Identification and screening of circular RNAs during adipogenic differentiation of ovine preadipocyte by RNA-seq

Abstract

Circular RNAs (circRNAs) are a class of non-coding RNAs that play important roles in preadipocyte differentiation and adipogenesis. However, little is known about genome-wide identification, expression profile, and function of circRNAs in sheep. To investigate the role of circRNAs during ovine adipogenic differentiation, the subcutaneous adipose tissue of Tibetan rams was collected in June 2022. Subsequently, the preadipocytes were immediately isolated from collected adipose tissue and then induced to begin differentiation. The adipocytes samples cultured on days 0, 2, and 8 of preadipocytes differentiation were used to perform RNA sequencing (RNA-seq) analysis to construct the expression profiles of circRNAs. Subsequently, the function of differentially expressed circRNAs was investigated by performing the Gene Ontology (GO) and Kyoto Encyclopedia of Genes and Genomes (KEGG) enrichment analysis of their parent genes. Finally, a circRNAs-miRNAs-mRNAs network involved in adipogenic differentiation was been analyzed. As a result, a total of 6,449 candidate circRNAs were identified in ovine preadipocytes. Of these circRNAs identified, 63 candidate circRNAs were differentially expressed among the three differentiation stages and their parent genes were mainly enriched in acetyl-CoA metabolic process, positive regulation of lipid biosynthetic process, positive regulation of steroid biosynthetic process, and focal adhesion pathway (P < 0.05). Based on a circRNAs-miRNAs-mRNAs regulatory network constructed, circ_004977, circ_006132 and circ_003788 were found to function as competing endogenous RNAs (ceRNAs) to regulate ovine preadipocyte differentiation and lipid metabolism. The results provide an improved understanding of functions and molecular mechanisms of circRNAs underlying ovine adipogenesis in sheep.

During the differentiation of ovine preadipocytes, the expression level of candidate circular RNAs related to lipid metabolism and synthesis was significantly changed at different differentiation stages. Several of these candidate circular RNAs may regulate adipogenesis by interacting with microRNAs.

Introduction

Proliferation and differentiation of preadipocytes are responsible for the number and size of adipocytes (Cornelius et al., 1994), respectively. They collectively influence fat development. Adipogenesis is a directed differentiation process of preadipocytes into mature adipocytes filled with triglycerides (Gregoire et al., 1998). The differentiation process mainly contains three adipogenic stages, namely proliferation, early differentiation, and terminal differentiation periods (Gregoire et al., 1998). The regulatory mechanisms of adipogenic differentiation are very complex, involving some coordinated changes in genes expression (Ibrahim, 2010; Cristancho and Lazar, 2011).

Recently, it was found that preadipocyte differentiation was also regulated by a class of non-coding RNAs called circular RNAs (circRNAs), which form a closed circular loop by back-splicing of pre-mRNA and present high evolutionary conservatism among various species (Li et al., 2018; Yu et al., 2021). Although, in general, circRNAs have lower levels of expression compared to the parent genes from which they derived, they play important roles in various biological process, including preadipocyte differentiation and lipid metabolism. For example, circSAMD4A derived from SAMD4A regulated adipogenesis of human by acting as a molecular sponge of miR-138-5p to abolish its inhibition effect on EZH2 expression (Liu et al., 2020). Similarly, a circFUT10-let‐7-PPARGC1B axis was also found to promote bovine preadipocyte proliferation and inhibit preadipocyte differentiation (Jiang et al., 2020). In addition, a recent study has also indicated that circRNA generated from ITGB1 (circITGB1) promoted proliferation and inhibited differentiation of ovine adipocytes via the miR-23a-ARRB1 pathway (Yue et al., 2023).

Although the importance of circRNAs during adipogenic differentiation has been revealed, most of these studies have focused on the expression profiles analysis of circRNAs at different differentiation stages using RNA sequencing (RNA-seq), and have investigated on humans (Sun et al., 2020), mice (Zhang et al., 2021), pigs (Tan et al., 2022), chickens (Zhang et al., 2020a), and yaks (Zhang et al., 2020b). These studies investigated the potential roles of differentially expressed circRNAs in preadipocytes by analyzing the functions of their parent genes or constructing circRNAs-related competing endogenous RNA (ceRNA) networks. Up to now, there was one study compared the expression profiles of circRNAs between preadipocytes and mature adipocytes in sheep (Xiao et al., 2021). In the study, four of 17 differentially expressed circRNAs identified may act as ceRNAs to regulate expression of adipogenesis-related genes (Xiao et al., 2021). However, this study did not contain the expression levels of circRNAs at early differentiation periods of ovine adipocytes. Therefore, to comprehensively investigate the roles of circRNAs in whole adipogenic differentiation process, the expression patterns of circRNAs at three differentiation stages, namely proliferation, early differentiation, and terminal differentiation period, need to be investigated.

In this study, to investigate the roles of circRNAs in ovine adipogenic differentiation and adipogenesis, we hypothesized that the expression level of circRNAs changed with the progression of ovine preadipocytes differentiation. To test this hypothesis, we compared the expression profiles of circRNAs in adipocytes of Tibetan sheep at three differentiation stages, and investigated the function of differentially expressed circRNAs by performing the Gene Ontology (GO) and Kyoto Encyclopedia of Genes and Genomes (KEGG) enrichment analysis of the parent genes. We also analyzed the circRNAs-miRNAs-mRNAs network involved in adipogenic differentiation. The results will improve the understanding of the roles of circRNAs in ovine preadipocytes differentiation and lipid metabolism, and then provide an opportunity to improve the meat quality and survival adaptability of Tibetan sheep.

Materials and Methods

Ethics statement

All animal procedures in this study were approved by Animal Experiment Ethics Committee of Gansu Agricultural University with an approval number of GSAU-ETH-AST-2021-028.

Culture and induction differentiation of ovine preadipocytes

The ovine preadipocytes used in this study were isolated from subcutaneous adipose tissue of three 1.5-year-old Tibetan rams. When the rams were slaughtered, their subcutaneous adipose tissues were collected and purified by removing visible blood vessels and connective tissues. The purified adipose tissues were mixed and chopped into 1.0 mm³ pieces. Subsequently, 20 mL of tissue digestion solution consisting of 1.0 U/mL neutral protease II solution and 0.75 U/mL collagenase IV solution, were used to digest about 1 g of minced adipose tissue at 37 °C for 1 h. Ovine preadipocytes were isolated from digested mixtures by filtrating and centrifugating at 1,500 × g for 5 min. Finally, the isolated cells were cultured at 5% CO₂ with 37 °C in growth medium composing of DMEM-F/12 medium (Hyclone, Logan, UT, USA) and 10% fetal bovine serum (Invigentech, Irvine, CA, USA). For every two hours, the cell suspension was re-seeded in a new culture dish (Corning, NY, USA) for purification. The purification procedure was performed at least three times.

The detailed methods about the cells culture and induction differentiation have also been descried by Hao et al. (2023). In this study, days 0 (D0), 2 (D2), and 8 (D8) of adipogenic differentiation were defined as preadipocytes stage, early differentiation stage, and terminal differentiation stage, respectively. Briefly, inducers of adipogenic differentiation (27.8 µg/mL 3-isobutyl-1-methylxanthine, 0.1 µg/mL dexamethasone, and 1 µg/mL insulin) were added to the growth medium to induce preadipocytes differentiation for 2 days. Subsequently, the induction medium was replaced by a maintenance medium (growth medium supplemented with 1 µg/mL insulin) and adipocytes were further cultured for 2 days. Finally, the maintenance medium was substituted with growth medium, and the culture of adipocytes continued until the differentiation process was finished.

Total RNA extraction and RNA-seq

The 12 isolated ovine adipocytes samples with the same density and growth status were cultured using the same growth medium at 37 °C with 5% CO₂. These 12 samples were randomly divided into three groups, which each group contained four replicates. The three groups were named days 0, 2, and 8, respectively. When the confluence of adipocytes in growth medium was over 90%, they were induced into differentiation described above. The groups of days 0, 2, and 8 were induced differentiation for 0 day, 2 days, and 8 days, respectively. After the induction of differentiation, the adipocytes samples of the three groups were used for further RNA-seq. Briefly, the Trizol reagent (Invitrogen, CA, USA) was used to isolate total RNA from the ovine adipocyte samples in the three groups. The concentration and integrity of the total RNA were evaluated using Nanodrop 2000 (Thermo Scientific, MA, USA) and Agilent 2100 Bioanalyzer (Agilent, CA, USA), respectively. Only qualified RNA samples with RNA integrity number > 7 were selected for the further use. After RNA quality detection, a Ribo-Zero Gold rRNA Removal Kit (Illumina, CA, United States) was used to remove ribosomal RNA (rRNA). The remaining RNA was used to generate cDNA libraries using a NEBNext Ultra RNA Library Prep Kit (New England Biolabs, MA, USA), and then performed a 150 bp paired-end sequencing by an Illumina HiSeq 4000 sequencer (Illumina, CA, USA).

Analysis of RNA-seq Data and Screening of Differentially Expressed circRNAs

For raw reads, their adapter reads, low quality reads with quality scores < Q20 and reads containing over 10% unknown nucleotides were removed to obtain clean reads, using fastp v0.18.0. In-house script was used to produce rRNA database by extracting ovine sequences with a key word of “rRNA” in GenBank (http://www.ncbi.nlm.nih.gov/genbank/). The clean reads were then mapped to the rRNA database to remove rRNA sequences using Bowtie2 v2.2.8. The remaining clean reads were mapped against Oar_rambouillet_v1.0 (https://www.ncbi.nlm.nih.gov/assembly/GCF_002742125.1/) using HISAT2 v2.1.0 with default parameters (Kim et al., 2015). Based on procedures developed by Memczak et al. (2013), 20-mers (anchors) sequences from both ends of the unmapped clean reads with ovine genome assembly were extracted uing Find_circ, and then re-aligned to ovine genome assembly using bowtie2 v2.2.8 with default parameters. The alignment results were used to identify circRNAs using software Find_circ (Memczak et al. 2013). To obtain highly reliable circRNAs, the following filter criteria were used: (1) The candidate circRNAs with only one clear breakpoint were retained. (2) For each read, the position overlap of its two anchors mapped to genome was no more than 2 bp. (3) For anchor reads mapping, only 2 bp base mismatch was allowed. (4) For each candidate circRNA, there must be more than two unique reads in anchors mapping results. (5) For one of the anchors of each read, the highest score of mapping results to genome was 35 points higher than the second highest score. (6) The number of unique reads supporting a specific candidate circRNA was more than half of the total number of samples. (7) The candidate circRNAs greater than 100k in length were removed. The circRNAs identified were characterized by analyzing their length, types, and chromosomal distribution, using in-house Perl scripts. The expression levels of circRNAs were normalized by calculating the reads per million mapped reads (RPM). Based on the negative binomial distribution test in DESeq v2.0 (Love et al., 2014), differentially expressed circRNAs between different differentiation stages were identified with Fold change > 2.0 and P-value < 0.05. In addition, expression patterns analysis of the differentially expressed circRNAs were analyzed by STEM tool (http://www.sb.cs.cmu.edu/stem/).

Validation of RNA-seq Data Using RT-PCR and RT-qPCR

To validate the authenticity of circRNAs obtained from RNA-seq, the reverse transcriptase-PCR (RT-PCR) and Sanger sequencing were performed. Briefly, nine differentially expressed circRNAs identified from three differentiation stages were randomly selected, including circ_001285, circ_005401, circ_003788, circ_001066, circ_002606, circ_000154, circ_005620, circ_004368, and circ_004915. The divergent primers (Table 1) were designed by Primer v3.0 to amplify back-spliced junction site of the selected circRNAs. The information about position of divergent primers, back-spliced junction site, and expected amplification size are showed in Supplementary File 1. The RNA samples that were used for RNA-seq were reverse-transcribed to produce cDNA, using a HiScript III 1st Strand cDNA Synthesis Kit (Vazyme, Nanjing, China). The PCR reaction was performed in a 20-µL system, containing 0.8-µL of the cDNA, 0.5 U of Taq DNA polymerase (Takara, Dalian, China), 150 µM of each dNTP (Takara, Dalian, China), 2.5 mM Mg²⁺, 0.25 µM of each primer, and 1 × PCR buffer supplied with DNA polymerase enzyme. Finally, PCR products were checked using 1.5% agarose gels electrophoresis and sequenced by Sanger sequencing. The sequences from Sanger sequencing were then aligned to the RNA-seq data to verify the location of the back-spliced junction sites of circRNAs.

Table 1.

PCR primers used for authenticity and reliability validation of nine differentially expressed circRNAs

CircRNA/gene	Forward primer	Reverse primer
circ_001285	CTCGAAGGACAAGCATCCAT	TTGGACTGCTACCACTGCTG
circ_005401	GCAGCTACCGCTTCCCCAAG	TTGGAAATGAGCAGGATGTG
circ_003788	ATGACTCTTGTGCTGCGTTG	AAGAGCCTTCACCAGCATGT
circ_001066	CCCCTGTAATGGCCTATCCT	GTCTGCTAACTGCGGGATGT
circ_002606	CCTATTCCGGGACAACAAGA	CTTCCCGGCACCTGATTCGC
circ_000154	AGCAGCAGCTCTTTCAGAGG	CCCCGAGATTCTAATGCAAG
circ_005620	CAACAAGCAGACCATCCTGA	GAACCAGAATGGGAAAAACG
circ_004368	TGACAGTGTGGAAGGCTCTG	CCTTGTATCCCAGGTTGGTG
circ_004915	TTTGCAAAATTGGCACTGAA	AGCAGCAGTGTCCTTCAACA
GAPDH	ACACTGAGGACCAGGTTGTG	GACAAAGTGGTCGTTGAGGG

To further validate the reliability of RNA-seq results, the expression levels of the same nine differentially expressed circRNAs were further detected by RT-quantitative PCR (RT-qPCR). The same primers and cDNA samples used for RT-PCR were also used for RT-qPCR analysis. The RT-qPCR was performed with a 2 × ChamQ SYBR qPCR Master Mix (Vazyme, Nanjing, China). The relative expression level of circRNAs was normalized by internal control gene GAPDH (Xiao et al., 2021) and then calculated by the 2^−ΔΔCt method (Livak and Schmittgen, 2001).

Function Annotation of the Parent Genes of Differentially Expressed circRNAs

To investigate biological function of differentially expressed circRNAs, their parent genes were performed GO enrichment analysis by the DAVID tool (https://david.ncifcrf.gov) (Ashburner et al., 2000). In addition, the KOBAS 3.0 (Bu et al., 2021) was also used to analyze the pathways enriched by the parent genes of differentially expressed circRNAs. The significant GO terms and KEGG pathways (P < 0.05) were defined based on hypergeometric test.

Construction of a ceRNA Network of circRNAs-miRNAs-mRNAs

According to our previous miRNAs and mRNAs sequencing data obtained from the same adipocyte samples (Hao et al., 2023), a ceRNA network of circRNAs-miRNAs-mRNAs was constructed based on the ceRNA hypothesis (Quan et al., 2021). Briefly, (1) the software miReap v0.2, Miranda v3.3a, and TargetScan v7.0 were collectively used to predict the target genes and binding circRNAs of differentially expressed miRNAs, and predicted results were then overlapped with differentially expressed genes (Hao et al., 2023) and differentially expressed circRNAs obtained from RNA-Seq in the study. The overlapped genes and circRNAs were finally used to construct the ceRNA network. (2) The correlation in expression between miRNAs and mRNAs or between miRNAs and circRNAs was, respectively, investigated by calculating Spearman Rank correlation coefficient (SCC). The miRNAs-circRNAs pairs or miRNAs-mRNAs pairs with SCC < -0.7 were selected for constructing ceRNA network. (3) Of the pairs identified above, the correlation in expression between circRNAs and mRNAs was evaluated by calculating Pearson correlation coefficient (PCC). Only pairs with PCC > 0.9 were selected for constructing ceRNA network. (4) The significant ceRNA pairs of circRNAs-miRNAs-mRNAs were identified based on hypergeometric cumulative distribution function test, and the ceRNA network was visualized by Cytoscape v3.5.1.

Statistical analysis

Statistical analysis was performed using SPSS v24.0 (IBM, Armonk, NY, USA), and data are presented as mean ± standard error. The expression differences among three groups were compared using one-way ANOVA analysis. Differences were considered significant at P < 0.05.

Results

Identification and Characterization of circRNAs

In this study, a total of 1,190,853,200 raw reads were obtained from the 12 cDNA libraries, and the raw reads have been submitted to Genbank with accession number of SRR19917845-SRR19917856. After quality control and removal of rRNA reads, an average of 38,692,391, 21,461,220, and 31,172,183 clean reads were obtained from ovine adipocytes on days 0, 2, and 8 of differentiation, respectively. Of these reads, 94.27%, 92.95%, and 93.00% reads can be well mapped against Ovine reference genome assembly (Oar_rambouillet_v1.0), respectively (Table 2). The average re-alignment rates of anchors on days 0, 2, and 8 of adipocytes differentiation were 51.87%, 47.21%, and 46.72%, respectively. From the re-alignment, a total of 19,792,461 splice sites were detected, which included 26,251 (0.13%) circular splice sites and 74,331 (0.38%) linear splice sites.

Table 2.

Overview of RNA-seq data

Samples	Average raw reads	Average clean reads	Q20	Average remaining clean reads	Average mapped reads	Mapped rate (%)
Day 0	104,875,453	104,689,034	98.40%	38,692,391	36,475,046	94.27
Day 2	96,118,943	95,949,195	98.53%	21,461,220	19,945,610	92.95
Day 8	96,718,905	96,278,852	97.84%	31,172,183	28,989,664	93.00

A total of 6,449 candidate circRNAs were identified in the three differentiation stages (Figure 1A, Supplementary Table S1). Of these circRNAs identified, 2,073, 1,133, and 1,553 circRNAs were specifically expressed on days 0, 2, and 8 of preadipocytes differentiation, respectively, while 575 candidate circRNAs were expressed across all the three-time points (Figure 1A, Supplementary Table S1). The majority of circRNAs identified in this study were annotated as annot_exons type with a proportion of 77.35%, followed by exon_intron (7.47%), one_exon (4.82%), intronic (4.56%), and antisense sequence types (4.12%), while intergenic sequences were the least common types (1.67%) (Figure 1B). The circRNAs identified were distributed across all the ovine chromosomes, but chromosomes 1-3 have the greatest number (Figure 1C). In addition, most of these circRNAs identified were less than 1 kb in length (Figure 1D).

Figure 1.

Characterization of circRNAs identified during different ovine preadipocytes differentiation periods. (A) Venn diagram summarizing the number of circRNAs specifically expressed on days 0, 2, and 8 of ovine preadipocytes differentiation, as well as the number of circRNAs expressed across all the three differentiation stages. The types (B), chromosome distribution (C), and length (D) of ovine circRNAs identified were exhibited.

Screening and Validation of Differentially Expressed circRNAs During Preadipocytes Differentiation

When comparing expression levels on days 2 to 0, days 8 to 0, and days 8 to 2 of preadipocytes differentiation, a total of 15, 32, and 16 differentially expressed circRNAs were found in day 2 vs day 0 (Supplementary Table S2), day 8 vs day 0 (Supplementary Table S3), and day 8 vs day 2 (Supplementary Table S4), respectively. It was noteworthy that some differentially expressed circRNAs were generated from genes related to adipocytes differentiation and lipid metabolism, such as COL6A2 circRNA (circ_005812) (Liu et al., 2017), CCDC88A circRNA (circ_001285) (Ghosh et al., 2011, Wang et al., 2020), MXRA5 circRNA (circ_000589) (Arner et al., 2018), IGF1R circRNA (circ_005401) (Wiper-Bergeron et al., 2003), and ZEB1 circRNA (circ_005620) (Saykally et al., 2009; Wang et al., 2021).

To validate the authentication of circRNAs identified, nine differentially expressed circRNAs were selected to perform RT-PCR and Sanger sequencing analysis. The results showed that the nine circRNAs were expressed in ovine adipocytes, and their sequences of head-to-tail splice junction site from Sanger sequencing were same as those from RNA-seq data (Figure 2A-C).

Figure 2.

Authentication verification of circRNAs identified from RNA-seq. (A) Schematic illustration of divergent primers designed for amplifying circRNAs in RT-PCR. (B) Detection of RT-PCR production of circRNAs amplified by divergent primers using agarose gel electrophoresis. (C) Identification of back-spliced junction sequences of circRNAs using Sanger sequencing. The junction site was pointed by black arrow.

To further validate the repeatability of the RNA-seq data, the same nine candidate circRNAs were also selected to perform RT-qPCR analysis. The results showed that the relative expression level of these differentially expressed circNRAs from RT-qPCR was consistent with the results from RNA-seq (Figure 3).

Figure 3.

RT-qPCR validation of expression levels of nine differentially expressed circRNAs identified by RNA-seq. Days 0, 2, and 8 represent samples from days 0, 2, and 8 of ovine preadipocytes differentiation, respectively. RT-qPCR data were shown as the means ± SD (n = 4). Different lowercase letters indicate signifcant diference between diferent differentiation periods (P < 0.05).

Expression Patterns Analysis of Differentially Expressed circRNAs During Adipocytes Differentiation

In this study, 63 differentially expressed circRNAs showed a total of eight expression patterns (Figure 4). It was noteworthy that profile 4 was significantly enriched by six differentially expressed circRNAs (Figure 4, Table 3), while profile 6 included the largest number of differentially expressed circRNAs (Figure 4). In profile 4, there was no significant change in expression levels of circRNAs between days 0 and 2 (P > 0.05), but their expression levels significantly increased from days 2 to 8 (Table 3, P < 0.05). In addition, circ_000589 and circ_004915 were only expressed on day 8 of preadipocytes differentiation (Table 3). These suggests that the circRNAs exhibited in profile 4 would function at late stage of preadipocytes differentiation and may play key roles in lipid accumulation. On the contrary, the expression level of 15 circRNAs in profile 6 were significantly increased from days 0 to 2 (P < 0.05), but did not significantly change from days 2 to 8 (P > 0.05) (Figure 4). These indicate that these circRNAs would mainly function in early stage of preadipocytes differentiation.

Figure 4.

Expression patterns analysis of differentially expressed circRNAs identified during ovine preadipocytes differentiation, using cluster analysis based on short time-series expression miner (STEM). The number of circRNAs enriched in each expression pattern is shown at the top of expression pattern figures. Colored profiles were significant expression patterns during preadipocytes differentiation (P < 0.05).

Table 3.

Expression pattern of differentially expressed circRNAs in profile 4

circRNA	Day 0_RPM	Day 2_RPM	Day 8_RPM	Parent gene
circ_000589	0	0	1860	MXRA5
circ_001285	1698	1165	4178	CCDC88A
circ_003788	135	201	1772	KIFAP3
circ_003918	376	200	1959	AGTPBP1
circ_004915	0	0	1278	DIAPH2
circ_005812	405	587	5731	COL6A2

Functional Annotation of the Parent Genes of Differentially Expressed circRNAs

GO enrichment analysis found that some important terms related to lipid metabolism and biosynthesis were significantly enriched by the parent genes of several differentially expressed circRNAs (Table 4), including Acetyl-CoA metabolic process, positive regulation of lipid biosynthetic process, and positive regulation of steroid biosynthetic process (P < 0.05). KEGG enrichment analysis found that aminoacyl-tRNA biosynthesis, lysine degradation, and focal adhesion pathway that were associated with adipogenesis were also significantly enriched by the parent genes of several differentially expressed circRNAs (P < 0.05) (Table 4).

Table 4.

The important GO terms and KEGG pathways enriched by the parent genes of differentially expressed circRNAs

Group	GO terms/KEGG pathway	P-value	Parent genes
Day 0 vs. day 2	GO: acetyl-CoA metabolic process	P < 0.05	AASS
	KEGG: aminoacyl-tRNA biosynthesis	P < 0.05	GARS
Day 0 vs. day 8	GO: acetyl-CoA metabolic process	P < 0.05	AASS
	GO: positive regulation of lipid biosynthetic process	P < 0.05	IGF1R
	GO: positive regulation of steroid biosynthetic process	P < 0.05	IGF1R
	GO: small GTPase binding	P < 0.05	RANBP17, DIAPH2
	KEGG: focal adhesion	P < 0.05	COL6A2, IGF1R
	KEGG: lysine degradation	P < 0.05	AASS, KMT2E
Day 2 vs. day 8	KEGG: lysine degradation	P < 0.05	NSD2

Construction of a ceRNA Network of circRNA-miRNA-mRNA

It was found that circRNA can function as miRNA sponges to regulate the expression level of the target genes of miRNA, with an accompanying regulation of preadipocytes differentiation and adipogenesis (Jiang et al., 2020). Based on the ceRNA hypothesis, a total of 49 ceRNA pairs of circRNA-miRNA-mRNA were defined. As shown in Figure 5, some ceRNA pairs were related to preadipocytes differentiation and lipid metabolism, including circ_004977-miR-200b-SLC6A20/Zdhhc15/ANK1, circ_006132-miR-106-5p/miR-17-5p/miR-20-5p/miR-150-FGF1, and circ_003788-miR-872-3p-OAS2 (Piórkowska et al., 2018; Jin et al., 2021; Cui et al., 2023).

Figure 5.

A ceRNA network of circRNA-miRNA-mRNA. The triangles represent differentially expressed circRNAs identified in ovine preadipocytes at different differentiation stages. The squares represent differentially expressed miRNAs sponged by these differentially expressed circRNAs, while the circles represent differentially expressed genes targeted by these differentially expressed miRNAs.

Discussion

In this study, a total of 6,449 candidate circRNAs were identified at the three differentiation stages of preadipocytes isolated from Tibetan sheep. It was noteworthy that 2,073, 1,133, and 1,553 candidate circRNAs were only expressed on days 0, 2, and 8 of preadipocyte differentiation, respectively, suggesting that developmental stage‐specific expression patterns of circRNAs, and crucial roles of circRNAs in adipogenesis. In this study, ovine chromosomes 1-3 produced the greatest number of circRNAs. This is reasonable as these chromosomes possess the largest size and comprise approximately 28.8% of the total length of the ovine genome (Oar_rambouillet_v1.0). Meanwhile, the majority of circRNAs identified in this study were generated from exons region of protein-coding genes, which may reflect that most of the circRNAs identified would act as molecular sponges of miRNAs to regulate the differentiation of ovine adipocytes.

Interestingly, we found that the candidate circRNAs derived from the same parent gene showed different expression patterns. For example, ovine SAMD4A is a parent gene for three candidate circRNAs (circ_006001, circ_003119, and circ_001397), but the three candidates exhibited different expression patterns (Table S1). In fact, this was not surprising as the follow reasons. On the one hand, the three circRNAs are derived from the different position of SAMD4A, which caused different splicing length and types of circRNAs. Of the three circRNAs, circ_006001 is a one_exon type of circRNA with 519 bp in length, circ_003119 is an intronic type of circRNA with 37,270 bp, while circ_001397 is an annot_exons type of circNRA with 783 bp. It is well known that different types of circRNAs play different roles in cells. The circRNAs with exons type are mainly localized in the cytoplasm and function as miRNA sponges, while exon–intron circRNAs predominantly localized in the nucleus can regulate the transcription of their parent genes in a cis-regulatory way (Hansen et al., 2013; Li et al., 2015). On the other hand, it has also been reported that different types of circRNAs have different cyclization mechanism, which may cause different splicing efficiency and expression abundance (Wilusz, 2018). Therefore, it was not surprising that different types of circRNAs derived from the same parent gene may have different expression patterns.

Of the 63 differentially expressed circRNAs identified, over half of them were screened in day 8 vs day 0. This was not surprising as there was the most significant phenotypic difference in lipid droplets accumulation between the two periods. Our expression patterns analysis showed that profile 4 was significantly enriched by six differentially expressed circRNAs, which may play key roles in lipid accumulation as their expression levels were significantly up-regulated on day 8 of preadipocytes differentiation, when compared to day 0 or day 2. Of the six circRNAs, circ_005812 was derived from COL6A2, which encodes a type of VI collagen. It has been observed that secretion of VI collagen in 3T3-L1 adipocytes significantly increased with the accumulation of lipid droplets. In principle, VI collagen promoted adipogenesis and lipid accumulation in 3T3-L1 preadipocytes by up-regulating expression of adipocyte-specific factors SREBP-1c, C/EBPβ, PPARγ2, and C/EBPα (Liu et al., 2017). In addition, CCDC88A circRNA (circ_001285) and MXRA5 circRNA (circ_000589) in profile 4 should also be paid attention to. CCDC88A has been known to be involved in phospholipid binding, and it can active PI3K-Akt signaling that was crucial for adipogenesis in adipocytes (Ghosh et al., 2011, Wang et al., 2020). MXRA5 was found to be involved in regulation of obesity of human (Arner et al., 2018).

Several circRNAs caught our attention as their parent genes were enriched in important biological process related to adipocytes differentiation, lipid and fatty acid synthesis. For example, IGF1R producing circ_005401 was involved in positive regulation of lipid biosynthetic and steroid biosynthetic process (P < 0.05). It was reported that steroid stimulated 3T3-L1 preadipocytes differentiation by targeting HDAC1 complex (Wiper-Bergeron et al., 2003). The up-regulated circ_004368 on days 2 and 8 of preadipocytes differentiation compared to day 0 was derived from AASS, which was involved in acetyl-CoA metabolic process. It was well known that acetyl-CoA metabolic was closely associated with fatty acid synthesis of animals (Yoshii et al., 2015). In addition, small GTPase binding that modulated the biogenesis of lipid droplets in 3T3-L1 preadipocytes, was enriched by the parent genes of DIAPH2 circRNA (circ_004915) and RANBP17 circRNA (circ_005540) (Martin and Parton, 2008; Tan et al., 2013).

In this study, several signaling pathways enriched by the parent genes of differentially expressed cirRNAs have also been reported in previous studies. For example, COL6A2 circRNA (circ_005812) and IGF1R circRNA (circ_005401) described above were also enriched in focal adhesion pathway (P < 0.05). The pathway has been reported to be involved in adipogenic differentiation of caprine intramuscular preadipocytes and was also enriched by the parent genes of differentially expressed circRNAs identified from different differentiation stages of preadipocytes in sheep (Xiao et al., 2021) and yak (Zhang et al., 2020b). Similarly, lysine degradation pathway enriched by the parent genes of AASS circRNA (circ_004368), KMT2E circRNA (circ_004757), and NSD2 circRNA (circ_002566) in the study, was also found to be enriched by the parent genes of circRNAs identified from adipocytes differentiation of yak (Zhang et al., 2020b). Additionally, in this study, the parent gene of GARS circRNA (circ_006408) was enriched in aminoacyl-tRNA biosynthesis pathway that promoted 3T3‐L1 preadipocytes differentiation by interacting with mTOR pathway (Guo et al., 2020).

The function of several differentially expressed circRNAs in ovine preadipocytes differentiation would also be reflected by their parent genes. For example, differentially expressed circ_005620 between days 0 and 2 is originated from zinc finger E-box binding homeobox 1 (ZEB1), which is a crucial transcription factor involved in bovine intracellular lipid metabolism and accumulation of mice adipose tissue (Saykally et al., 2009; Wang et al., 2021). The circ_002606 may regulate ovine adipocytes differentiation as knockdown of its parent gene MYH10 in mice adipocytes has been observed to significantly inhibit adipogenesis and reduce cell area (Kislev et al., 2022). The circ_000154 was down-regulated on day 8 compared to day 2, and its parent gene PUM2 has been reported to promote lipid accumulation in zebrafish mesenchymal stem cell (Lee et al., 2020).

Given the finding that cicr_004977 was not expressed on day 0, but its expression was significantly up-regulated on day 8, it was therefore inferred that circ_004977 may regulate late differentiation stage of ovine preadipocytes. It was noteworthy that miR-200b was involved in three ceRNA pairs associated with circ_004977, namely circ_004977-miR-200b-SLC6A20, circ_004977-miR-200b-Zdhhc15, and circ_004977-miR-200b-ANK1. Our previous study has confirmed that miR-200b plays a negative role in preadipocytes differentiation and lipid droplets accumulation of Tibetan sheep (Jin et al., 2021). Based on the finding that circRNAs may play an opposite role with miRNAs sponged by them, circ_004977 would therefore promote lipid synthesis in preadipocytes of Tibetan sheep. Similarly, circ_004915 specifically expressed on day 8 of preadipocytes differentiation, would also regulate lipid droplets accumulation as it was a potential molecular sponge of miR-200b. Other ceRNA pairs regulating adipocytes differentiation and adipogenesis included circ_006132-miR-106-5p/miR-17-5p/miR-20-5p/miR-150-FGF1 and circ_003788-miR-872-3p-OAS2. FGF1 has been reported to promote differentiation of caprine intramuscular and subcutaneous preadipocytes (Cui et al., 2023). This suggests that circ_006132 would upregulate the expression of FGF1 through abolishing the suppression effect of miR-106-5p/miR-17-5p/miR-20-5p/miR-150 on FGF1, eventually positively regulating ovine preadipocytes differentiation. The speculation was confirmed by our results that circ_006132 was significantly up-regulated on day 8 compared to day 0. In addition, the ceRNA network of circ_003788-miR-872-3p-OAS2 may also be involved in ovine preadipocytes differentiation as OAS2 was participated in lipid metabolism process (Piórkowska et al., 2018).

It was worth noting that several important circRNAs previously reported associated with adipocytes differentiation and proliferation were also found in this study. For example, circITGB1, which has been reported to promote proliferation and inhibit differentiation of ovine adipocytes, was identified as circ_002775 in this study (Supplementary Table S1). It was found that circ_002775 showed a lower expression level on day 2 of differentiation compared to on days 0 and 8. The result may reflect the roles of circITGB1 in proliferation and differentiation of adipocytes (Yue et al., 2023). Similarly, in a previous study, bovine circRNF111 has been confirmed to regulate adipocyte differentiation by elevating PPARγ expression via miR-27a-3p (Shen et al., 2023). The circRNA was also found in this study (named circ_003566; Table S1). These results indicate high conservatism of circRNAs across species.

This study investigated the potential roles of circRNAs in ovine adipocytes differentiation and adipogenesis, by analyzing the expression patterns of circRNAs at different differentiation periods. However, some of these results were based on the bioinformatics analysis or functional prediction. The specific function of these circRNAs need to be further investigated in vivo or in vitro in future.

Conclusion

This study identified 63 differentially expressed circRNAs during adipogenic differentiation of ovine preadipocytes. The circ_005812, circ_001285, circ_000589, circ_005401, circ_004368, circ_004368, circ_006408, circ_005620, circ_002606, and circ_00015 were found to be derived from important functional genes related to adipogenesis and lipid metabolism. In addition, circ_004977, circ_004915, circ_006132, and circ_003788 would function as molecular sponges of miRNAs to regulate preadipocytes differentiation. The study provides an improved understanding of the roles of circRNAs in adipogenic differentiation of sheep.

Glossary

Abbreviations

C/EBP

CCAAT/enhancer binding protein

ceRNA

competing endogenous RNA

circRNA

circular RNA

GO

Gene Ontology

KEGG

Kyoto Encyclopedia of Genes and Genomes

lncRNA

long non-coding RNA

miRNA

microRNA

PCC

Pearson correlation coefficient

PPARγ

peroxisome proliferator activated receptor gamma

RNA-seq

RNA sequencing

SCC

Spearman rank correlation coefficient

SREBP

sterol regulatory element binding protein

Conflict of Interest Statement

All authors declared no conflict of interest.

Author Contributions

Jiyuan Shen: Writing–original draft, Validation, Conceptualization. Xiayang Jin: Formal analysis. Zhiyun Hao: Investigation. Jiqing Wang: Conceptualization, Funding acquisition, Writing–review and editing. Jiang Hu: Methodology. Xiu Liu: Project administration. Shaobin Li: Resources. Fangfang Zhao: Software. Mingna Li: Software. Zhidong Zhao: Supervision. Bingang Shi: Validation. Chunyan Ren: Visualization.

Data Availability

The raw data of RNA sequencing is available at the NCBI GenBank Sequence Read Archive (SRA), under accession number PRJNA854878.