toxicity in 16HBE cells_.

Figure 5.

The hub role analysis of hsa-miR-663b-related circRNAs in the circRNA-miRNA-mRNA regulation network. (A) The interaction network between hsa-miR-663b and its related circRNAs and mRNAs. Blue ellipse represents circRNA, the yellow “V” represents miRNA, and red triangle represents mRNA. (B) The PPI network of the target genes of hsa-miR-663b. The node size represent degree and the edge size represent combined score.

Figure 6.

The hub proteins engaged in the JAK–STAT signaling pathway after the beryllium sulfate treatment in 16HBE cells. The red indicated the hub proteins.

Discussion

To date, previous study has reported miRNA played an important role in beryllium toxicity, but the role of circRNA in the mechanism of beryllium toxicity have not been elucidated [6]. Here, high-throughput sequencing analysis was performed to detect the different expression profiles of circRNAs in the BeSO₄-induced 16HBE cells, and bioinformatics tools were used to explore the underlying mechanism of circRNA in beryllium toxicity.

Liu et al [4]found beryllium inhibited apoptosis through mitochondrial apoptosis pathway in the rat model of pulmonary disease induced by beryllium oxide. Mitochondria, as the place of the cellular energy metabolism, plays important roles in the process of cell signal transduction, apoptosis, and cancer [28]. Previous studies indicated that oxidative stress involved in the beryllium toxicities [29, 30]. We found that 36 significantly up-regulated and 35 significantly down-regulated circRNAs were identified in total. Gene functions analysis revealed that the enriched GO terms were closely related to the cellular energy metabolism and oxidative stress. This suggests that the toxicity mechanism of beryllium sulfate may be related to the cellular energy metabolism dysregulation and oxidative stress.

Researches have shown that beryllium toxicity mechanisms were associated with the biological processes of oxidative stress, apoptosis, inflammation, and so on [4, 5, 31]. In the functional analysis of circRNA-miRNA-mRNA network, the most enriched pathway was cellular senescence, which is a complex phenotype that always occurs in response to oxidative stress, DNA damage, and mitochondrial dysfunction [32, 33]. It also showed other enriched pathways involved in the process of oxidative stress, inflammation and apoptosis, such as Hippo signaling pathway, NF-kappa B signaling pathway, HIF-1 signaling pathway, TNF signaling pathway, etc. The Hippo pathway as a tumor suppressor inhibit cell proliferation, apoptosis, and autophagy [34]. The study found tectorigenin alleviated inflammation and apoptosis in rat tendon-derived stem cells via modulating NF-Kappa B and MAPK pathways [35]. Curcumin prevents chronic intermittent hypoxia-induced myocardial HIF-1 activation, oxidative stress, inflammation, and apoptosis [36]. In addition, there was evidence of inflammation, apoptosis, and autophagy involved in the toxic mechanism of beryllium sulfate [5, 6], suggesting that the above enriched signaling pathways yielded several insights into the mechanism of beryllium toxicity.

It has been reported that the role of hsa-miR-663b affected the process of proliferation, invasion, and migration in many other diseases, such as osteosarcoma and pancreatic cancer [37, 38]. Hsa-miR-663b has the most binding sites with circRNAs in the constructed circRNA-miRNA-mRNA network. Since circRNAs can serve as ceRNAs to sponge miRNAs [12], we infer that circRNAs in the network may possibly act as the sponge of hsa-miR-663b to regulate the target genes and then exert its biological functions in the damage of 16HBE cells induced by BeSO₄. Researches show that JAK–STAT signaling pathway is essential in immunity, proliferation, differentiation, apoptosis, and tumorigenesis [39–41]. The PPI network showed that some of these hub proteins were involved in the JAK–STAT signaling pathway, which indicates that JAK–STAT signaling pathway may play a crucial role in the potential mechanism of beryllium exposure to 16HBE cells. Therefore, research on the oncogenic role of miR-663b and its binding circRNAs through JAK–STAT signaling pathway should go further.

In conclusion, the aberrant circRNA expression profiles were screened following exposure to BeSO₄ for 48 hours in 16HBE cells through high-throughput sequence analysis. GO and pathway analysis revealed the potential role of the target genes of circRNAs in 16HBE cells treated by BeSO₄, and we found they were related to the cellular energy metabolism and oxidative stress. The circRNA-miRNA-mRNA network annotated that cellular senescence, TNF signaling pathway, NF-kappa B signaling pathway, HIF-1 signaling pathway, and Hippo signaling pathway were enriched in 16HBE cells induced by BeSO₄. In addition, the PPI network indicated the most circRNAs might participate in the BeSO₄ toxic effects by acting as a sponge for the miR-663b through JAK–STAT signaling pathway in 16HBE cells by bioinformatics analysis. Taken together, our results suggested that circRNAs were associated with BeSO₄ toxicity in 16HBE cells.

Funding

This study was supported by the National Natural Science Foundation of China (81573193), the Natural Science Foundation of Hunan Province (2020JJ4082), and Hunan Provincial Innovation Foundation For Postgraduate (CX20200963).

Conflict of interest statement

The authors declare that there are no conflict of interest.