A Mutation in Endogenous saRNA miR-23a Influences Granulosa Cells Response to Oxidative Stress

Phenotypes are the result of the interaction between the gene and the environment, so the response of individuals with different genotypes to an environment is variable. Here, we reported that a mutation in miR-23a influences granulosa cells (GCs) response to oxidative stress, a common mechanism of environmental factors affecting female reproduction. We showed that nuclear miR-23a is a pro-apoptotic miRNA in porcine GCs through the activation of the transcription and function of NORHA, a long non-coding RNA (lncRNA) induces GC apoptosis and responses to oxidative stress. Mechanistically, miR-23a acts as an endogenous small activating RNA (saRNA) to alter histone modifications of the NORHA promoter through the direct binding to its core promoter. A C > T mutation was identified at −398 nt of the miR-23a core promoter, which created a novel binding site for the transcription factor SMAD4 and recruited the transcription repressor SMAD4 to inhibit miR-23a transcription and function in GCs. Notably, g.−398C > T mutation in the miR-23a promoter reduced GCs response to oxidative stress. In addition, g.−398C > T mutation was significantly associated with sow fertility traits. In short, our findings preliminarily revealed the genetic basis of individual differences in the response to oxidative stress from the perspective of a single mutation and identified miR-23a as a candidate gene for the environmental adaptation to oxidative stress.


Introduction
miRNAs are a class of small RNA molecules consisting of 21-25 nucleotides (nt) that are widely conserved in evolution. Since the discovery that lin-4, the first miRNA, controls temporal pattern formation in Caenorhabditis elegans, miRNAs have been proved to be involved in almost all physiological and pathological processes through a mechanism known as RNA interference (RNAi) [1,2]. Briefly, within the cytoplasm, mature miRNAs are first loaded into miRNA-induced silencing complexes (miRISCs), a ribonucleoprotein complex, and then bind to miRNA response elements (MREs) in the 3 untranslated regions (UTRs) of mRNAs by base pairing, which in turn initiates mRNA decay or repress translation to post-transcriptionally repress target expression [3]. Notably, in addition to the cytoplasm, some mature miRNAs have recently been detected in the nucleus [4,5]. Interestingly, nuclear mature miRNAs can activate rather than repress the transcription of target mRNAs through a new mechanism termed RNA activation (RNAa) [6]. Briefly, nuclear mature miRNAs act as an endogenous small activating RNA (saRNA) directly interacting with MRE motifs in the promoter region of target mRNAs by base pairing in order to alter histone modifications and recruit RNA polymerase II, finally affecting the transcriptional activity of target mRNAs [6][7][8]. However, only a few miRNAs have been identified as controlling the transcription of target mRNAs through an RNAa mechanism.
Single nucleotide variants (SNVs) are one of the most common types of heritable variations. SNVs in the key protein-coding genes are usually associated with health and Waltham, MA, USA), and its quality was checked by a microspectrophotometer (Thermo, Waltham, MA, USA), ensuring that the values of OD280/OD260 were between 1.9 and 2.0. Then, 500 ng of total RNA was reverse-transcribed into cDNA by using the HiScript II Q Select RT SuperMix (R232-01, Vazyme, Nanjing, China) according to the manufacturer's protocol. The cDNA was diluted to appropriate concentrations for qRT-PCR on a StepOnePlus System (Applied Biosystems, Waltham, MA, USA) for a total of 40 cycles by using the SYBR Green Master Mix Kit (Q111-02/03, Vazyme, Nanjing, China). The reaction system was 10 µL, containing 0.2 µL each of the upstream and downstream primers (primer concentration 10 µmol/L) and 10% cDNA, which contained a denaturation temperature of 95 • C and an annealing temperature of 60 • C. Among them, the determination of primer efficiency and the PCR amplification efficiency was achieved by a series of dilutions of the cDNA template ranging from 95% to 105%, and the presence of a single peak in 80-90 • C confirmed the specificity of the PCR amplification product when the melting curve analysis was performed after qRT-PCR. U6 was used as an internal reference for the miRNA, and GAPDH was used as an internal reference for the other genes [12]. Finally, the relative expression level was worked out by the 2 −∆∆CT method. The primers for the RT-PCR are shown in Table S2.

Flow Cytometry
After transfection for 48 h, 5 × 10 5 cells were harvested after trypsinization (25200072, Thermo, Waltham, MA, USA) and were washed twice with pre-chilled PBS. The staining was performed by adding 100 µL Binding Buffer, 5 µL annexin V, and 5 µL Propidium Iodide (PI) (A211-01/02, Vazyme, Nanjing, China) stain for 10 min. The samples were finally diluted by adding 400 µL Binding Buffer and were sequentially detected using a flow cytometer (Becton Dickinson, Franklin Lakes, NJ, USA). The results were analyzed using flowjo V7.6 software. The data from 10,000 cells were collected for each data file and analyzed by dividing the complete granule cell community using a cross shaped gating, guaranteeing the same gating settings between the treatment groups.

Subcellular Localization of miR-23a
First, 5 × 10 5 pGCs were harvested after trypsinization and lysed in 500 µL lysis buffer (PBS, 0.1% NP-40, 10 mmol/L RNase inhibitor) for 10 min on ice. Then, the cells were centrifuged at 5000× g for 3 min at 4 • C. The supernatant was collected for cytoplasmic RNA extraction. The remaining particles were lysed and centrifuged again, and then the particles were collected again for nuclear RNA extraction. The nuclear and cytoplasmic RNA were validated using U6 and GAPDH, respectively, and the nuclear and cytoplasmic RNA were used for the subcellular localization analysis of miR-23a using qRT-PCR.

Plasmid Construction and Dual-Luciferase Reporter Assays
The overexpression plasmids pcDNA3.1-SMAD4 and pcDNA3.1-NORHA were previously constructed [19,20]. For the plasmid pcDNA3.1-NR2C1, the coding sequence (CDS) region of NR2C1 was amplified from the cDNA of pGCs by using primers containing enzymatic digestion sites and cloned into a pcDNA3.1-basic vector (Promega, Madison, WI, USA). To construct luciferase reporters, the ssc-miR-23a promoter with four deletion fragments (−1446/−101, −1118/−101, −658/−101, and −289/−101) and different alleles for SNV was amplified from porcine genomic DNA by primers containing enzymatic digestion sites and cloned into a pGL3-basic vector (Promega, Madison, WI, USA). The vector and amplified products were digested with both NheI and XhoI for 2 h at 37 • C, purified, and ligated with T4 ligase (EL0016, Thermo, Waltham, MA, USA) at 16 • C overnight. The ligation products were transformed into competent cells (TSC-C01, Tsingke, Beijing, China) and plated, and single colonies to be grown were sequenced. Reporters of NORHA containing MRE-wt and MRE-mut were constructed by Tsingke (China). The primers are shown in Table S3. pGCs were assayed for luciferase activity using the dual-luciferase reporter assay system after 24 h of transfection (Promega, Madison, WI, USA).

ChIP
First, 1 × 10 7 pGCs were cross-linked in 1% formaldehyde for 10 min and sonicated 13 times (10 s each) with a microneedle on ice with an output control setting of 30%. The chromatin solution was collected after the ultrasound and half was left as the positive control (input). Then, the remaining ultrasound treated chromatin was diluted 2.5 times and incubated with 40 protein A/G agarose at 4 • C for 1 h. The supernatant was then collected and incubated with an anti-SMAD4 antibody (D120124, Biotechnology), anti-H3K9me2 antibody (A2359, ABclonal), anti-H3K4me3 antibody (A2360, ABclonal), and anti-Ago2 antibody (A6023, ABclonal), respectively, at 4 • C overnight to pull down the DNA-protein complex. After de-crosslinking, DNA fragments were detected for enrichment by qRT-PCR. The antibody against IgG (D110058, Biotechnology) was taken as the internal control, and the unprocessed chromatin was taken as the input control. The primers for ChIP are shown in Table S4.

H 2 O 2 Treatment
For the H 2 O 2 treatment, the culture medium was first changed by DMEM/F12 without FBS for 6 h, and then H 2 O 2 (Sigma, shanghai, China) was added into the medium with the final concentration at 200 µM for 2 h [21].

DNA Extraction, Genotyping, and Association Analysis
First, 20-40 mg ear tissue was digested with 200 µL DNA lysis buffer (100 mmol Tris-Hcl, 25 mmol EDTA, 500 mmol NaCl, and 1% SDS) containing protease K (100 µg/mL) at 55 • C for 12 h. The impurities were then removed by sequential saturation with Tris saturated phenols, chloroform, and isoamyl alcohol. The DNA was diluted to 50 ng/µL as a template for PCR amplification. The SNVs were screened using pooled-DNA sequencing and genotyped by DNA sequencing. The primers are shown in Table S5. The general linear model (GLM) of the SAS v9.2 software (SAS Institute, Cary, NC, USA) was used to analyze the correlation between the SNVs in miR-23a and the sow fertility traits, as described in our previous report [12].

Statistical Analysis
The data are presented as the mean ± S.E.M, and each experiment was carried out at least three biological replicates. The statistical analysis and graphing were performed using the GraphPad Prism v5 software (GraphPad Software, San Diego, CA, USA), and the significance was tested using a two-tailed t-test or a one-way ANOVA with a least significant difference test and a multiple comparisons test.

miR-23a Is a Pro-Apoptotic miRNA in pGCs
The multi-sequence alignment showed that the porcine miR-23a gene shares high nucleotide identities with other mammals ( Figure S1). Furthermore, both the mature and seed sequences of miR-23a are completely conserved in mammals including pigs ( Figure 1A). To understand the potential function of miR-23a, a total of 111 putative targets were predicted using three bioinformatics tools-TargetScan, TarBase, and miRDB ( Figure 1B and Table S6)-which were mainly distributed in biological processes and pathways related to cancer, transcriptional regulation, and apoptosis ( Figure S2). Combined with previous studies in hGC apoptosis [18], we next focused on the role of miR-23a in pGC apoptosis. The flow cytometry showed that there was a greater induction of the apoptotic rate of pGCs when it was transfected with miR-23a mimics (an endogenous analog of miR-23a) compared to the transfection of mimics NC (nonsense scrambled) ( Figure 1C,D), whereas the opposite was observed when it was transfected with miR-23a inhibitor (a suppressor of miR-23a) compared to transfection of inhibitor NC (nonsense scrambled) ( Figure 1E,F). These results suggest that miR-23a is a pro-apoptotic miRNA in the GCs of the porcine ovary.
analyze the correlation between the SNVs in miR-23a and the sow fertility traits, as described in our previous report [12].

Statistical Analysis
The data are presented as the mean ± S.E.M, and each experiment was carried out at least three biological replicates. The statistical analysis and graphing were performed using the GraphPad Prism v5 software (GraphPad Software, San Diego, CA, USA), and the significance was tested using a two-tailed t-test or a one-way ANOVA with a least significant difference test and a multiple comparisons test.

miR-23a Is a Pro-Apoptotic miRNA in pGCs
The multi-sequence alignment showed that the porcine miR-23a gene shares high nucleotide identities with other mammals ( Figure S1). Furthermore, both the mature and seed sequences of miR-23a are completely conserved in mammals including pigs ( Figure  1A). To understand the potential function of miR-23a, a total of 111 putative targets were predicted using three bioinformatics tools-TargetScan, TarBase, and miRDB ( Figure 1B and Table S6)-which were mainly distributed in biological processes and pathways related to cancer, transcriptional regulation, and apoptosis ( Figure S2). Combined with previous studies in hGC apoptosis [18], we next focused on the role of miR-23a in pGC apoptosis. The flow cytometry showed that there was a greater induction of the apoptotic rate of pGCs when it was transfected with miR-23a mimics (an endogenous analog of miR-23a) compared to the transfection of mimics NC (nonsense scrambled) ( Figure 1C,D), whereas the opposite was observed when it was transfected with miR-23a inhibitor (a suppressor of miR-23a) compared to transfection of inhibitor NC (nonsense scrambled) ( Figure 1E,F). These results suggest that miR-23a is a pro-apoptotic miRNA in the GCs of the porcine ovary.  targets for miR-23a was performed using TarBase (blue), TargtScan (red), and miRDB (green). (C,D) After transfection with miR-23a mimics, miR-23a levels were examined by qRT-PCR at 24 h (C), and the apoptosis rate was examined by fluorescence-activated cell sorting (FACS) at 48 h (D) in pGCs. (E,F) After transfection with the miR-23a inhibitor, miR-23a levels were examined by qRT-PCR at 24 h (E), and the apoptosis rate was examined by FACS at 48 h (F) in pGCs. Data are expressed as the mean ± S.E.M., and the significance was tested using a t-test. (* p < 0.05, *** p < 0.001).

miR-23a Activates the Transcription and Function of lncRNA NORHA
Nuclear and cytoplasmic miRNAs have been shown to control cell functions by different mechanisms (RNAi or RNAa) [22]. We performed a subcellular localization assay by separating nuclear and cytoplasmic RNA and showed that miR-23a is more enriched in the nucleus than it is in the cytoplasm of pGCs (Figure 2A), indicating that miR-23a induces pGC apoptosis, possibly via an RNAa mechanism. Interestingly, an MRE motif was detected in the promoter region of the porcine NORHA, a pro-apoptotic lncRNA in pGCs ( Figure 2B), when we previously analyzed its promoter sequence [19]. Furthermore, the two have a strong potential interaction ability with the -20 kcal/mol of the minimum free energy (MFE) ( Figure 2C), so NORHA was selected for further investigation. As expected, we noted that the overexpression of miR-23a significantly activates NORHA transcription in pGCs at 48 h, but not at 24 h ( Figure 2D), which is in line with the first feature of the RNAa mechanism: delayed action time. Similarly, the knocking down of miR-23a significantly inactivates NORHA transcription in pGCs at 48 h, but not at 24 h ( Figure 2E). To investigate whether miR-23a induces pGC apoptosis through the activation of NORHA transcription, both the miR-23a inhibitor and pcDNA3.1-NORHA were co-transfected into pGCs cultured in vitro. The overexpression of NORHA significantly reversed the reduction of the pGC apoptosis rate caused by the miR-23a inhibitor ( Figure 2F). In summary, these data suggest that miR-23a enhances pGC apoptosis through the activation of NORHA transcription. The prediction of potential targets for miR-23a was performed using TarBase (blue), TargtScan (red), and miRDB (green). (C,D) After transfection with miR-23a mimics, miR-23a levels were examined by qRT-PCR at 24 h (c), and the apoptosis rate was examined by fluorescence-activated cell sorting (FACS) at 48 h (d) in pGCs. (E,F) After transfection with the miR-23a inhibitor, miR-23a levels were examined by qRT-PCR at 24 h (E), and the apoptosis rate was examined by FACS at 48 h (F) in pGCs. Data are expressed as the mean ± S.E.M., and the significance was tested using a t-test. (* p < 0.05, *** p < 0.001).

miR-23a Activates the Transcription and Function of lncRNA NORHA
Nuclear and cytoplasmic miRNAs have been shown to control cell functions by different mechanisms (RNAi or RNAa) [22]. We performed a subcellular localization assay by separating nuclear and cytoplasmic RNA and showed that miR-23a is more enriched in the nucleus than it is in the cytoplasm of pGCs (Figure 2A), indicating that miR-23a induces pGC apoptosis, possibly via an RNAa mechanism. Interestingly, an MRE motif was detected in the promoter region of the porcine NORHA, a pro-apoptotic lncRNA in pGCs ( Figure 2B), when we previously analyzed its promoter sequence [19]. Furthermore, the two have a strong potential interaction ability with the -20 kcal/mol of the minimum free energy (MFE) ( Figure 2C), so NORHA was selected for further investigation. As expected, we noted that the overexpression of miR-23a significantly activates NORHA transcription in pGCs at 48 h, but not at 24 h ( Figure 2D), which is in line with the first feature of the RNAa mechanism: delayed action time. Similarly, the knocking down of miR-23a significantly inactivates NORHA transcription in pGCs at 48 h, but not at 24 h ( Figure 2E). To investigate whether miR-23a induces pGC apoptosis through the activation of NORHA transcription, both the miR-23a inhibitor and pcDNA3.1-NORHA were co-transfected into pGCs cultured in vitro. The overexpression of NORHA significantly reversed the reduction of the pGC apoptosis rate caused by the miR-23a inhibitor ( Figure 2F). In summary, these data suggest that miR-23a enhances pGC apoptosis through the activation of NORHA transcription.  (F) After co-transfection with the miR-23a inhibitor and pcDNA3.1-NORHA, the pGC apoptosis rate was examined by FACS at 48 h. Data are expressed as the mean ± S.E.M., and significance was tested using a t-test (D,E) and a one-way ANOVA (F) (ns, not significant, * p < 0.05, ** p < 0.01).

miR-23a Is a saRNA of lncRNA NORHA
To test the miR-23a activation of lncRNA NORHA transcription via miR-23a MRE in its promoter, two reporter vectors of the porcine NORHA promoter were generated ( Figure 3A). The luciferase reporter assays showed that the luciferase activities of the NORHA promoter were significantly increased upon the overexpression of miR-23a in pGCs, while no significant changes were observed in the luciferase activities of the promoter with a mutation-type MRE motif of miR-23a ( Figure 3B,C), indicating that miR-23a enhances the promoter activity of NORHA via an MRE motif of miR-23a in its promoter. The activation of target transcription by saRNAs has been shown to depend on AGO2 [23]. We therefore performed a ChIP assay by a specific anti-AGO2 antibody to assess whether miR-23a guides AGO2 to bind to the NORHA promoter in pGCs. Compared with the X site (without the MRE motif of miR-23a) within the intron 1 of NORHA, a significant enrichment of AGO2 binding was observed in the MRE motif of miR-23a in the region of the NORHA promoter ( Figure 3D), indicating that AGO2 binding was guided by miR-23a. In addition, RNAa has also been shown to be involved in the histone modification at the target promoter region [22,24]. As expected, the overexpression of miR-23a resulted in the decreased enrichment of H3K9me2 and H3K9me3 at the MRE motif for miR-23a in the region of the NORHA promoter ( Figure 3E). Overall, these data suggest that miR-23a is a saRNA of lncRNA NORHA in pGCs.  (top) and ChIP images (bottom). F1/R1 and F2/R2 are the primers used to amplify the MRE site and X site, respectively. The X site is an intronic region of NORHA that does not contain the MRE of miR-23a. M, DNA marker. (E) ChIP assays were performed using H3K9me2-and H3K9me3-specific antibodies. Data are expressed as the mean ± S.E.M., and significance was tested using a t-test (ns, not significant, ** p < 0.01).

g.−398C > T Mutation in the miR-23a Promoter Decreases Its Transcriptional Activity
Interestingly, a C/T point mutation was detected at −398 nt of the promoter region of the porcine miR-23a gene, which was named g.−398C > T ( Figure 4A). To investigate whether the SNV g.−398C > T is located in the core promoter of the porcine miR-23a gene, we identified its core promoter by luciferase reporter assays using deletion constructs. As shown in Figure 4B, a DNA region from −658 to −289 nt is the core promoter of the porcine miR-23a gene, and the SNV g.−398C > T was located in this region. To further determine whether the SNV g.−398C > T affects miR-23a promoter activity, two reporter vectors of the core promoter with alleles C and T were constructed ( Figure 4C). The luciferase activities of the reporter vector with allele C were significantly higher than that with allele T in the pGCs ( Figure 4D). Furthermore, the expression levels of miR-23a in the pGCs of sows with genotype CC were significantly higher than those with genotype TT ( Figure 4E). In general, these data suggest that the SNV g.−398C > T is strongly involved in the transcriptional activity of miR-23a in pGCs through the influence of its promoter activity.

g.−398C > T Mutation in the miR-23a Promoter Creates a Novel Binding Site for the Transcription Factor SMAD4
To investigate the mechanism by which the SNV g.−398C > T decreases miR-23a promoter activity, we predicted the binding sites for the transcription factors in its core promoter. Notably, a putative binding site for the transcription factor NR2C1 was only detected in the miR-23a promoter with allele C for the SNV g.−398C > T, and a putative binding site for the transcription factor SMAD4 was only detected in the miR-23a promoter with allele T for the SNV g.−398C > T ( Figure 4F). The luciferase reporter assays showed that the transcription factor SMAD4 significantly reduces the activities of the miR-23a promoter with allele T, but not that with allele C (Figures 4G and S3A). However, the transcription factor NR2C1 has not affected the activities of the miR-23a promoter with both allele C and allele T (Figures 4G and S3B). The ChIP experiments showed that the transcription factor SMAD4 directly binds to the miR-23a promoter in the GCs from sows with genotype TT for the SNV g.−398C > T, but not those with genotype CC ( Figure 4H). These results suggest that the SNV g.−398C > T produces a binding site for the transcription factor SMAD4, which decreases the transcriptional activity of miR-23a through recruiting the transcription factor SMAD4.

The SNV g.−398C > T Affects the SMAD4 Inhibition of miR-23a Expression
To investigate whether SMAD4 regulates miR-23a expression, we isolated the GCs from sows with different genotypes for the SNV g.−398C > T and treated them with pcDNA3.1-SMAD4. The overexpression of SMAD4 significantly inhibited the expression of miR-23a in the pGCs with genotype TT, but not those with genotype CC (Figure 5A), indicating that SMAD4 is a transcription repressor of miR-23a in the pGCs with genotype TT for the SNV g.−398C > T. Furthermore, a co-transfection experiment with pcDNA3.1-SMAD4 and miR-23a mimics showed that SMAD4 also significantly reduces the expression of NORHA, a direct target of miR-23a, while miR-23a can rescue this situation in pGCs ( Figure 5B). Importantly, SMAD4 is an anti-apoptotic transcription factor in pGCs [20]. We therefore investigated whether miR-23a mediates the anti-apoptotic effect of SMAD4. Flow cytometry confirmed that SMAD4 significantly reduces the apoptosis rate of pGCs, while the overexpression of miR-23a can reverse the anti-apoptotic influence of SMAD4 in pGCs ( Figure 5C,D). The above data suggest that SMAD4 is embroiled in miR-23a functions, including suppressing the downstream target NORHA expression, as well as the pGC apoptosis, via the regulation of miR-23a.  [20]. We therefore investigated whether miR-23a mediates the anti-apoptotic effect of SMAD4. Flow cytometry confirmed that SMAD4 significantly reduces the apoptosis rate of pGCs, while the overexpression of miR-23a can reverse the anti-apoptotic influence of SMAD4 in pGCs ( Figure 5C,D). The above data suggest that SMAD4 is embroiled in miR-23a functions, including suppressing the downstream target NORHA expression, as well as the pGC apoptosis, via the regulation of miR-23a.

The miR-23a SNV g.−398C>T Affects the Resistance of GCs to Oxidative Stress
Our previous study demonstrated that FoxO1, an effector for oxidative stress, is downstream of NORHA in pGCs [19]. Western blotting showed that the overexpression of miR-23a significantly induces the FoxO1 protein level in pGCs ( Figure 6A). Notably, the mRNA expression of NORHA and FoxO1 in pGCs with genotype TT for the SNV g.−398C>T were significantly lower than that in pGCs with genotype CC (Figure 6B-C). To investigate whether the miR-23a SNV g.−398C>T influences the pGCs response to oxidative stress, we induced oxidative stress in pGCs by H2O2 exposure. Flow cytometry showed that pGCs with genotype TT for the SNV g.−398C>T have a significantly lower apoptosis rate than those with genotype CC ( Figure 6D). This is consistent with the low expression of the pro-apoptotic factor miR-23a, NORHA, and FoxO1 in pGCs with genotype TT for the SNV g.−398C>T. Interestingly, under H2O2 exposure, the increase in the apoptosis rate of pGCs with genotype TT for the SNV g.−398C>T was also lower than that of genotype CC (Figure 6D), indicating that pGCs with genotype TT for the SNV g.−398C>T are protected against oxidative stress relative to those with genotype CC. These data suggest that the miR-23a SNV g.−398C>T affects the response of pGCs to oxidative stress.

The miR-23a SNV g.−398C > T Affects the Resistance of GCs to Oxidative Stress
Our previous study demonstrated that FoxO1, an effector for oxidative stress, is downstream of NORHA in pGCs [19]. Western blotting showed that the overexpression of miR-23a significantly induces the FoxO1 protein level in pGCs ( Figure 6A). Notably, the mRNA expression of NORHA and FoxO1 in pGCs with genotype TT for the SNV g.−398C > T were significantly lower than that in pGCs with genotype CC (Figure 6B-C). To investigate whether the miR-23a SNV g.−398C > T influences the pGCs response to oxidative stress, we induced oxidative stress in pGCs by H 2 O 2 exposure. Flow cytometry showed that pGCs with genotype TT for the SNV g.−398C > T have a significantly lower apoptosis rate than those with genotype CC ( Figure 6D). This is consistent with the low expression of the pro-apoptotic factor miR-23a, NORHA, and FoxO1 in pGCs with genotype TT for the SNV g.−398C > T. Interestingly, under H 2 O 2 exposure, the increase in the apoptosis rate of pGCs with genotype TT for the SNV g.−398C > T was also lower than that of genotype CC (Figure 6D), indicating that pGCs with genotype TT for the SNV g.−398C > T are protected against oxidative stress relative to those with genotype CC. These data suggest that the miR-23a SNV g.−398C > T affects the response of pGCs to oxidative stress.

The miR-23a SNV g.−398C>T is Significantly Associated with Sow Fertility Traits
To investigate whether the SNV g.−398C>T in the miR-23a promoter is associated with sow fertility traits, the SNV g.−398C>T was genotyped in a Yorkshire sow population (n = 346) by direct sequencing. Three genotypes (CC, CT, and TT) with frequencies of 0.41, 0.47, and 0.12, respectively, were observed ( Figure 7A,B). Moreover, the allele frequencies were 0.65 (allele C) and 0.375 (allele T) ( Figure 7C), indicating that allele C is the dominant allele. To determine the effect of the SNV g.−398C>T on the fertility traits in Yorkshire sows, an association analysis was carried out using a mixed model. For multiparity, the total number of piglets born (TNB) of sows with genotype TT was significantly higher than that of those with genotype CC ( Figure 7D). However, there is no significant difference in the other fertility traits, including the total number of piglets born alive (NBA), the number of stillbirths (NSB), and the litter weight (LW) among three genotypes (CC, CT, and TT) of the SNV g.−398C>T in this Yorkshire population ( Figure S4). In summary, these results indicate that the miR-23a SNV g.−398C>T is significantly associated with sow fertility traits, and miR-23a is a candidate miRNA for sow fertility traits in the Yorkshire population.

The miR-23a SNV g.−398C > T is Significantly Associated with Sow Fertility Traits
To investigate whether the SNV g.−398C > T in the miR-23a promoter is associated with sow fertility traits, the SNV g.−398C > T was genotyped in a Yorkshire sow population (n = 346) by direct sequencing. Three genotypes (CC, CT, and TT) with frequencies of 0.41, 0.47, and 0.12, respectively, were observed ( Figure 7A,B). Moreover, the allele frequencies were 0.65 (allele C) and 0.375 (allele T) ( Figure 7C), indicating that allele C is the dominant allele. To determine the effect of the SNV g.−398C > T on the fertility traits in Yorkshire sows, an association analysis was carried out using a mixed model. For multiparity, the total number of piglets born (TNB) of sows with genotype TT was significantly higher than that of those with genotype CC ( Figure 7D). However, there is no significant difference in the other fertility traits, including the total number of piglets born alive (NBA), the number of stillbirths (NSB), and the litter weight (LW) among three genotypes (CC, CT, and TT) of the SNV g.−398C > T in this Yorkshire population ( Figure S4). In summary, these results indicate that the miR-23a SNV g.−398C > T is significantly associated with sow fertility traits, and miR-23a is a candidate miRNA for sow fertility traits in the Yorkshire population. (D) Association analysis between the SNV g.−398C>T and TNB of sows for multiparities. Data are expressed as the least-squares mean ± S.E.M., n = 346, and significance was tested using a GLM program. Significant differences between groups (p < 0.05) are indicated by different lowercase letters, and insignificant differences (p > 0.05) are indicated by the same letter.

Discussion
In porcine folliculogenesis, more than 95% of follicles will undergo atresia, which severely limits sow fertility. Endogenous miRNAs are strongly involved in porcine follicular atresia and control key characteristics during follicular atresia, such as miR-144 [25] and miR-2337 [22] control pGC apoptosis, the main causes of follicular atresia. Besides, the balance of progesterone (P4)/17β-estradiol (E2) concentration in follicular fluid, a core classification index of atretic follicles, is also modulated by miRNAs including miR-1275 [26] and miR-31-5p [27]. In this study, we identified miR-23a as a novel miRNA that induces pGC apoptosis in pigs. Previous studies showed that miR-23a promotes cell apoptosis in hGCs via the FasL-Fas, ERK1/2, and CDC42/PAK1 signaling pathways, which are related to ovarian reserve function and premature ovarian failure [18,28,29]. In hGCs, miR-23a has also been shown to block the cell cycle on the G0/G1 phase and suppress cell proliferation [28]. In addition to GCs, miR-23a is also an important regulator in many cell

Discussion
In porcine folliculogenesis, more than 95% of follicles will undergo atresia, which severely limits sow fertility. Endogenous miRNAs are strongly involved in porcine follicular atresia and control key characteristics during follicular atresia, such as miR-144 [25] and miR-2337 [22] control pGC apoptosis, the main causes of follicular atresia. Besides, the balance of progesterone (P4)/17β-estradiol (E2) concentration in follicular fluid, a core classification index of atretic follicles, is also modulated by miRNAs including miR-1275 [26] and miR-31-5p [27]. In this study, we identified miR-23a as a novel miRNA that induces pGC apoptosis in pigs. Previous studies showed that miR-23a promotes cell apoptosis in hGCs via the FasL-Fas, ERK1/2, and CDC42/PAK1 signaling pathways, which are related to ovarian reserve function and premature ovarian failure [18,28,29]. In hGCs, miR-23a has also been shown to block the cell cycle on the G0/G1 phase and suppress cell proliferation [28]. In addition to GCs, miR-23a is also an important regulator in many cell types such as human acute lymphoblastic T-cells [30], airway epithelial cells and fibroblasts [31], and oral squamous cell carcinoma cells [32]. In summary, our findings not only extend the function of miR-23a-regulating GC apoptosis from humans to domestic animals, they once again confirm the conservation of the function of miRNAs in the same cell types among different species and also provide a new potential non-hormonal drug for sow reproductive regulation.
lncRNAs and miRNAs are the two most studied endogenous non-coding RNAs (ncRNAs), which form a complex regulatory network with protein-coding genes through various forms of dialogue mechanisms to control health and disease. LncRNAs, for instance, (i) act as a molecular sponge of miRNAs [19], (ii) compete for protein-coding genes with miRNAs [33], (iii) act as a precursor or host of miRNAs [34], and (iv) contribute to miRNA biogenesis [35]. Meanwhile, miRNAs also inhibit lncRNA expression via an RNAi mechanism, just as they inhibit the 3 -UTRs of protein-encoding genes [36]. In this study, we identified a novel mechanism by which miRNAs regulate lncRNA-miR-23a acts as an endogenous small saRNA to promote the transcription of lncRNA NORHA through direct binding to the promoter of lncRNA NORHA. saRNAs are a kind of small double-stranded RNA (dsRNA) with a length of 19-21 nt which target the promoter regions of genes and induce gene transcription via an RNAa mechanism [6]. In the mid 2000s, Li et al. [24] found that chemically synthesized exogenous saRNA activates the transcription of target genes in mammalian cells. Subsequently, some exogenous saRNAs such as the CEBPA-saRNA targets CEBPA [37] and the saRNA-158 targets VMP1 [38] have been identified. Since the mature sequence of miRNAs is similar to that of saRNAs in terms of length, scientists attempt to assess their feasibility as endogenous saRNAs [39]. In recent years, several miRNAs including miR-4281 [40], miR-320 [41], and miR-2337 [22] have been identified as the endogenous saRNAs target important functional genes. In conclusion, our findings prove for the first time that miRNA can serve as the saRNA of lncRNA, which establishes a new form of connection between miRNAs and lncRNAs. At the same time, our study also provides a theoretical basis for using miRNAs as endogenous saRNAs in order to enhance the transcriptional activation of favorable genes for female reproduction and improve female fertility.
Variants in the promoter region are strongly involved in gene transcription [42]. One of the important mechanisms is to affect the binding with transcription factors by creating a new transcription factor binding site or by losing an existing transcription factor binding site [43]. In this study, we identified, for the first time, the SNV g.−398C > T in the miR-23a promoter and showed that a new binding site for the transcription factor SMAD4 in the miR-23a promoter was created by the SNV g.−398C > T, which significantly inhibits the transcriptional activity and expression of miR-23a in pGCs with genotype TT, thereby reducing pGC apoptosis and increasing sow fecundity. As an important effector of the TGF-β signaling pathway and an anti-apoptotic factor in mammalian GCs, SMAD4 is well known to be involved in female reproduction, such as the fate (maturation or atresia) of follicular development [20], ovulation [44], and fertility [45]. Furthermore, SMAD4 generally functions as a transcription factor, directly acting on the key genes including the miRNAs of important events in the female reproductive cycle [46]. For example, the key genes HAS2, for inducing ovulation [47], and ALK3, for maintaining embryo implantation [48], are all transcriptional targets of SMAD4. Recently, several miRNAs involved in critical events in the female reproductive cycle, such as miR-29c, for influencing the follicular development fate [12], have also been shown to be directly regulated by the transcription factor SMAD4. Taken together, our findings not only identify an miRNA affecting GC apoptosis but also reveal a molecular mechanism by which the miR-23a SNV g.−398C > T affects miR-23a transcription and GC apoptosis. Meanwhile, our findings also define a new signaling pathway, SMAD4/miR-23a/NORHA, affecting GC apoptosis, which provides new insights into the mechanism of GC apoptosis and female fertility in mammals.
Oxidative stress is a major trigger of follicular atresia and GC apoptosis which impairs follicular development, leads to a variety of ovarian diseases, and limits the fertility of female animals including pigs [49][50][51]. FoxO1 is a major sensor and effector of oxidative stress [52] which can promote cell apoptosis in response to oxidative stress by activating downstream apoptosis-related proteins such as Bim [53], P21 [54], and Fasl [55]. Similarly, it has also been reported that inhibiting FoxO1 by different means can effectively block the effect of oxidative stress [56]. Here, we found that the level of FoxO1, downstream of NORHA (a direct target of miR-23a), was decreased in pGCs with genotype TT for the miR-23a SNV g.−398C > T, indicating that the SNV g.−398C > T may affect the response of pGCs to oxidative stress. Subsequently, this hypothesis was confirmed by an in vitro oxidative stress model for pGCs. Multiple environmental stresses in swine production often cause the elevation of oxidative free radicals between the organism and tissues, so oxidative stress is a common mechanism by which multiple stressors impair reproductive performance [57]. At present, there are many studies on the mechanisms by which oxidative stress affects reproductive performance in sows, but there are few studies on the differentiation of sows in response to oxidative stress and the underlying mechanisms [58]. In addition, the growing practice suggests that differences between the rearing environment of breeding pigs (friendly environment) and that of commercial pigs (a more challenging environment) result in some individuals often failing to exhibit their due genetic potential, and, therefore, genotype-by-environment interactions (GEI) are increasingly being considered in pig breeding programs in order to improve the environmental adaptation, productivity, and economic benefits for the pig industry [59][60][61]. Therefore, GEI is becoming an important focus for the pig industry. In conclusion, our findings reveal that a mechanism underlying mutation affects sow fertility by exerting different responses to the external environment, which provide a potential molecular marker for selecting sows with high environmental adaptation and fertility. Meanwhile, our findings also shed new light on the molecular mechanism of GEI in regulating economically important traits (i.e., reproductive and growth traits) in livestock and the personalized regulation of sow reproduction. miRNAs are required for female reproduction and are involved in all the important events in the mammalian reproductive cycle from the establishment of the primordial follicle pool to pregnancy and parturition [62]. The knockout of key genes such as Dicer and Drosha/DGCR8 in the miRNA biogenesis pathway can lead to changes in the female fertility of mice [63]. Indeed, a significant association between miR-604A>G and miR-938G>A polymorphism and idiopathic recurrent pregnancy loss were observed in Korean women [64]. In sows, miRNAs can control the fate of follicular development, and their levels are involved in sow fertility [65,66]. However, no candidate miRNAs for sow fertility traits have been identified. Although two SNVs (g.−573G>A and g.−40T>C) were identified in the miR-130a promoter region by our group, they were not associated with the fertility traits in a Yorkshire sow population [67]. In addition, a T > C variant has been revealed to be associated with the TNB and NBA traits in a DIV sow population; however, it is located at 18 nt downstream of the pre-miR-27a [65]. In this study, we showed that the SNV g.−398C > T in the miR-23a promoter is associated with the TNB trait, demonstrating that miR-23a is a candidate miRNA for the fertility traits of Yorkshire sows. Together, our findings identified, for the first time, a causal miRNA for sow fertility traits, which provides a potential genetic marker for molecular breeding in Yorkshire sows and confirms that miRNA can serve as a new carrier for the genetic markers of sow fertility traits. However, this candidate genetic marker remains to be verified in pig breeding.

Conclusions
In conclusion, we demonstrated that miR-23a acts as a saRNA to activate lncRNA NORHA transcription in pGCs through the alteration of histone modifications in the promoter region by directly binding to its promoter in order to induce GC apoptosis. Interestingly, we identified the SNV g.−398C > T in the miR-23a promoter, which is strongly involved in miR-23a transcription and functions by recruiting the transcription repressor SMAD4. Importantly, this SNV also causes a decrease in the sensitivity of GCs to oxidative stress, thereby increasing sow fertility (Figure 8). Our findings provide an endogenous saRNA for regulating sow reproduction and potential molecular markers for breeding sows with strong environmental adaptability and high fecundity. stress, thereby increasing sow fertility (Figure 8). Our findings provide an endogenous saRNA for regulating sow reproduction and potential molecular markers for breeding sows with strong environmental adaptability and high fecundity.

Institutional Review Board Statement:
This study was authorized by the Animal Ethics Committee (AEC) of Nanjing Agricultural University, Nanjing, China.
Informed Consent Statement: Not applicable.

Data Availability Statement:
No new data were created or analyzed in this study. Data sharing is not applicable to this article.

Conflicts of Interest:
The authors declare that they have no conflict of interest.