COX15 deficiency causes oocyte ferroptosis

Significance

1) We identified the mitochondria-associated mutant gene COX15 in patients with oocyte defects in a recessive inheritance pattern. This finding expands the spectrum of mitochondrial disorders to oocyte defects and female infertility. 2) All patients with COX15 pathogenic variants show clinical manifestation of oocyte defects, while the somatic tissues are normal, suggesting the different role of mitochondria between somatic cells and germ cells. 3) We observed an unexpected pathological mechanism in which COX15 deficiency results in oocyte defects by triggering oocyte ferroptosis. 4) Inhibition of ferroptosis rescued the oocyte ferroptosis phenotype in vitro and ex vivo triggered by COX15 dysfunction, thus providing a potential therapeutic strategy in the future.

Abstract

Mitochondria play diverse roles in mammalian physiology. The architecture, activity, and physiological functions of mitochondria in oocytes are largely different from those in somatic cells, but the mitochondrial proteins related to oocyte quality and reproductive longevity remain largely unknown. Here, using whole-exome sequencing data from 1,024 women (characterized by oocyte maturation arrest and degenerated or morphologically abnormal oocytes) and 2,868 healthy controls, we performed a population and gene–based burden test for mitochondrial genes and identified a candidate gene, cytochrome c oxidase assembly protein 15 (COX15). We report that biallelic COX15 pathogenic variants cause human oocyte ferroptosis and female infertility in a recessive inheritance pattern. COX15 variants impaired mitochondrial respiration in Saccharomyces cerevisiae and led to reduced protein levels in HeLa cells. Oocyte-specific deletion of Cox15 led to impaired Fe²⁺ and reactive oxygen species homeostasis that caused mitochondrial dysfunction and ultimately sensitized oocytes to ferroptosis. In addition, ferrostatin-1 (an inhibitor of ferroptosis) could rescue the oocyte ferroptosis phenotype in vitro and ex vivo. Our findings not only provide a genetic diagnostic marker for oocyte development defects but also expand the spectrum of mitochondrial disorders to female infertility and contribute to unique insights into the role of ferroptosis in human oocyte defects.

Mitochondrial disorders are among the most common types of inherited diseases. They are usually caused by genetic variants (either in the mitochondrial or nuclear genome) and affect the energy-generating processes of oxidative phosphorylation (1–3). Numerous genes have been reported to cause mitochondrial diseases, the majority of which are involved in the assembly of respiratory complexes (Complex I-V) (2, 4, 5). However, the clinical manifestations of most reported mitochondrial diseases are seen in somatic tissues, especially those with high energy demand such as the brain, heart, and muscles (2). It has been reported that mitochondria are a key indicator of oocyte quality, and mitochondrial dysfunction can affect the development and quality of female germ cells, thus leading to female infertility (6–8).

Compared with mitochondria in somatic cells, oocyte mitochondria have several unique characteristics such as spherical morphology, fewer cristae, fragmented mitochondrial networks, and low respiratory activity (9, 10). In somatic cells, the normal activity of the electron transport chain, including complex I, is essential for energy metabolism and for the biosynthesis of essential biomolecules, which in turn generates reactive oxygen species (ROS) (11). In contrast, in early-stage oocytes, the activity of respiratory complex I is suppressed to prevent the generation of toxic levels of ROS (12). These differences indicate the distinct role of mitochondria in oocytes, but currently known mitochondrial genetic variants show severe effects on somatic organs, and thus, the fertility status of affected individuals has been unavailable due to developmental delays and death prior to sexual maturity (5, 13). In addition, the high phenotypic heterogeneity of mitochondrial disorders resulting from different variants makes it difficult to establish clear genotype–phenotype associations between pathogenic variants and infertility (5, 14). Thus, until now, there have been no studies providing convincing evidence that pathogenic variants in genes encoding mitochondrial proteins cause human oocyte defects and infertility.

Here, to explore the role of mitochondrial proteins in oocyte developmental competence and female fertility, we screened for pathogenic variants in mitochondria-associated genes in infertile women presenting with oocyte defects. We found that biallelic variants of COX15 were enriched in patients with oocyte defects. COX15 (NM_078470.5; OMIM: 603646) encodes cytochrome c oxidase assembly protein 15 (COX15) (15), a vital regulating factor for respiratory complex IV (COXIV) assembly (16). The effects of COX15 variants on mitochondrial function were investigated in cultured cells, Saccharomyces cerevisiae, and mice models. We found that COX15 deficiency causes mitochondrial dysfunction and triggers oocyte ferroptosis, thus leading to oocyte developmental defects and female infertility.

Results

Enrichment of Compound-Heterozygous COX15 Variants in Patients with Oocyte Defects.

To identify the mitochondria-associated proteins responsible for oocyte developmental defects and female infertility, we performed whole-exome sequencing of 1,024 patients (characterized by oocyte maturation arrest and degenerated or morphologically abnormal oocytes) and 2,868 healthy controls (Fig. 1A). Then the functional variants of 1,688 genes encoding human mitochondria-related proteins according to the MitoProteome Database (http://mitoproteome.org/) were extracted and a standard collapsing analysis under a gene-based burden test for recessive diplotypes was performed (Fig. 1A). The most enriched gene in patients was COX15, which is involved in the assembly of mitochondrial respiratory chain COXIV. Compound-heterozygous COX15 variants were observed in seven independent patients, and none were observed in healthy controls (Fig. 1 A and B and SI Appendix, Fig. S1 A and B and Table S1). The patients from family 1, 2, 4, 5, 6, and 7 all carried the recurrent variant c.C664T [p.R222C] combined with c.1200_1203delGCTG [p. W400*], c.564_565delTG [p.C188Wfs*26], c.C507G [p.Y169*], c.C372A [p.Y124*], c.C821T [p.T274M], and c.C293T [p.S98L], respectively (SI Appendix, Fig. S1C), while the patient in family 3 carried variants c.A655G [p.S219G] and c.G980A [p.R327Q]. All the variants were confirmed by Sanger sequencing and suggested a recessive inheritance pattern (SI Appendix, Figs. S1C and S2A). Variants p.S98L, p.Y124*, p.Y169*, p.S219G, p.R222C, and p.T274M were conserved across different species, including humans, chimps, mice, pigs, sheep, dogs, elephants, chickens, and zebrafish (SI Appendix, Fig. S2B). The variants’ frequencies and positions are shown in SI Appendix, Table S2. This genetic evidence strongly suggests that COX15 is involved in oocyte development and female fertility.

Fig. 1.

Identification of COX15 variants. (A) Strategy for identifying variants related to mitochondrial oxidative phosphorylation leading to poor oocyte quality. (B) The bubble plot for genes with enrichment of rare protein-altering variants. COX15 was the gene with the highest enrichment in rare protein-altering variants. (C) The morphology of normal and affected individuals’ GV, MI, and PB1 oocytes by light microscopy. The normal oocytes extruded PB1 and entered metaphase II. Oocytes from family 1 (II-1), family 5 (II-1), and family 6 (II-1) showed obvious oocyte defects with abnormal morphology. (Scale bar, 25 μm.) (D) The percentage of defective oocytes from the patients’ medical records.

Clinical Features.

All individuals carrying COX15 variants had suffered from primary infertility with unknown causes for many years. They had normal menstrual cycles and their male partners had normal sperm counts and normal sperm morphologic features and motility. According to medical records and patients’ self-reports, all the affected individuals had no other obvious abnormalities or medical disorders. They all underwent several failed in vitro fertilization (IVF) intracytoplasmic sperm injection (ICSI) attempts. The clinical records showed that the majority of the recruited oocytes were immature, degenerated, or had abnormal morphology (Fig. 1 C and D and SI Appendix, Table S1).

The proband in family 1 (II-1) was diagnosed with primary infertility of unknown cause for 13 y. In her first and second IVF cycles, 22 oocytes were retrieved, 15 of which were immature. A total of 5 out of 7 metaphase II (MII) oocytes were fertilized and developed into cleaved embryos, but none resulted in pregnancy. In her third ICSI treatment cycle, 6 oocytes were retrieved, 4 of which were immature oocytes that were degenerated or had abnormal morphology, and the 2 MII oocytes failed to be fertilized (Fig. 1 C and D and SI Appendix, Table S1).

For the proband in family 2 (II-1), in her first IVF cycle, 2 of the 10 retrieved oocytes were dead, 7 were immature, and the 1 MII oocyte failed to be fertilized. In her second ICSI cycle, a total of 16 oocytes were retrieved, and 10 of them were degenerated or had abnormal morphology (Fig. 1D and SI Appendix, Table S1).

For the proband from family 3 (II-1), 6 of 7 oocytes were immature, degenerated oocytes with abnormal morphology in the first ICSI cycle. In her second ICSI cycle, a total of 12 oocytes were retrieved, and the majority of them were degenerated, while 2 MII oocyte could be fertilized but failed to develop into cleaved embryos (Fig. 1D and SI Appendix, Table S1).

For the proband from family 4 (II-1), in her first ICSI cycle, 1 of 3 oocytes was degenerated, and the 1 MII oocyte failed to develop into a cleaved embryo. In her second ICSI cycle, 2 oocytes were immature with unavailable morphology information (Fig. 1D and SI Appendix, Table S1).

For the proband from family 5 (II-1), a total of 20 oocytes were retrieved in her first IVF and second ICSI cycles, and 18 of them were immature oocytes, but the morphology of the oocytes was not recorded. The 2 remaining MII oocytes could be fertilized and developed into cleaved embryos. In her third ICSI cycle, 7 oocytes were retrieved, among which 4 oocytes were degenerated (Fig. 1 C and D and SI Appendix, Table S1).

For the proband from family 6 (II-1), of the 20 oocytes retrieved in her first IVF cycle, 18 oocytes were immature. In her second IVF cycle (15 oocytes) and third IVF cycle (19 oocytes), a total of 34 oocytes were obtained and 30 of them were immature. A total of 16 of the 30 immature oocytes were degenerated or had abnormal morphology (Fig. 1 C and D and SI Appendix, Table S1). In addition, 2 of 4 MII oocytes could be fertilized but failed to form cleaved embryos.

For the proband from family 7 (II-1), in her three IVF cycles, a total of 3 oocytes were obtained, and all of them were degenerated (Fig. 1D and SI Appendix, Table S1).

Together, all seven infertile patients exhibited similar phenotypes characterized by oocyte development defects.

COX15 Variants Impaired Mitochondrial Respiration in S. cerevisiae.

To determine the effects of the identified variants on COX15 function, we first measured the protein level of COX15 by transfecting HeLa cells with plasmids containing wildtype (WT) or mutant plasmids. R217W was used as the positive control because it has previously been reported to cause mitochondrial complex IV deficiency nuclear type 6 (MC4DN6, 615119), an autosomal recessive multisystem metabolic disorder with clinical manifestations including encephalomyopathy, cardiomyopathy, overall developmental delay, and premature death (17). We found that all the nonsense and frameshift variants led to premature termination codons in COX15 (Fig. 2A). The missense variants S219G, R222C, and R217W significantly decreased the COX15 protein level, while variants S98L, T274M, and R327Q had little effect (Fig. 2B and SI Appendix, Fig. S3A). To further evaluate the functional impairment induced by COX15 variants, we performed spot assays using a yeast system. Cox15-deficient yeast strains were generated, and these were transfected with WT and mutant COX15 plasmids. Both WT and mutant strains grew normally on solid Yeast Extract Peptone Dextrose (YPD) medium (which allows for fermentation, a metabolic process that occurs in the absence of oxygen). However, when spotted on the solid Yeast Extract Peptone Glycerol (YPG) medium (which allows for aerobic respiration), the growth of all mutant yeast strains was inhibited to different degrees compared with WT yeast. For variants S98L, T274M, R327Q, Y124*, Y169*, C188Wfs*26, and W400*, as well as the positive control R217W, the yeast could hardly grow on the YPG plates (Fig. 2C). The S219G variant grew on the YPG plates, but the growth rate was significantly reduced compared to WT. For the recurrent variant R222C, the growth rate was also slightly affected (Fig. 2C and SI Appendix, Fig. S3B). These results indicated that the COX15 variants disrupted aerobic respiration in S. cerevisiae and that the physiological function of COX15 was impaired by the patient-identified variants, thus demonstrating the pathogenic effects of these variants.

Fig. 2.

Pathogenic variants in COX15 reduced the protein level and disrupted respiration in S. cerevisiae. (A) Western blot analysis of COX15 nonsense variants in HeLa cells. R217W was the positive control from the MC4DN6 patient. (B) Western blot analysis of COX15 missense variants in HeLa cells. R217W was the positive control from the MC4DN6 patient. (C) The spotting assay of WT and mutated YPH149 yeast on YPD and YPG medium.

Cox15 Deficiency Impaired Fertility in Mice Due to Oocyte Defects.

Next, we asked whether COX15 loss of function causes oocyte defects in vivo. We constructed knock-in mice by introducing variant T277M (corresponding to human T274M) that severely disrupts the function of COX15. As expected, homozygous T277M/T277M female mice were completely infertile. However, these mice also showed a phenotype of severe growth retardation and premature ovarian failure (POF) at an early age (SI Appendix, Fig. S4 A–C), thus making it difficult for us to evaluate the phenotype at the oocyte level. Considering the fact that Cox15 homozygous knockout mice are lethal (18–20), we generated an oocyte-specific Cox15 knockout mouse (referred to as Cox15^{ZP3-Cre, loxP/loxP} or Cox15^OO–/–) (SI Appendix, Fig. S5 A–C). At the age of 10 to 12 wk, homozygous female Cox15^OO–/– mice were subfertile (Fig. 3A and SI Appendix, Fig. S5 D and E), while from 10 mo of age, the mice became completely infertile due to POF with very few follicles remaining in the ovary (SI Appendix, Fig. S5 F–I). To evaluate the effects of COX15 deficiency on oocyte developmental competence, germinal vesicle (GV) oocytes were collected after ovulation stimulation. A total of 20.8% of the GV oocytes from Cox15^OO–/– female mice at 10 to 12 wk of age showed morphological defects (Fig. 3 B and C), and this proportion rose to 72.73% at 10 mo of age (SI Appendix, Fig. S5 J and K). Although cytoplasmic abnormalities (dark cytoplasmic and cytoplasmic granularity) were seen in Cox15-deleted oocytes, these oocytes were alive (SI Appendix, Fig. S6 A and B). The F-actin staining showed that Cox15 deletion had no effects on F-actin intensity or localization in oocytes (SI Appendix, Fig. S6C). Furthermore, the proportion of mature oocytes (MII oocytes) was significantly reduced in Cox15^OO–/– mice (43.5%) compared with control mice (97.03%) (Fig. 3 D and E). The in vitro maturation assay showed that the GV breakdown (GVBD) rate of Cox15 null oocytes was significantly lower than control oocytes (SI Appendix, Fig. S6 D and E). To further assess the developmental potential of the oocytes, we performed IVF with MII oocytes. Most of the MII oocytes derived from Cox15^OO–/– mice could be fertilized, but the proportions of 2-cell, 4- to 8-cell, morula, and blastocyst stage embryos were significantly decreased compared with the control mice (Fig. 3 F and G). These data collectively suggested that COX15 deficiency caused oocyte defects and resulted in impaired fertility in mice, thus recapitulating the phenotype seen in our identified patients.

Fig. 3.

Cox15 oocyte depletion caused female subfertility with oocyte defects and mitochondrial dysfunction. (A) Average pups per litter for control (NC, n = 10) and Cox15^OO–/– (n = 9) female mice. (B) Images of control and Cox15^OO–/– GV oocytes. (C) Percentage of defective GV oocytes from control (n = 219) and Cox15^OO–/– (n = 226) female mice. (D) Representative images of oocytes after hCG stimulation. (E) The PB1 rate of control (n = 115) and Cox15^OO–/– (n = 62) oocytes after ovulation stimulation. (F) The rate of control (n = 67) and Cox15^OO–/– (n = 27) embryo development. (G) Representative images of control and Cox15^OO–/– embryos at the 2-cell, 4- to 8-cell, morula, and blastocyst stages. (H) Mitochondrial morphology analysis of control and Cox15^OO–/– GV oocytes by TOM20 staining. Cox15^OO–/– oocytes showed abnormal mitochondrial location and morphology. (I) Percentage of oocytes with normal and abnormal (accumulated and condensed) mitochondrial location and morphology. Control: n = 59, Cox15^OO–/–: n = 78. (J) Representative images of JC-1 staining. (K) MMP of GV oocytes from control (n = 41) and Cox15^OO–/– (n = 34) oocytes. (L) ROS levels in control (n = 41) and Cox15^OO–/– (n = 45) oocytes. (M) Intensity analysis of (L).

COX15 Deficiency Induced Mitochondrial Dysfunction and ROS Accumulation.

Because COX15 is essential for the physiological function of the mitochondria, we hypothesized that the oocyte defects caused by COX15 deficiency are associated with mitochondria dysfunction. We first evaluated the morphology and distribution of mitochondria by TOM20 staining. The results showed that 56.4% of Cox15^OO–/– oocytes exhibited abnormal mitochondrial distribution (Fig. 3 H and I). In particular, 33.3% of the oocytes showed abnormal accumulation of mitochondria around the GV (Fig. 3 H and I), and 23.1% of the oocytes manifested with mitochondrial aggregation into large condensates (Fig. 3 H and I). We then assessed the mitochondrial membrane potential (MMP) by JC-1 probes and found that the MMP was significantly decreased in Cox15^OO–/– oocytes (Fig. 3 J and K). Taken together, our results indicate that COX15 deficiency alters the dynamic properties of mitochondria.

Because the main function of mitochondria is energy production, we next measured the intracellular ATP level of GV oocytes, MII oocytes, and 2-cell embryos. Unexpectedly, the ATP levels in Cox15-deficient oocytes and embryos were comparable with the control group (SI Appendix, Fig. S6 F–H). This result was further confirmed by the observation that there was no difference in the cellular energy sensor p-AMP-dependent kinase (pAMPK) between the two groups (SI Appendix, Fig. S6 I and J). Considering that increased ROS levels usually accompany mitochondrial dysfunction, we next measured ROS levels and found that ROS levels were significantly increased in Cox15^OO–/– oocytes (Fig. 3 L and M). It has been reported that excessive ROS may cause DNA damage and lead to apoptotic cell death in cultured cell lines (21, 22), and we further checked the status of DNA damage and apoptosis by γH2AX staining (a sensitive molecular marker of DNA damage and repair), terminal deoxynucleotidyl transferase (TdT) dUTP nick-end labeling (TUNEL) staining and the cleaved caspase-3 staining. However, both DNA damage and apoptosis remained at low levels in both Cox15^OO–/– and control oocytes, and no difference was found between the two groups. In contrast, UV irradiation treatment (the positive control) resulted in high γH2AX and TUNEL signal (SI Appendix, Fig. S7 A and B). These results suggested that COX15 deficiency caused mitochondrial dysfunction and ROS accumulation without effects on DNA damage and apoptosis.

COX15 Deficiency Triggers Ferroptosis Leading to Oocyte Defects.

To determine the molecular mechanism that directly caused the oocyte defects, we performed RNA sequencing (RNA-seq) of GV oocytes and 2-cell embryos from Cox15^loxP/loxP mice (controls) and Cox15^OO–/– mice. The transcriptome of Cox15^OO–/– oocytes and embryos was significantly changed (Fig. 4 A and B and SI Appendix, Fig. S8A). Interestingly, we found that Mt1 was the only overlapping gene among the top 10 differentially expressed genes (DEGs) between GV oocytes and 2-cell embryos (SI Appendix, Fig. S8B). qRT-PCR confirmed the significant upregulation of Mt1 induced by COX15 deficiency (SI Appendix, Fig. S8 C and D). Mt1 encodes metallothionein 1, which belongs to the family of small cysteine-rich metal-binding proteins involved in metal homeostasis and detoxification (23). Thus, we hypothesized that the Cox15 deficiency-induced oocyte defects are modulated by metal ion-mediated cell death (cuproptosis or ferroptosis). Considering that the MT1 protein has a higher affinity for copper ions, we first checked the level of FDX1, a marker of cuproptosis, and found that Cu2+ and FDX1 protein levels showed little difference between Cox15^OO–/– and control oocytes (SI Appendix, Fig. S8 E–H), suggesting that cuproptosis was not activated.

Fig. 4.

The RNA-seq analysis of GV oocytes and 2-cell embryos indicated the occurrence of metal ion-mediated cell death. (A) Volcano plot for control (n = 5) and Cox15^OO–/– (n = 5) oocytes. (B) Volcano plot for control (n = 5) and Cox15^OO–/– (n = 5) 2-cell embryos. (C) FerroOrange (an Fe²⁺ indicator) staining of control oocytes (n = 26) and Cox15^OO–/– oocytes (n = 35). (D) The intensity analysis of (C). (E) The intensity analysis of Nile red staining. (F) Nile red staining of control oocytes (n = 32) and Cox15^OO–/– oocytes (n = 42). (G) The relative intensity of BODIPY 581/591 C11 probe staining of oxidized lipids and reduced lipids in control oocytes (n = 22) and Cox15^OO–/– oocytes (n = 18). (H) Percentage of abnormal GPX4 localization in control oocytes and Cox15^OO–/– oocytes. (I) GPX4 staining of control (n = 40) and Cox15^OO–/– oocytes (n = 35).

We next used FerroOrange staining [an orange fluorescent probe that specifically detects labile iron (II) ions (Fe²⁺)] to test for ferroptosis and found a significant accumulation of Fe2+ in Cox15^OO–/– oocytes (Fig. 4 C and D), suggesting that ferroptosis was activated. We confirmed this by checking several other markers of ferroptosis. Nile red staining showed obvious lipid accumulation (Fig. 4 E and F). The BODIPY 581/591 C11 probe (a lipid peroxidation sensor) was further used to detect ROS levels in cells and membranes. The results showed that oocytes from Cox15^OO–/– mice exhibited significantly increased lipid ROS production (SI Appendix, Fig. S9A and Fig. 4G). To further confirm the role of Cox15 in ferroptosis, we measured the protein level of glutathione peroxidase 4 (GPX4, another marker for ferroptosis). The immunostaining results showed that 22.2% of GV oocytes from Cox15^OO–/– mice showed abnormal protein localization (Fig. 4 H and I). Although the intensity quantification of GPX4 showed no obvious difference between control and Cox15-deleted oocytes (SI Appendix, Fig. S9B), the western blot of GPX4 showed an increased protein level in Cox15^OO–/– oocytes (SI Appendix, Fig. S9 C and D). In addition, the average size of the mitochondria in Cox15^OO–/– oocytes was also smaller than controls (SI Appendix, Fig. S9 E and F). It has been reported that ferroptosis can change mitochondrial morphology and cristae structure causing smaller mitochondria than normal controls (24, 25). Thus, the presence of smaller mitochondria in Cox15-deleted oocytes may be the consequence of ferroptosis activation. To determine the effects of Cox15 deletion on growing oocytes, control and Cox15-deleted oocytes at postnatal day 7(P7) were retrieved. Although the morphology of growing Cox15^OO–/– oocytes showed no obvious defects (SI Appendix, Fig. S10A), the FerroOrange staining indicated a significant accumulation of Fe2+ in growing Cox15^OO–/– oocytes (SI Appendix, Fig. S10 A and B). Also, JC-1 staining showed a significant decrease of MMP in growing Cox15^OO–/– oocytes (SI Appendix, Fig. S10 C and D). These findings indicated that COX15 is required for oocyte developmental competence by maintaining mitochondrial function during early stages of oogenesis. To determine the underlying mechanism, we reanalyzed the RNA seq data from GV stage oocytes and found that the expression of Pde family genes was significantly decreased in COX15-deleted oocytes (SI Appendix, Fig. S11A). This result was confirmed by using qRT-PCR for Pde4d, Pde11a, and Pde7b (SI Appendix, Fig. S11B). Previous studies showed that PDE4D deletion or inhibiting its activity caused GV arrest (26, 27). Therefore, we inferred the GV arrest phenotype in COX15-deleted oocytes may result from the downregulation of Pde family genes.

To further confirm the effects of Cox15 variants on oocyte quality and female fertility, we generated double knock-in (DKI) mouse models R225C/S101L [corresponding to the variant R222C/S98L in family 7(II-1)] and R225C/T277M [corresponding to the variant R222C/T274M in family 6(II-1)]. The litter sizes of both R225C/S101L and R225C/T277M female mice were significantly decreased compared with normal controls, indicating a subfertility phenotype (Fig. 5A). The body size, ovary size, and oocyte GVBD rate of both DKI mice showed no obvious difference compared to control mice (SI Appendix, Fig. S12 A and B and Fig. 5B). However, the first polar body (PB1) rate and the morphology of oocytes from R225C/T277M mice were significantly affected compared with normal controls (Fig. 5 C and D). Meantime, the developmental competence of embryos was also decreased in both DKI mice (Fig. 5E and SI Appendix, Fig. S12C). Furthermore, the GV oocytes from both DKI female mice showed increased Fe2+ and ROS levels, indicating the activation of ferroptosis (Fig. 5 F and G and SI Appendix, Fig. S12 D and E). Taken together, the DKI mice with the patient-derived variants could recapitulate the oocyte defects phenotype, thus establishing a direct causal relationship between COX15 variants and oocyte quality. We thus concluded that COX15 deficiency triggers ferroptosis, which contributes to the observed oocyte defects.

Fig. 5.

The R225C/S101L and R225C/T277M DKI female mice showed oocyte defects and female subfertility. (A) Average pups per litter for control and DKI mice. (B) The GVBD rate of control and DKI female mice. (C) The PB1 rate of control and DKI female mice. (D) The representative images of MII oocytes after in vivo ovulation stimulation. (E) The rate of control and DKI embryo development. (F) The relative intensity of Fe²⁺ in control and DKI oocytes. (G) The relative intensity of ROS in control and DKI oocytes.

The Activation of Ferroptosis is Associated with COXIV Disruption in Oocytes.

COX15 participates in COXIV assembly by regulating the maturation of COX1 (a core subunit of cytochrome c oxidase) (16). To determine whether Cox15 deletion disrupted COXIV assembly, we measured the protein level of COX1. We found that the COX1 level was sharply decreased in Cox15^OO–/– oocytes, implying the disruption of COXIV assembly (Fig. 6 A and B). To further confirm that the activation of ferroptosis was the result of COXIV disruption, we inhibited the activity of COXIV using NaN3 and found that inhibition of COXIV caused morphological defects in WT mouse oocytes (Fig. 6 C and D). Meanwhile, the rate of GVBD and PB1 extrusion were both significantly reduced following treatment in WT mouse oocytes (Fig. 6 E and F). Of note, the NaN3 treatment significantly increased lipid peroxidation, Fe²⁺, and ROS levels in oocytes (Fig. 6 C and D and SI Appendix, Fig. S13 A–D). The dosage of NaN3 needed to activate ferroptosis had no obvious effects on apoptosis (SI Appendix, Fig. S13 E and F). Consistent with the results in mouse oocytes (SI Appendix, Fig. S13 A–H), COXIV inhibitor treatment caused human oocyte defects by inducing ferroptosis (Fig. 6 G and H). We also performed serial assays showing the similarities between Cox15^oo−/− and NaN3 treatment. First, NaN3 treatment could slightly increase the intensity of GPX4 in GV oocytes but had little effect on the protein level of COX1(SI Appendix, Fig. S14 A–D). Second, the qRT-PCR showed that NaN3 treatment could significantly increase the expression level of Mt1 in growing oocytes (SI Appendix, Fig. S14E). The effects of NaN3 on Mt1 and GPX4 were consistent with Cox15 deletion, while there is a different effect of Cox15 deletion and NaN3 treatment on COX1. Apart from inhibiting COXIV, NaN3 treatment can also decrease the ATP levels by inhibiting F0F1 ATPase (SI Appendix, Fig. S14F). Then, to exclude the possibility that ferroptosis activation was associated with F0F1 ATPase inhibition, we treated oocytes with a specific F0F1 ATPase inhibitor Oligomycin A. The data showed that Oligomycin A treatment could decrease the ATP level, but didn’t elevate the Fe²⁺ and ROS levels (SI Appendix, Fig. S14 G–K). All these results showed that activation of ferroptosis was due to the inhibition of COXIV and not F0F1 ATPase. Taken together, these results suggest that COX15 deficiency-triggered ferroptosis is associated with the disruption of COXIV assembly.

Fig. 6.

Fer-1 rescued the low maturation rate and poor morphology of oocytes in vitro and ex vivo. (A) MT-COX1 immunofluorescence indicated a sharp decrease in protein level in Cox15^OO–/– oocytes (n = 24) compared with control oocytes (n = 23). (B) Intensity analysis of (A). (C) COXIV inhibition caused Fe²⁺ accumulation and abnormal morphology in oocytes, thus indicating the activation of ferroptosis. Fer-1 rescued the changes in Fe²⁺ levels and oocyte morphology induced by COXIV inhibitor treatment. (D) The intensity analysis of (C). (E) The GVBD rate in oocytes with vehicle control treatment, NaN3 treatment, and NaN3+Fer-1 treatment. COXIV inhibition caused a low GVBD rate, and Fer-1 could partly rescue the decreased GVBD rate. (F) The PB1 rate for oocytes with control treatment, NaN3 treatment, and NaN3+Fer-1 treatment. COXIV inhibition caused a low PB1 rate, and Fer-1 could partly rescue the decreased PB1 rate. (G) Representative images of human GV oocytes with or without COXIV inhibitor, n = 3. (H) Fe²⁺ levels in human GV oocytes with or without COXIV inhibitor treatment, n = 3. (I) Representative images of PB1 oocytes from control and Cox15^OO–/– female mice with or without Fer-1 treatment. (J) The PB1 rate of oocytes from control and Cox15^OO–/– female mice with or without Fer-1 treatment.

Inhibition of Ferroptosis Alleviates Oocyte Defects.

Because COX15 deficiency led to oocyte defects by inducing ferroptosis, we next used the inhibitor of ferroptosis Fer-1 to test whether inhibiting ferroptosis could rescue the phenotype. Compared with oocytes treated with NaN3, Fer-1 inhibited the activation of ferroptosis in cultured oocytes (Fig. 6 C and D and SI Appendix, Fig. S13 A–F). Correspondingly, the oocyte morphological defects, GVBD rate, and PB1 rate were also significantly rescued (Fig. 6 E and F). However, Fer-1 did not reduce lipid peroxidation in Cox15-deleted oocytes in vitro (SI Appendix, Fig. S13 G and H). There might be two possible reasons for this. First, Cox15 deletion affected the growing oocyte as early as P7(SI Appendix, Fig. S10), and thus the intervention at the GV stage might be too late. Second, the limited time window for in vitro treatment might be insufficient to rescue the lipid peroxidation. To further confirm this effect in vivo, Fer-1 was intraperitoneally injected into Cox15^OO–/– mice and control mice every 2 d for 1 mo, with an equal dose of corn oil used as the vehicle control. The results showed that Fer-1 treatment could clearly improve the oocyte quality as indicated by the significantly reduced proportion of abnormal oocytes and increased rate of PB1 oocytes (Fig. 6 I and J). We also tried other rescue strategies. CAY10650, an inhibitor of cytosolic phospholipase A2α (cPLA2α) which involved in lipid droplet formation (28), could effectively decrease the intensity of lipid droplets at 12 nM and 120 nM after a 12 h treatment (SI Appendix, Fig. S15 A and B). However, CAY10650 treatment had no obvious effect on the morphology of control and Cox15-deleted oocytes, indicating that removal of lipid droplets could not improve the morphological defects of Cox15-deleted oocytes (SI Appendix, Fig. S15C). We injected Cox15 cRNA into Cox15-deleted oocytes at the GV stage. However, the oocyte morphology and GVBD rate were not rescued (SI Appendix, Fig. S15 D and E). In addition, the antioxidant N-Acetyl Cysteine (NAC) could not rescue the reduced GVBD rate of Cox15-deleted oocytes (SI Appendix, Figs. S6 D and E and S15 F and G). Because Cox15 depletion affected the growing oocytes as early as at P7 (SI Appendix, Fig. S10), intervention at the GV stage might be too late. These findings suggested that inhibition of ferroptosis can alleviate the oocyte defects triggered by COX15 dysfunction.

Discussion

Proper oocyte maturation is required for successful human reproduction, and oocyte defects usually result in female infertility, which manifests in different forms including oocyte maturation arrest, fertilization failure, zygotic cleavage failure, and early embryonic arrest (29). Mitochondrial dysfunction and ROS accumulation have been proposed to be among the main causes of the decline of oocyte quality that leads to lower rates of fertilization and reduced embryo survival with age (10). However, there are no convincing studies on oocyte defects and female infertility resulting from variants in genes encoding mitochondrial proteins. In this study, we report that biallelic COX15 variants cause human oocyte ferroptosis and infertility and show that Cox15 oocyte deletion impaired Fe²⁺ and ROS homeostasis thus causing mitochondrial dysfunction and ultimately sensitizing oocytes to ferroptosis (SI Appendix, Fig. S16). Our findings provide direct evidence that variants in COX15 are responsible for abnormal oocyte development and competence and expand our understanding of mitochondrial disorders to female infertility.

The biallelic hotspot variant R217W in COX15 has previously been shown to cause MC4DN6 (13, 17, 30, 31), an autosomal recessive multisystem metabolic disorder, whereas the compound heterozygous hotspot variant R222C in COX15 that we identified here was found to be responsible for female infertility mainly characterized by oocyte developmental defects (Fig. 1 C and D and SI Appendix, Fig. S2B). There are some possible explanations for the two distinct diseases resulting from different hotspot variants in COX15. First, the R217W variant from the affected individuals with MC4DN6 usually causes severe dysfunction of both alleles (13, 17, 30, 31), while the R222C variant we identified had a weaker effect than R217W (Fig. 2). This suggests that different dosage effects of COX15 variants have different effects on human oocytes and somatic cells, respectively, which results in the two distinct diseases. Second, compared to somatic cells, oocytes undergo long periods of dormancy in the ovary and lack the ability for self-renewal, which may lead to increased accumulation of the effects of corresponding variants in oocytes. Thus, variants with only low or moderate effects in somatic cells might cause severe phenotypes in oocytes. This was further confirmed by the observation that aged Cox15^OO–/– mice had more severe phenotypes than young mice (Fig. 3 A–G and SI Appendix, Fig. S5). Similar results were also observed in our previous study showing that different types of TRIP13 recessive pathogenic variants cause distinct diseases through different effects on mitosis (Wilms tumors) and meiosis (oocyte maturation arrest) (32, 33).

Lipid peroxidation is a hallmark of ferroptosis. Previous work showed that increasing lipid peroxidation could raise membrane tension, which in turn activated Piezo1 and transient receptor potential mechanosensitive ion channels and caused the aberrant movement of ions across the plasma membrane (34). This led to a loss of ionic homeostasis and osmotic cell swelling that eventually resulted in plasma membrane deformation (35). Thus, the abnormal morphology phenotype in Cox15-deleted oocytes might be caused by lipid peroxidation of the plasma membrane.

Human primordial oocytes are established during fetal development and survive in the ovary for up to 50 y (12). During their long-term dormancy, oocytes must maintain a low ROS metabolism in order to avoid the toxic effects on nucleic acids (DNA and RNA), proteins, and lipids (20, 36). A previously proposed mechanism is that early oocytes prevent the generation of toxic ROS by suppressing complex I (12). COXIV, the terminal enzyme in the respiratory chain, is responsible for ROS clearance by promoting electron binding to the ultimate O₂ acceptor to generate H2O. However, the role of COXIV in oocyte developmental competence (oocyte quality) and reproductive longevity is less well understood. We found that Cox15 deficiency impaired ROS and Fe²⁺ homeostasis in human and mouse oocytes thus causing mitochondrial dysfunction and ultimately sensitizing the oocytes to ferroptosis, thus providing direct evidence for the necessity of the oocyte to maintain an ROS-free metabolic state. Unexpectedly, Cox15 deficiency had little effect on DNA damage and apoptosis, which usually lead to the accumulation of ROS and lead to apoptosis in cultured cell lines, further suggesting the unique feature of oocytes compared to somatic cells (21, 22). Oocytes in primordial follicles undergo apoptosis by the TAp63-dependent signaling pathway (37), and the decreasing expression level of TAp63 in growing oocytes reduces the capacity of oocytes to respond to apoptotic signals (38). Furthermore, pro-apoptotic factors such as Pmaip1 show low expression levels during the oocyte maturation process, but they become highly expressed after the 8-cell stage, while anti-apoptotic factors such as BCL2L10 and BCL2L2 show high expression in early oocytes but low expression after the 8-cell stage (39–41). In this process, the apoptosis pathway gradually switches to the DNA repair pathway, and this can effectively safeguard against DNA damage (38). All these findings indicate that oocytes in growing and maturation stages are more sensitive to ferroptosis, but not to apoptosis. Thus, even in the presence of increased ROS, Cox15 deletion had little effect on DNA damage.

It has been reported that during oocyte development, both OXPHOS and glycolysis are involved to support ATP production for oocytes (42, 43). COX15 depletion can disrupt OXPHOS, glycolysis may compensate for the ATP production. In addition, mouse oocytes can generate ATP from AMP via the adenosine salvage pathway (44). These observations indicated that compensatory energy production pathway may exist to maintain the ATP levels in oocytes. Thus, we propose that glycolysis and adenosine salvage pathways might complement the ATP levels in Cox15-deleted oocytes.

Mitochondrial replacement therapies or techniques including pronuclear transfer, maternal spindle transfer, and first or second polar body transfer have been an effective reproductive treatment to overcome the transmission of maternally inherited mitochondrial DNA variants (45, 46). But there is still a lack of effective strategies for the treatment of mitochondrial disorders. In this study, we tried to rescue the phenotype of oocyte-specific Cox15-deleted mice by inhibiting ferroptosis. The results showed that inhibition of ferroptosis alone could alleviate oocyte defects triggered by COX15 dysfunction. This may be a potential individualized therapy for patients with COX15 dysfunction or ferroptosis activation.

In summary, we identified biallelic pathogenic variants in COX15 that cause female infertility and human oocyte ferroptosis. We found that Cox15 deletion impaired the redox balance and lipid metabolism in oocytes causing mitochondrial dysfunction and ultimately sensitizing oocytes to ferroptosis. Our findings not only show direct evidence for the essential roles of mitochondrial proteins in human oocyte development but also expand the spectrum of mitochondrial disorders related to female infertility and provide unique insights into the role of ferroptosis in oocyte defects.

Materials and Methods

Human Subjects.

Unrelated patients with oocyte maturation arrest and healthy controls were recruited from the Reproductive Medicine Center of the Shanghai Ninth Hospital affiliated with Shanghai Jiao Tong University, the Department of Reproductive Medicine of Binzhou Medical University Hospital, the Reproductive Medicine Center of the Third Hospital Affiliated to Guangxi Medical University, the Reproductive Medicine Center of Shaanxi Maternal and Child Care Service Center, the Department of Reproductive Medicine of Kunming Angel Women’s and Children’s Hospital, and the Shanghai Ji’ai Genetics and IVF Institute. Studies of human subjects were approved by the Fudan University Medicine Institutional Review Board. All oocytes from controls and patients were obtained with written informed consent signed by the donor couples. The study was approved by the Reproductive Study Ethics Committee.

Genetic Studies.

The procedure of whole-exome sequencing and quality control has been described in previous studies (47). According to the MitoProteome Database (http://mitoproteome.org/), 1,705 mitochondrial genes were mapped with our WES data, and then the variants of the 1,688 genes related to mitochondrial function and oxidative phosphorylation were extracted. The gene burden of rare variants in cases and unaffected controls was processed with the denovolyzeR package according to a previous report (47). The enrichment factors of the 1,688 genes related to mitochondrial function and oxidative phosphorylation were ranked according to P-values. Details are shown in SI Appendix.

Generation of CRISPR/Cas9 Genome Editing Mouse Models.

The oocyte deletion mice (Cox15^OO–/–) and knock-in mice (R225C, T277M, and S101L) models were generated by Cyagen Biosciences Inc. Details are shown in SI Appendix. Primers used in this study are shown in SI Appendix, Table S3.

Expression Constructs, Transfection, and Western Blotting.

The coding sequence of human COX15 (NM_078470) was cloned into the PCMV6-entry vector. Variants in COX15 were introduced using the site-directed KOD-Plus-Mutagenesis Kit (Toyobo Life Science). Cell culture and western blot were performed according to a standard protocol (48). Details are shown in SI Appendix. Antibodies used in this study are listed in SI Appendix, Table S4.

Yeast Strains and Handling.

All S. cerevisiae strains used in this study were derived from YPH499. Generation of COX15 deletion and mutant yeast stains and growth tests assay were performed according to previous procedure (16). Details are shown in SI Appendix.

ATP Measurements.

For intracellular ATP measurement, 5 to 10 GV oocytes, MII oocytes, and 2-cell embryos per group were lysed with 200 μL of lysis buffer for 5 min and centrifuged at 4,000×g at room temperature for 30 s. A mixture of 30 μL of supernatant and 30 μL of luciferine-luciferase was assayed, as described previously (48). The relative ATP concentration was expressed as the ratio of all values with respect to the WT group.

Mouse Oocyte Collection, Fertilization, and Embryo Culture In Vitro.

The oocyte retrieving, IVF, and embryo culture were performed as described previously (49). All mouse experiments were reviewed and approved by the Shanghai Medical College of Fudan University.

Transmission Electron Microscopy.

The ovaries of WT and Cox15^OO–/– female mice were collected and fixed for transmission electron microscopy (TEM, Philips CM-120, Netherlands) analysis according to standard TEM procedures previously described (50).

Immunofluorescence.

Immunostaining of oocytes was performed according to a standard protocol. The primary antibodies anti-COX1, anti-GPX4, anti-TOM20, and anti-phos-γh2AX were used to label target proteins. Details are shown in SI Appendix.

Evaluation of Oocyte Mitochondria, Ferroptosis, and Cuproptosis.

Mitochondrial ROS levels were measured using a Reactive Oxygen Species Assay Kit (S0033S, Beyotime, China). Nile red, BODIPY^TM 581/591 C11 (D3861, Invitrogen), FerroOrange (F374, Dojindo, Japan), rhodamine B hydrazide and FDX1 was staining for ferroptosis and cuproptosis analysis. Details are shown in SI Appendix.

RNA-Seq Analysis.

GV oocytes and 2-cell embryos were collected from control and Cox15^OO–/– female mice (10 to 12 wk old), and total RNA from oocytes was extracted using a RNeasy Mini Kit (Qiagen). The Single Cell/Low Input RNA sequencing was performed. Details are shown in SI Appendix.

In Vitro and Ex Vivo Treatment with Fer-1.

For the in vitro rescue assay, Fer-1 was added along with 0.01% NaN3 for 4 h. For the ex vivo assay, 10-wk-old female mice were treated with Fer-1 for 4 wk. The mice received intraperitoneal injections every 2 d of Fer-1 (1 mg/kg) or vehicle (corn oil). Details are shown in SI Appendix.

Statistics.

Statistical analysis was performed using GraphPad Prism. Values were analyzed by two-tailed Student’s t test for two experimental groups or by one-way ANOVA for more than two groups. P-values were calculated with two-tailed unpaired Student’s t test or chi-square test as described in the figure and table legends. *P < 0.05, **P < 0.01, and ***P < 0.001 (P > 0.05, not significant).

Study Approval.

Studies of human subjects were approved by the Fudan University Medicine Institutional Review Board and the Reproductive Study Ethics Committee of the Shanghai Ninth Hospital affiliated with Shanghai Jiao Tong University and the Shanghai Ji’ai Genetics and IVF Institute. Written informed consent was obtained from all the participants. The animal experiments were performed according to the procedures approved by the Experimental Animal Ethics Committee of Fudan Shanghai Medical College.

Supplementary Material

Appendix 01 (PDF)

Dataset S01 (XLS)

Data, Materials, and Software Availability

The raw data used in the study have been deposited in the Genome Variation Map and Genome Sequence Archive at the National Genomics Data Center, Beijing Institute of Genomics, Chinese Academy of Sciences under accession numbers GVM000497 (47) and GSA:CRA014380 that are publicly accessible at https://ngdc.cncb.ac.cn/gsa. Variants of the control cohort used in this study were generated by the HuaBiao project and can be obtained from https://www.biosino.org/wepd/.