NORHA, A Novel Follicular Atresia-related Lncrna, Promotes Granulosa Cell Apoptosis via miR-183-96-182 Cluster and FoxO1 Axis


 BackgroundFollicular atresia has been shown to be strongly associated with low follicle utilization rate and female infertility, which are regulated by many factors such as miroRNAs (miRNAs), a class of non-coding RNAs (ncRNAs). However, little is known about long non-coding RNAs (lncRNAs), another ncRNAs, which regulate follicular atresia. ResultsA total of 94 differentially expressed lncRNAs, including 74 up-regulated and 20 down-regulated lncRNAs, were identified in early atretic follicles compared to healthy follicles by RNA-sequencing. We identified and characterized a non-coding RNA that was highly expressed in atretic follicles (NORHA), an intergenic lncRNA, was the most significantly elevated lncRNA in early atretic follicles. Functionally, RT-PCR，flow cytometry and western blot results showed that NORHA was associated with follicular atresia by influencing GC apoptosis. Mechanistically, bioinformatics analysis, luciferase reporter assay and RNA immunoprecipitation assay results showed that NORHA acted as a ‘sponge’, which directly bound to the miR-183-96-182 cluster, and therefore resisted their targeting inhibition of FoxO1, the major sensor and effector of oxidative stress. Furthermore, NORHA and oxidative stress synergistically induced GC apoptosis. ConclusionsWe provide a comprehensive perspective of lncRNAs regulation of follicular atresia, and demonstrate that NORHA, a novel lncRNA related to follicular atresia, induces GC apoptosis through affecting the miR-183-96-182 cluster and Foxo1 axis


Introduction
The overall genome, non-coding RNAs (ncRNAs) such as microRNAs (miRNAs), ribosomal RNAs (rRNAs), transfer RNAs (tRNAs), long non-coding RNAs (lncRNAs), piwi-interacting RNAs (piRNAs), and circular RNAs (circRNAs), represent a larger part of the transcriptome than coding genes [1][2][3]. LncRNAs refer to a subclass of RNA polymerase II transcripts with a length of not less than 200 nucleotides (nt) and weak or non-coding potential [4,5]. Compared with protein-coding genes and other ncRNAs, the lack of sequence conservation among species is a feature highlight of lncRNAs [6]. This confounded efforts to predict the sequences, functions, and mechanisms of action of lncRNAs across species [7]. For scientists, speciesspeci c lncRNAs are like mysterious treasure, so using high-throughput technology to uncover speciesspeci c lncRNAs is the focus of current research, and huge numbers of lncRNAs in various species have been identi ed [8]. Only a few lncRNAs have been functionally characterized, and as of 2018, only 156 lncRNAs had known biological functions [6,9]. However, their functions are very extensive, and have been reported as essential regulators in various cell biological processes, such as cell survival [10], differentiation [11], cell cycle [12], cell apoptosis [13], pluripotency [14], and susceptibility to infection [15].
In addition, unlike the relatively simple mechanism (RNA-RNA interaction) of action for miRNAs, the mechanism of function of lncRNAs is very diverse, and enables complex interactions with DNA, RNA, and proteins [9].
In the ovary, emerging data have demonstrated that lncRNAs are widely enriched in oocytes, various somatic cell types including granulosa cells (GCs), cumulus cells (CCs) and ovarian surface epithelial stem cells, and follicular uid [16,17]. These cells are associated with multiple ovarian functions, from maintenance and initiation of the primordial follicle pool to ovulation, corpus luteum formation, and ovarian aging [18][19][20][21]. Numerous lncRNAs were identi ed in somatic cells and oocytes at all stages of follicular development using high-throughput technology. For example, a total of 20,563 lncRNAs were identi ed in human CCs [22], and 4,926 differentially expressed lncRNAs (DELs) were identi ed in goat ovaries between the luteal and follicular phases [23]. However, only a few lncRNAs are well-characterized in terms of function and mechanism in humans and rodents [19,24]. Nuclear enriched abundant transcript 1 (NEAT1), for instance, a lncRNA implicated in many major physiological events [25,26]. NEAT1 is abundant in human metaphase II (MII) oocytes and mouse corpus luteum, and a Neat1 knockout (KO) mouse model revealed that luteal tissue formation was seriously diminished in nearly half of the Neat1 KO mice [16,19]. Furthermore, ovarian lncRNAs have been shown to be related to livestock fecundity [27]and human infertility [28]. In addition, dysregulation and dysfunction of lncRNAs will result in ovarian diseases such as premature ovarian insu ciency (POI), polycystic ovary syndrome (PCOS), ovarian cysts, and cancer [24,29].
In summary, there are many lncRNAs related to ovarian function identi ed by high-throughput methods [22,23]; however, the function and molecular mechanism of only a small subset of lncRNAs are known. In this study, we performed RNA-seq to investigate lncRNA expression pro les in healthy follicles (HF) and early atretic follicles (EAF) of pig ovaries. Furthermore, we investigated the role and mechanism of NORHA, a lncRNA that was the most up-regulated in atretic follicles, to better understand the regulation of lncRNAs in follicular atresia.

Follicles collection and classi cation
Antral follicles with a diameter of 3-5 mm were procured from ovaries of mature sows, and then classi ed into HF and EAF group based on the ratio of progesterone (P4)/17β-estradiol (E2) concentration in follicular uid, the GC density in follicular uid, and the morphological features of follicles as previously described [30]. All animal experiments follow the ARRIVE guidelines and approved by the Ethical Committee of Nanjing Agriculture University.

RNA isolation and Sequencing
TRIzol reagent (Invitrogen, Shanghai, China) was used to isolate total RNA from follicles. The quality and integrity of total RNA were evaluated using an Agilent 2100 Bioanalyzer (California, USA) and agarose gel electrophoresis. The integrity score was no less than 7.0, and the ratio of 28S/18S of ribosomal RNA greater than 0.7 was set as the acceptable standard. After removing rRNA, total RNA of six healthy follicles or six early atretic follicles were constructed into one RNA-seq library. Paired-end sequencing was performed using an Illumina HiSeq2000.

Sequencing data analysis
The genome mapping of reads was performed by the TopHat v2.0.9 software (https://ccb.jhu.edu/software/tophat/index.shtml). The mapped reads were assembled by using a Cu inks tool (http://cole-trapnell-lab.github.io/cu inks/). Cuffcompare program (http://cole-trapnelllab.github.io/cu inks/cuffcompare/) was used to compare candidate sequences with known lncRNAs, and the differential expression analysis was performed by DESeq2 in the R package (https://www.rproject.org/). DELs was screen by the following criteria: (i) the length of transcripts ≥ 200 bp; (ii) the expression value with FRPKM > 0; (iii) the transcripts were predicted as a non-coding RNA in pig reference genome database (Sus scrofa 11.1); (iv) The screening criteria of DELs were established by a P-value ≤0.05 and a |logarithm to the fold-change in the expression of two| ≥ 0.5. Chromosomal locations, gene blast, annotated transcripts and sequence mapping were visualized on the genome data from NCBI database (https://www.ncbi.nlm.nih.gov/gene). Two online tools CPC (http://cpc.cbi.pku.edu.cn/) and CAPT (http://lilab.research.bcm.edu/cpat/index.php) were used to evaluate the coding potential of transcripts.

Functional enrichment analysis
To predict the potential functions of DELs, the cis-target mRNAs, which were selected within the range of 200 kb in the upstream and downstream of DELs, were annotated and classi ed by Gene ontology (GO) and Kyoto encyclopedia of genes and genome (KEGG) pathway analysis using DAVID Bioinformatics Resources v6.7 (https://david-d.ncifcrf.gov/).

Quantitative real-time RT-PCR
Total RNAs from tissues, follicles or GCs were extracted as described in RNA-sequencing, and the rststrand cDNA was synthesized from 500 ng of total RNA using the PrimeScript™ RT Master Mix (TaKaRa, Dalian, China). Differently, the reverse transcription of miRNAs was performed by TransScript miRNA First-Strand cDNA Synthesis SuperMix (TransGen, Beijing, China). RT-PCR was performed as described previously [31], and the primers are shown in Table S6.
Rapid ampli cation of cDNA ends The 5'-and 3'-rapid ampli cation of cDNA ends (RACE) was performed to obtain the full-length of lncRNA. The rst-strand cDNA was synthesized by using the SMARTer RACE cDNA Ampli cation Kit (TaKaRa, Dalian, China). PCR was performed to amplify cDNA product from the 5' end or 3' end of NORHA with a gene-speci c primers shown in Table S6.

Cell culture and treatment
GCs and HEK293T cells were cultured as described previously [31]. After the density reached 70-80%, cells were transfected by the Lipofectamine 3000 reagent (Invitrogen, Shanghai, China) following the manufacturer's speci cations. For H 2 O 2 treatment, the medium was replaced with serum-free medium, different concentrate H 2 O 2 (30 % [w/w] in H 2 O, Sigma, Shanghai, China) were added to the medium for 90 min.

Plasmid construction
To construct the overexpression plasmid pcDNA3.1-NORHA, the full length of NORHA RNA was ampli ed from cDNA of porcine GCs by using a speci c primer (Table S6) with HindIII and XhoI enzymes sites. The PCR product was digested with restriction enzymes and then inserted into the pcDNA3.  Table S6.

Apoptosis analysis
GCs apoptosis rate were measured by Apoptosis Detection Kit (Vazyme, Nanjing, China). Brie y, GCs were stained with 5 μl of Annexin FITC V and 5 μl of PI solution for 10 min, and the apoptosis rate was counted by BD FACScan ow cytometry (Becton Dickinson, Franklin, NJ, USA).
Dual-luciferase reporter assay Luciferase activities were detected by Luciferase Assay Buffer II and Stop & GLO Substrate with Modulus™ Assay System (Turner Biosystems, San Francisco, CA, USA) as described previously [31].
RNA immunoprecipitation (RIP) assay RIP experiment was performed by using the EZ Magna RIP kit (Millipore, Billerica, MA, USA). The cells were pelleted and lysed in complete RIP lysis buffer containing protease inhibitor cocktail and RNase inhibitor. Homogenates were re-suspended to a single cell suspension and store at -80℃. Magnetic beads were washed several times with RIP wash buffer and then labelled with antibodies for 30 min. The extract was incubated with magnetic beads linking to AGO2 (CST, BMA, USA) or IgG antibodies (Santa Cruz, CA, USA) for overnight at 4 ˚C with head-to-head rotation. Finally, RT-PCR was performed to detect the expression of NORHA by speci c primers shown in Table S6.

ROS detection
The ROS detection was performed by Reactive Oxygen Species Assay Kit (KeyGen, Shanghai, China). Brie y, GCs were incubated with DCFH-DA (diluted at 1:1000 in serum-free medium ) for 20 min in 37 ℃, and the uorescence of DCF in cells was detected by using BD FACScan ow cytometry (Becton Dickinson, Franklin, NJ, USA).

Statistical analysis
Statistical analysis was performed by using Graphpad Prism v5.0 software (San Diego, CA, USA). The level of signi cance was set to 0.05 and used the Student's t-test or one-way analysis of variance test. The correlation detection was analyzed using the Pearson test model. A P value of < 0.05 was considered to be associated, and the value of r represents the levels of correlation.

Genome-wide identi cation of follicular atresia-related lncRNAs
To investigate the role of lncRNAs during follicular atresia, HF and EAF were isolated from porcine ovaries according to the P4/E2 ratio, the GC density, and the morphological features of follicles (Fig. 1a); and RNA-seq was used to construct the expression pro les of lncRNAs. A total of 10,066 transcripts and 1,918 lncRNAs were identi ed in follicles, 1,807 and 1,899 lncRNAs were detected in HF and EAF, and 1,574 lncRNAs were found in both HF and EAF (Fig. 1b, Table S1). The chromosomal distribution of follicular lncRNAs was not uniform, and the number of lncRNAs transcribed from chromosome 1 was the highest, while chromosome 10 was the lowest (Fig. 1c). Notably, a total of 94 DELs, including 74 up-regulated and 20 down-regulated lncRNAs, were identi ed in EAF compared to HF (Fig 1d, Table S2). To con rm the RNA-seq data, 7 DELs were selected for quanti cation by qRT-PCR. The data showed that the qRT-PCR of 6 DELs con rmed the accuracy of the RNA-seq data (Fig. 1e).
A total of 492 cis-target mRNAs of DELs were identi ed (Table S3), and GO analysis revealed that a total of 23 signi cant GO terms were enriched, for example, the positive regulation of apoptotic process, negative regulation of cell proliferation, sequence-speci c DNA binding, and nucleoplasm ( Fig. S1, Table  S4). Besides, KEGG pathway analysis revealed that multiple signi cant pathways were enriched, such as the MAPK signaling pathway, oxytocin signaling pathway, DNA replication, and Insulin secretion (Fig. S2, Table S5).

NORHA was a novel cytoplasmic lncRNA involved in follicular atresia
Notably, in the rst ve DELs up-regulated during follicular atresia, qRT-PCR showed that LOC102167901 was the most signi cantly elevated DEL, which was selected for further analysis. We rst isolated the fulllength RNA sequence of porcine LOC102167901 from porcine ovaries by performing a RACE (Figs. 2a, and S3a). The full-length LOC102167901 novel transcript was 1,566 bp ( Fig. S3b), which was different from the original sequences of LOC102167901 in the GenBank database (XR_304632, 7082 bp, predicted). The protein-coding ability scores of the LOC102167901-derived novel transcript were 0.067 (CPAT method) and -0.92330 (CPC method), which was close to that of other known lncRNAs (e.g., MALAT1 and H19) (Figure 2b, c), indicating that the novel transcript was devoid of protein-coding potential; it was a real lncRNA. We therefore termed this novel lncRNA as a non-coding RNA that was highly expressed in atretic follicles or NORHA.
NORHA is a sense lncRNA located in a region from 100135521 nt to 100137345 nt on the pig chromosome 7, and consists of two exons and one intron (Fig. 2a). The expression pattern revealed that NORHA was highest expression in ovary, but very weak in heart (Figs. 2d, and S4). The subcellular location showed that NORHA was enriched in the cytoplasm of porcine GCs (Fig. 2e). In ovarian follicles, NORHA levels were positively correlated with the P4/E2 ratio (r=0.3858, P < 0.05), a biomarker for follicular atresia, indicating that NORHA was involved in porcine follicular atresia (Fig. 2f). Taken together, we identi ed a novel cytoplasmic lncRNA, NORHA, in GCs and demonstrated that this lncRNA was involved in porcine follicular atresia.

NORHA induced GC apoptosis
In pigs and other mammals, follicular atresia has been shown to be triggered by GC apoptosis [31,32]. To understand the relationship between NORHA and GC apoptosis in vivo, we examined the levels of NORHA and Caspase3, a pro-apoptotic marker for porcine GC apoptosis in follicles, and found that NORHA levels were positively correlated with Caspase3 mRNA levels (r=0.5845, P < 0.01) (Fig. 3a). To further determine the role of NORHA in porcine GC apoptosis, we overexpressed and silenced endogenous NORHA in GCs by treatment with pcDNA3.1-NORHA vector and NORHA-siRNA (Fig. 3b). Flow cytometry showed that NORHA overexpression increased the apoptosis rate of GCs (Fig. 3c), whereas silencing of NORHA decreased the apoptosis rate of GCs (Fig. 3d). In addition, cleaved Caspase3 (C-CASP3) protein levels were increased in GCs after overexpression of NORHA (Fig. 3e), whereas decreased in GCs after knockdown of NORHA (Fig. 3f). These results suggested that NORHA was a pro-apoptotic lncRNA in porcine GCs.
NORHA was a molecular sponge of the miR-183-96-182 cluster The main mode of action of cytoplasmic lncRNA is to function as a "miRNA sponge" [33,34]. To explore the molecular mechanism underlying cytoplasmic NORHA enhancement of porcine GC apoptosis, the coexpression network of NORHA and miRNAs (unpublished) during follicular atresia was constructed. The results suggested that NORHA potentially interacted with 21 miRNAs that were differentially expressed during follicular atresia (Fig. 4a). Interestingly, the miR-183-96-182 cluster, which are transcripts from a common genomic region (nt 18982506 -nt 18982590) on pig chromosome 18 (Figs. 4b, and S5), were predicted to bind to NORHA using RNAhybrid (Fig. 4c, d), indicating that the lncRNA NORHA was a putative molecular sponge of the miR-183-96a-182 cluster in sows.
To con rm whether NORHA bound to the miR-183-96-182 cluster, we generated a dual-luciferase reporter vector harboring the response element of the miR-183-96-182 cluster (Fig. 4e), and co-transfected it into HEK293T cells with miR-183, miR-96, or miR-182 mimics, respectively. Luciferase assays revealed that these miRNAs cluster signi cantly decreased the luciferase activity of the reporter vector (Fig. 4f). However, the three miRNAs had no effect on the luciferase activity of the reporter construct with a mutated binding site (Fig. 4g). These results suggested that NORHA could bind to the miR-183-96-182 cluster. RIP assay shown that the NORHA were enriched in AGO2, a member of RISC (RNA-induced silencing complex) (Fig. 4h), indicating that NORHA had the potential to bind to the miR-183-96-182 cluster. Furthermore, the miR-183-96-182 cluster was up-regulated in NORHA-silenced GCs (Fig. 4i). Together, our data suggested that NORHA directly interacted with, and could sponge, the miR-183-96-182 cluster in porcine GCs. miR-183-96-182 cluster inhibited GC apoptosis Our previous miRNA pro le (unpublished) showed that the levels of three members of the miR-183-96-182 cluster were decreased in porcine EAF compared to HF (Fig. 5a). We then con rmed this using RT-PCR (Fig. 5b), indicating that this cluster participated in porcine follicular atresia. To evaluate the function of the miR-183-96-182 cluster on follicular atresia, miRNA mimics were used to exogenously overexpress the levels of these miRNAs in GCs. Overexpression of the miR-183-96-182 cluster obviously decreased the percentage of apoptotic cells (Fig. 5c), and markedly suppressed levels of C-CASP3 (Fig. 5e), indicating that the miR-183-96-182 cluster inhibited porcine GC apoptosis. In contrast, the percentage of apoptotic cells (Fig. 5d) and C-CASP3 levels (Fig. 5f) were signi cantly increased in the miR-183-96-182 clustersilenced GCs. Our results suggested that the miR-183-96-182 cluster could impede GC apoptosis and follicular atresia in sows.

FoxO1 was a common target of the miR-183-96-182 cluster in GCs
To explore the mechanism of the miR-183-96-182 cluster prevention of GC apoptosis, we next identi ed their targets. A total of 148, 383, and 411 potential targets were predicted for miR-183, miR-96, and miR-182, respectively (Fig. S6a). Seventeen genes were commonly targeted by the miR-183-96-182 cluster (Fig. 6a). Of these, FoxO1, a core member of the forkhead box O (FoxO) family, which plays an extremely important role in GC apoptosis in mammals [35,36], was selected as a candidate target of the miR-183-96-182 cluster for further study. RNAhybrid showed that the FoxO1 3'-UTR of mammals, including pigs, not only had a high interaction capacity with the miR-183-96-182 cluster (Fig. 6b), but also contained a miRNA response element (MRE) of the miR-183-96-182 cluster at 2261 nt -2267 nt (GenBank ID: NM_214014) (Figs. 6c, and S6b), indicating that FoxO1 was a potential target of the porcine miR-183-96-182 cluster.
To determine whether the miR-183-96-182 cluster directly targeted the porcine FoxO1 3'-UTR, we constructed a pmirGLO dual-luciferase reporter vector of the porcine FoxO1 3'-UTR containing the MRE motif ( Figure S7a). HEK293T cells were transiently co-treated with this reporter plasmid and the miRNA mimics. The luciferase activity of the reporter vectors were signi cantly attenuated in cells treated with the miRNA mimics (Fig. 6d), but the luciferase activity of the MRE-mutated reporter vectors had no altered (Figs. 6e, and S7b), indicating that FoxO1 was a direct target of the miR-183-96-182 cluster in pigs. We next investigated whether the miR-183-96-182 cluster controlled FoxO1 expression in porcine GCs. As shown in Fig. 6f, g, the miR-183-96-182 cluster signi cantly restrained FoxO1 expression in GCs, at both the transcription and translation levels. Together, these results suggested that FoxO1 was a direct and functional target of the miR-183-96-182 cluster in porcine GCs.
NORHA, the miR-183-96-182 cluster, and FoxO1 are involved in GC apoptosis, which has been con rmed by the above results and other studies [35,36]. We therefore investigated whether NORHA affected FoxO1-mediated cell apoptosis by competing for the miR-183-96-182 cluster in porcine GCs. Overexpression of FoxO1 induced GC apoptosis, whereas knockdown of FoxO1 reduced GC apoptosis, indicating that FoxO1 promoted porcine GC apoptosis (Fig. 7f), as observed in mice [35,36]. Furthermore, we showed that FoxO1-induced GC apoptosis could be reversed by miRNA-182, and this process could also be inhibited by NORHA (Fig. 7g, h). These analyses suggested that NORHA induced FoxO1-mediated cell apoptosis by competing for the miR-183-96-182 cluster in porcine GCs.

NORHA and oxidative stress synergistically induced GC apoptosis
FoxO1 is widely regarded as a major sensor and effector of oxidative stress; oxidative stress is a strong inducer of GC apoptosis in mammals [35,37]. We therefore hypothesized that NORHA were involved in oxidative stress-induced GC apoptosis. To validate this hypothesis, we rst investigated the effects of oxidative stress driven by H 2 O 2 treatment on porcine GC apoptosis. As shown in Fig. 8a, H 2 O 2 can induce an increase in ROS levels in GCs. Meanwhile, the cell apoptosis rate and C-CASP3 levels were upregulated in GCs with continuing H 2 O 2 stimulation (Fig. 8b, c), indicating that short-term oxidative stress driven by H 2 O 2 treatment promoted porcine GC apoptosis. Furthermore, overexpression of NORHA induced the apoptosis rate in GCs with H 2 O 2 exposure (Fig. 8d), revealing that NORHA and oxidative stress could synergistically induce GC apoptosis.

Discussion
A large number of primordial follicles exist in the ovarian follicle pool in mammals; approximately 420,000 in pigs and 400,000 in humans. However, less than 1% of follicles mature and ovulate, while most follicles are atretic and degenerate [38]. Therefore, follicular atresia not only restricts the effective utilization of the primordial follicle pool, but also limits the potential reproductive ability of domestic animals. Many factors (e.g., follicle-stimulating hormone (FSH), Liver receptor homolog-1 (LRH-1), and Xlinked inhibitor of apoptosis (XIAP)) and signaling pathways (e.g., death receptor-mediated, follicle stimulating hormone receptor (FSHR), and TGF-β signaling pathways) related to follicular atresia have been identi ed [31,35,[39][40][41][42]. In recent years, high-throughput technology has gradually become an important way to fully understand the molecular events during follicular atresia, and the expression pro le of miRNAs [30] and mRNAs [43] in follicular atresia has been revealed. Here, we pro led lncRNA expression patterns during follicular atresia using high-throughput RNA-seq, and 94 DELs were identi ed. Regulation of adjacent or host genes via cis regulation is a main functional model of lncRNAs [44,45]. Notably, among the cis-target mRNAs of these DELs, multiple genes, such as CYP11A1 [46], BABAM2 [47], broblast growth factor 18 (FGF18) [48], and neurogenic locus notch homolog protein 2 (NOTCH2) [49] have been implicated in GC functions (e.g., steroid hormone synthesis), follicular atresia, ovarian functions, and female fertility. Our ndings are the rst to identify lncRNAs during follicular atresia and provide an understanding of lncRNAs that are involved in follicular atresia.
We further demonstrated that NORHA, the most up-regulated lncRNA during porcine follicular atresia, was strongly involved in porcine follicular atresia, through enhancing GC apoptosis. In GCs, lncRNAs have been reported to be associated with various cellular functions, such as proliferation and cycle [24,28], and the secretion of steroid hormones, including E2, P4, and testosterone (T) [50,51]. However, few studies have investigated the regulation of GC apoptosis by lncRNAs. A recent study showed that lncRNA steroid receptor RNA activator (SRA), an important player in transcriptional regulation, is thought to interact with a DNA-binding protein by binding to speci c DNA sequences [52], which induces the release of E2 and P4, and reduces cell apoptosis in mouse GCs [51]. The Prader-Willi region non-protein coding RNA 2 (PWRN2), a CCs-expressed lncRNA, is thought to be associated with oocyte nuclear maturation through sponging miR-92b-3p in the human ovary [53]. In addition, in ovarian cancer cells, some lncRNAshave also been demonstrated to be involved in various cellular processes [29]. MLK7-AS1, for instance, an lncRNA that is speci cally upregulated in ovarian cancer tissues, which controls multiple cellular processes (e.g., it sti es cell invasion, proliferation, wound healing, and advances cell apoptosis), modulates the epithelial-mesenchymal transition (EMT) process by in uencing the miR-375/Yesassociated protein 1 (YAP1) axis [29]. Taken together, our ndings are the rst to identify and characterize lncRNAs associated with follicular atresia, and provide evidence that NORHA could serve as a potential diagnostic biomarker and rescue target for follicular atresia.
As an important regulatory RNA, lncRNAs exert their biological function mainly by regulating target expression at various levels, from transcription to protein localization, and stability [9]. The subcellular localization (cytoplasm and/or nucleus) of lncRNAs is the principal determinant of their molecular function and mode of action [54]. The most common mechanism of action of both cytoplasmic and nuclear lncRNAs is to function as a ceRNA, regulating target gene expression via an lncRNA-miRNA-target model [9,33,54]. For instance, lncRNA-protein phosphatase 1 nuclear-targeting subunit (PNUTS), a noncoding isoform of the protein-coding gene PNUTS, serves as a ceRNA of miR-205 to in uence the EMT migration and invasion by controlling the miR-205/ZEB/E-cadherin axis [55]. More recently, temozolomide -associated lncRNA (TALC), a highly expressed lncRNA in temozolomide-resistant glioblastoma, has been demonstrated to function as a ceRNA to competitively bind miR-20b-3p to enhance c-Met expression [33].
TALC then activates the Stat3/p300 complex to increase the transcriptional activity of the O 6methylguanine-DNA methyltransferase by modulating the acetylation of H3, including H3K9, H3K27, and H3K36 [33]. Here, based on the NORHA-miRNA interactive network, we detected a putative binding site of the miR-183-96-182 cluster in NORHA, and demonstrated that NORHA functioned as a ceRNA for the miR-183-96-182 cluster, relieving its inhibition of the target FoxO1 and promoting cell apoptosis in porcine GCs.
The miR-183-96-182 cluster (or miR-183 cluster) is a polycistronic miRNA cluster, which is located within a 5-Kb genomic region on chromosome 7 in humans, chromosomes 6 in mice, and chromosome 18 in pigs. Importantly, this family is not only highly conserved among different species, but also has similar seed sequences among different members of the same species, implying that the members of this cluster may share the same targets and biological functions [56,57]. Consistent with this hypothesis, recent reports showed that the miR-183-96-182 cluster common targets are HDAC9, which encodes a histone deacetylase that in uences memory formation [58], Cacna2d2, which encodes the auxiliary voltage-gated calcium channel subunits α2δs to scale mechanical pain sensitivity [56], and DAP12 and Nox2, to control macrophage functions in response to Pseudomonas aeruginosa [59]. In addition, increased the miR-183-96-182 cluster in luteal relative to follicular tissues could enhance cell survival and P4 release in ovarian luteal cells of both bovine and humans by targeting bovine and human FoxO1, respectively [60]. Herein, we demonstrated that the miR-183-96-182 cluster was down-regulated in atretic follicles, inhibited the common target FoxO1 gene and mediated NORHA regulation of cell apoptosis in porcine GCs. In the ovary, the miR-183-96-182 cluster was mainly expressed in both follicular GCs and the corpus luteum, which plays a vital role in cell proliferation and the cell cycle of bovine GCs [61]. Together, our work indicated that the miR-183-96-182 cluster was related to follicular atresia by repression of GC apoptosis.
Determination of the mechanism of the miR-183-96-182 cluster repression of GC apoptosis led us to identify FoxO1 as a functional target of the miR-183-96-182 cluster, which mediated the anti-apoptotic function of the miR-183-96-182 cluster in GCs. FoxO1, a member of the FOXO family, has been shown to be a direct target of dozens of miRNAs, including the miR-183-96-182 cluster in mammals [35,57]. As a functional target of the miR-183-96-182 cluster, FoxO1 participated in the miR-183-96-182 cluster regulation of multiple functions in various cells and tissues, such as cell death in endometrial cancer [62], pathogenicity in Th17 cells [57], adipogenesis in C2C12 myoblasts [63], and sperm quality in mouse testes [64]. Notably, in ovarian GCs, FoxO1 also mediates the miR-183-96-182 cluster regulation of GC functions [61]. In bovine GCs, for instance, FoxO1 is inhibited by the miR-183-96-182 cluster and can abate the proportion of cells in the S phase [61]. In addition, FoxO1 is thought to be associated with other GC functions, such as autophagy [34], apoptosis [35], proliferation [61] and differentiation [65], and response to FSH [66].

Conclusion
Collectively, our work detected alterations in lncRNA expression in health and early atretic follicles, and identi ed a novel lncRNA, NORHA, which was highly expressed in atretic follicles. NORHA induced follicular atresia and GC apoptosis via a miR-183-96-182 cluster/FoxO1 axis by competitively sponging the miR-183-96-182 cluster (Fig. S8). Our ndings uncovered new epigenetic mechanisms of follicular atresia and GC apoptosis, providing evidence that NORHA could serve as a potential diagnostic biomarker and rescue target for follicular atresia, as well as a novel candidate for the improvement of female fertility.

Competing interests
The authors declare that they have no competing interests.

Ethics approval and consent to participate
All animal experiments follow the ARRIVE guidelines and approved by the Ethical Committee of Nanjing Agriculture University.

Consent for publication
Not applicable     NORHA is a molecular sponge of the miR-183-96-182 cluster. a An interaction network between NORHA and miRNAs that down-regulated in EAF. The data of miRNA-seq were unpublished. The red triangle represents NORHA that was up-regulated in EAF, and blue ellipses represent miRNAs that down-regulated in EAF. b Chromosome location of miR-183, miR-96, and miR-182 show that they are a miRNA cluster from a common genome region. c Schematic showing the interactions of miR-183, miR-96, and miR-182 seed sequences, with NORHA RNA sequence. d Minimum free energy (MFE) of miR-183, miR-96, and miR-182 binding to NORHA were predicted by RNAhybrid. e Schematic showing the reporter constructs of NORHA. NORHA sequence containing wild-typed and mutant-typed miRNA binding site was ampli ed and inserted into prim-GLO plasmid to construct luciferase report vector. f and g Wild-(f) or mutant-typed (g) report vector were co-transfected with miR-183 mimics (mi), miR-96 mimics (mi), and miR-182 mimics (mi) into HEK293T cells, luciferase activity was measured. h RIP assay by AGO2-speci c antibody. The levels of NORHA enriched on AGO2 protein were detected by qRT-PCR. IgG antibody was used as a negative control. i The levels of miR-183, miR-96 and miR-182 in GCs knockdown of NORHA. Data are represented as mean ± S.E.M. with at last three independent experiments. * P<0.05. ** P<0.01. ns, not signi cant.    were co-transfected with mimics (mi) of miR-183, miR-96, or miR-182, and pcDNA3.1-NORHA, protein levels of FoxO1 were measured by western blot. f FoxO1 contributes to GC apoptosis. pcDNA3.1-FoxO1 (upper part), or FoxO1-siRNA (lower part) were transfected into GCs, and apoptosis rate was measured by ow cytometry. g, h NORHA induces GC apoptosis via the miR-182/FoxO1 axis. GCs were co-transfected with pcDNA3.1-FoxO1, miR-182 mimics (mi) and pcDNA3.1-NORHA, apoptosis rate (g), and protein levels of T-CASP3 and C-CASP3 (h) were detected. Data are represented as mean ± S.E.M. with at last three independent experiments. * P<0.05. ** P<0.01.

Figure 8
NORHA and oxidative stress synergistically induce GC apoptosis. a H2O2 induces oxidative stress.
Porcine GCs were treated with H2O2 (0, 500, and 1000 μM) for 90 min, and reactive oxygen species (ROS) levels were detected by ow cytometry. b, c Oxidative stress induced by H2O2 induces GC apoptosis. The apoptosis rate (b), and protein levels of T-CASP3 and C-CASP3 (c) were measured in GCs with 1000 μM H2O2 treatment. d NORHA and oxidative stress synergistically induced GC apoptosis. GCs were treated with pcDNA3.1-NORHA with or without H2O2 treatment, and the apoptosis rate was calculated. Data are represented as mean ± S.E.M. with at last three independent experiments. * P<0.05. ** P<0.01.

Supplementary Files
This is a list of supplementary les associated with this preprint. Click to download.