Aqp8 ^−/− mice (C57BL/6 genetic background) were generated by targeted gene disruption (Yang et al., 2005). For this experiment, 5–6-week-old female mice were used. Mice were allowed free access to water and food in 12 h light and 12 h dark conditions. All animal experiments were reviewed and approved by the Committee on the Ethics of Animal Research of Dalian Medical University.

Aqp8 ^+/+ and Aqp8 ^−/− were intraperitoneally injected with 0.1 ml pregnant mare serum gonadotropin (PMSG) (1000 IU in 10 ml 0.9% NaCl, Ningbo a second hormone factory) or vehicle control (0.9% NaCl, 0.1 ml) at 16:00 into the mice. Mice were killed by cervical dislocation 0, 1, 2, 3, 4, and 5 days after PMSG treatment, and the ovary samples were excised. The ovaries were used for the collection of granulosa cells.

Ovaries were excised from mice and placed in DMEM/F12 (GIBCO-BRL, 11039021) that was supplemented with 10% fetal bovine serum (GIBCO-BRL), 10 mg/ml of streptomycin sulfate (Sigma), and 75 mg/ml of penicillin G (Sigma). Granulosa cells were harvested by follicle puncture using a 25-gauge needle. After follicle puncture, granulosa cells were suspended in the appropriate solution for immunoblotting or H₂O₂ concentration.

For in vitro culture of granulosa cells under serum-free conditions, ovaries were collected, and granulosa cells were collected by follicle puncture, as described previously. The cells were seeded in 24-well plates and were allowed to attach overnight. The next morning, the medium and unattached cells were removed and replaced with serum-free media. After 24 h, the granulosa cells were fixed for immunofluorescence.

The adenovirus expressing GFP-LC3B (Ad-GFP-LC3B, C3006, Beyotime) was transfected into GCs grown on coverslips. After 24 h, cells were rinsed using PBS and then exposed to 2 h of H₂O₂ incubation. The distribution and fluorescence emitted by GFP-LC3B puncta were then observed under a live cell station (GE). Experiments were repeated three times.

Protein was isolated by the previous description (Huang et al., 2020a). The freshly isolated granulosa cells were lysed with ice-cold radioimmunoprecipitation assay (RIPA) buffer that was supplemented with a protease inhibitor PMSF (Beyotime). To facilitate the complete solubilization of the cellular proteins, the cell lysates were incubated on ice for 30 min and then centrifuged (13000 g at 4°C for 30 min). The protein concentration was tested by using a BCA Protein Assay Kit. The whole-cell lysates (20 mg/lane) were separated by sodium dodecyl sulfate (SDS)–polyacrylamide gel electrophoresis and transferred to a polyvinylidene difluoride (PVFD) membrane (Millipore). After the nonspecific binding sites were blocked with 5% skim milk, the membrane was treated with the anti-LC3 rabbit polyclonal antibody (1:2000, Novus Biologicals), cleaved caspase-3 rabbit polyclonal antibody (diluted 1:1000, Cell Signaling Technology), and β-actin (1:1000, Beyotime) overnight at 4°C. The primary antibodies for Akt, pAkt-Ser473, Bax, Bcl-2, and P-Tyr were obtained from Cell Signaling Technology (Beverly, MA). The primary antibodies for Atg3 were obtained from Proteintech (Wuhan, China), while Beclin-1 was obtained from Santa Cruz Biotechnology (Santa Cruz, CA, United States). The immunoreactive bands were demonstrated by incubation with horseradish peroxidase (HRP)-conjugated goat anti-rabbit IgG (1:10000, Zhongshan Biotechnology) at room temperature for 1 h. The peroxidase activity was visualized with the enhanced chemiluminescence detection system (WBKLS500, Millipore). Integrated optical intensities of the immunoreactive protein bands were detected by using the DNR bioimaging system MicroChemi 4.2 and the quantified analysis by ImageJ software, normalized to β-actin values.

To identify autophagic vacuoles at the ultrastructural level, the ovary was fixed with 2.5% glutaraldehyde in 0.1 M cacodylate buffer (pH 7.4) for more than 24 h at 4°C, rinsed in cacodylate buffer, postfixed in 1% OsO₄ in cacodylate buffer, dehydrated, and embedded in Epon. Ultrathin sections were briefly contrasted with uranyl acetate and photographed with a transmission electron microscope (Hitachi 7100 Japan).

Intracellular hydrogen peroxide was measured by using the Hydrogen Peroxide Assay Kit (Beyotime, Nanjing, China). Briefly, 100–200 μl of hydrogen peroxide was added to detect the ratio of lysate to lysate. One million granulosa cells were treated with 100–200 μl of the hydrogen peroxide lysate, followed by sufficient homogenization to break and lyse the cells. It was then centrifuged at 12000 g at 4°C for 3–5 min, and the supernatant was collected. Then, 50 µl of the sample or standard was treated with 100 µl of the hydrogen peroxide detection reagent at room temperature (15–30°C). And then, it was immediately measured at A560 nm.

To evaluate intracellular H₂O₂ levels, 1×10⁶ cells/ml were washed twice in HBSS and incubated with 20 µM 2′,7′-dichlorofluorescein diacetate (DCFH-DA) (sigma). Granulosa cells were incubated for 30 min at 37°C. DCFH-DA was a small nonpolar and nonfluorescent molecule that diffuses into the cells, where it is enzymatically deacetylated by intracellular esterases to a polar nonfluorescent compound, which is oxidized to the highly green fluorescent 2,7-dichlorofluorescein (DCF). DCF fluorescence was measured using a multiwell plate reader (Synergy Neo HTS multimode microplate reader, BioTek) at excitation and emission wavelengths of 485 and 535 nm, respectively.

Total RNA was isolated by the previous description (Huang et al., 2020b) using the TRIzol reagent (10296010, Invitrogen) and quantified using a NanoVueTM Plus Spectrophotometer (GE Healthcare, Buckinghamshire, United Kingdom). The A260/A280 ratio of the optical density was measured using the NanoVueTM Plus Spectrophotometer. The ratio was between 1.9 and 2.1 for all samples, indicating a good quality of RNA purity and yield. cDNA was synthesized from 2 µg of total RNA using the RT reagent Kit with gDNA Eraser (TaKaRa, RR047A). The qPCR solution contained 1 µl of cDNA, 1 µl of specific primers, and 10 µl of a 2X PCR SuperMix (AS111, TransGen Biotech) in a final volume of 20 µl. All primers were produced by lifespan (Milan, Germany) and listed as follows in the 5′-3′ direction: Aqp8-F: ACA​CCA​ATG​TGT​AGT​ATG​GAC​CT; Aqp8-R: TGA​CCG​ATA​GAC​ATC​CGA​TGA​AG; β-actin-F: TGG​AAT​CCT​GTG​GCA​TCC​ATG​AAA​C; β-actin-R: TAA​AAC​GCA​GCT​CAG​TAA​CAG​TCC​G.

The reaction conditions were as follows: polymerase activation and DNA denaturation (one cycle at 95°C for 30 s); denaturation, annealing, and extension (40 cycles at 95°C for 10 s and 60°C for 30 s); melting curve (65°C, with the temperature, gradually increased 0.5°C up to 95°C). mRNA expression was normalized to the level of β-actin mRNA. Changes in mRNA expression were calculated according to the 2^−△△Ct method. The amplification of β-actin mRNA was utilized to normalize the data.

Raw transcriptome sequencing data were stored in FASTQ document format by Bcl2fastq (v2.17.1.14). Sequencing data quality was assessed by FastQC (v0.10.1). The raw data were preprocessed, the low-quality data were filtered, and the contamination and joint sequences were removed by cutadapt (version 1.9.1). Then, gene expressions were calculated by HTSeq (V 0.6.1) with FPKM (Fragments per Kilobase per Million reads) (Mortazavi et al., 2008). Differentially expressed genes (DEGs) in the ovary (group: control and PMSG with treatment for 2 days or 4 days) were analyzed by using the R package edgeR (v3.4.6) using |fold change|>2 and a p-value < 0.05.

Statistical analysis was performed by analysis of variance (ANOVA). Significant differences between treatment groups were determined by Duncan’s multiple range tests. p < 0.05 was considered statistically significant. For statistical comparisons between two groups, an independent-sample t-test was used.

Microtubule-associated light-chain protein 3 (LC3) is an autophagy marker used to evaluate granulosa cell autophagy. Granulosa cells freshly isolated from Aqp8 ^+/+ and Aqp8 ^−/− mice were used. The results of Western blot analysis using specific antibodies against the autophagy marker LC3 and apoptosis marker cleaved caspase-3 were shown in Figure 1. Densitometric analysis normalized by the β-actin content indicated that the protein contents of LC3 and cleaved caspase-3 significantly decreased compared with the control, suggesting that AQP8 was involved in the development of follicular granulosa cells via autophagy and apoptosis.

Transmission electron micrographs showed that autophagosomes existed in the follicular granulosa cells of Aqp8 ^+/+ and Aqp8 ^−/− mice. Fewer autophagosomes localized in the granulosa cells of the Aqp8 ^−/− mice than in those of the Aqp8 ^+/+ mice (Figure 2).

Considering that AQP8 is expressed on granulosa cells (Su et al., 2010), we used Aqp8 ^−/− mice to determine whether or not AQP8 is required for the efficient entry of exogenous H₂O₂. To investigate H₂O₂ transport across the plasma membrane, we incubated DFCH-DA into granulosa cells. The addition of PBS or 50 µM H₂O₂ clearly activated DFCH-DA to DCF, as determined by 488/510 nm shifts on a microplate reader (Figure 3A). The aforementioned experiments provided additional evidence that H₂O₂ cannot freely permeate through the plasma membrane and identified AQP8 as an efficient transporter. Clearly, H₂O₂ import was severely impaired with AQP8 silencing.

Meanwhile, we investigated whether or not AQP8 can transport extracellular H₂O₂ into granulosa cells. Intracellular H₂O₂ was measured in primary granulosa cell cultures from Aqp8 ^+/+ and Aqp8 ^−/− mice after the extracellular addition of 10–100 μM H₂O₂ using the fluorescent dye CM-H2DCFDA, which reacted with ROS, including H₂O₂. Results showed that the cellular H₂O₂ level was significantly higher in the granulosa cells of the Aqp8 ^+/+ mice than that in those of the Aqp8^−/− mice (Figure 3B), indicating the involvement of AQP8 in H₂O₂ transport in granulosa cells. The ability of AQP8 to transport H₂O₂ across the plasma membrane was tested. Figure 3B showed that the basal intracellular H₂O₂ level was significantly affected by AQP8, revealing that the channel activity of this AQP isoform modulated the entry of physiologically produced H₂O₂ into the cells.

Autophagy, as an important process of programmed cell death, regulates follicular homeostasis in rats (Choi et al., 2010). Follicular development and atresia in vivo model was established in immature mice. Granulosa cell autophagy was determined by measuring the expression levels of LC3-II/LC3-I. As shown in Figure 4A, LC3-II expression in the granulosa cells was significantly downregulated 1 and 2 days after PMSG injection compared with the granulosa cells of immature mice without the treatment of exogenous gonadotropin (day 0). LC3-II expression was upregulated on day 3 and maintained until day 5. Furthermore, the expression of Aqp8 was measured. After PMSG injection, the expression of Aqp8 was significantly downregulated on days 1 and 2 and then increased on days 3, 4, and 5 (Figure 4B). Ovarian granulosa cells were isolated from each stage of the atresia model, and the concentration of H₂O₂ in granulosa cells was measured. The concentration of H₂O₂ in granulosa cells decreased during follicular development but increased during atresia (Figure 4C), suggesting that AQP8 mediated H₂O₂ transport and autophagy in granulosa cells to regulate follicular development and atresia.

The mechanism by which AQP8 regulates the autophagy of granulosa cells was further explored. Results showed that AQP8 can facilitate the absorption of H₂O₂ into granulosa cells and mediate downstream intracellular signaling. As shown in Figure 5A, the phosphorylation level of Akt in the granular cells of the Aqp8 ^−/− mice was increased. The expression levels of proteins related to apoptosis (Bax and caspase-3) and autophagy (Beclin-1 and Atg-3) decreased in the granulosa cells of the Aqp8 ^−/− mice. Meanwhile, phosphorylated tyrosine signaling was activated in the granulosa cells of the Aqp8 ^−/− mice (Figure 5B). And the absorption of H₂O₂ in the granulosa cells of the Aqp8 ^+/+ mice inhibited the PI3K signaling pathway and promoted granulosa cell death and autophagy.

We next investigated whether or not AQP8 influences autophagy by oxidative stress. An in vitro culture experiment revealed that AQP8 was involved in granulosa cell autophagy through the transport of H₂O₂. Primary cultured GCs were transfected with the GFP-LC3B plasmid, incubated with or without 200 μM H₂O₂ for 2 h, and then rinsed with PBS. Immunofluorescence images of granulosa cells were obtained in a live cell station. As shown in Figure 6, AQP8 obviously induced autophagosome formation in the cells with H₂O₂ exposure, displaying that AQP8 facilitated intracellular autophagy of granulosa cells through H₂O₂ absorption.

Transcriptome analysis of the ovary was performed in mice with PMSG treatment for 2 days or 4 days. Principal components, differentially expressed genes (DEGs), and KEGG pathways were analyzed. The principal components of each group between parents and offspring showed obvious differences (Figure 7A). To further study the mechanisms underlying follicular development and atresia, we analyzed the DEGs and KEGG pathways between different groups. In total, 1,421 DEGs with |fold change|> 2 and p-value < 0.05 were annotated in the ovary (groups: control and injected with PMSG for 2 days), of which 690 genes were upregulated and 731 genes were downregulated (Figure 7B). In total, 1,237 DEGs with |fold change| > 2 and p value < 0.05 were annotated in the ovary (groups: control and injected with PMSG for 4 days), of which 622 genes were upregulated and 615 genes were downregulated (Figure 7C). In addition, DEG analyses of the three cohorts (control, injected with PMSG for 2 days or 4 days) overlapped, with 52 common DEGs obtained (Figure 7D). KEGG pathway enrichment analysis was also performed to study the pathway with significant DEG enrichment. Functional enrichment results revealed that 30 KEGG pathways of DEGs in the ovaries of the mice injected with PMSG for 2 days or 4 days were statistically significantly enriched, respectively (Figures 7E,G). The KEGG pathways in the control mice and mice injected with PMSG for 2 days included five categories: organismal systems, metabolism, human diseases, environmental information processing, and cellular processes in the ovary (Figure 7F). The KEGG pathways in the control mice and mice injected with PMSG for 4 days included four categories: organismal systems, metabolism, human diseases, and environmental information processing in the ovary (Figure 7H). Organismal systems and metabolism are the most enriched pathways, suggesting that they might play important roles in follicular development and atresia.