Maternal MLH3 dysfunction drives unexplained recurrent pregnancy loss via impaired oocyte maturation and defective decidualization.

We performed whole exome sequencing (WES) on peripheral blood samples from 94 unexplained recurrent pregnancy loss (URPL) patients who experienced more than 2 times first trimester abortion (Supp. Table 2 ). Strikingly, patient 1 exhibited a notable history of four early pregnancy losses with her former partner and one with her current husband (Supp. Table 1 , Fig.  1 A). WES revealed one frameshift deletion (p.G1302Efs*21) and one missense variant (p.G752E) in MLH3 gene, both located within functionally critical domains (Fig.  1 B-C). Subsequent WES of chorionic villus tissue detected only the p.G1302Efs*21 variant, confirming a compound heterozygous state with the p.G752E variant. Importantly, embryonic karyotyping ruled out aneuploidy as a contributing factor, and comprehensive clinical evaluation of patient 1 revealed no systemic abnormalities or aberrant tumor markers that could account for the reproductive phenotype (Supp. Table 2 ). Furthermore, cohort screening identified four additional URPL cases harboring rare mutations in the MLH3 gene (Fig.  1 B-C). In total, five patients in our cohort carried rare MLH3 variants, all of which were absent or extremely rare in population databases (gnomAD frequency: 0–0.011%). Notably, all identified variants affect amino acid residues that are highly or moderately conserved across diverse vertebrate species, suggesting a potential deleterious impact on MLH3 protein structure or function (Supp. Fig.  1 ). Consistently, none of the four newly identified cases showed systemic abnormalities or aberrant tumor markers (Supp. Table 2 ). The observed mutation frequency in our URPL cohort (5/94, 5.32%) thus significantly exceeds that in the general population, collectively suggesting a pathogenic role for MLH3 dysfunction in URPL. Moreover, DevOmics data [ 10 ] confirmed predominant MLH3 expression during oogenesis (Supp. Fig. 2 ). Accordingly, we performed spatiotemporal localization analysis in mouse oocytes, demonstrating consistent Mlh3 expression across all meiotic stages (GV, GVBD, MI, and MII; Figure. 1 D). These findings collectively implicate MLH3 dysfunction as a potential contributor to URPL, possibly through compromising oocyte quality and meiotic integrity, which may in turn undermine embryonic developmental competence and predispose to early pregnancy loss. Fig. 1 Five URPL patients in our cohort carried rare MLH3 variants. A . Abortion history of patient 1. B - C . MLH3 mutational sites and functional impacts in URPL patients. D . Mlh3 localization in mouse oocytes at stages of GV, GVBD, MI and MII. bar = 20 μm Five URPL patients in our cohort carried rare MLH3 variants. A . Abortion history of patient 1. B - C . MLH3 mutational sites and functional impacts in URPL patients. D . Mlh3 localization in mouse oocytes at stages of GV, GVBD, MI and MII. bar = 20 μm To functionally characterize Mlh3’s role in oocyte development, we employed Trim-Away technology to specifically deplete Mlh3 protein in mouse GV-stage oocytes (Fig.  2 A). Quantitative immunofluorescence analysis revealed a significant reduction in Mlh3 signal intensity (83.88% decrease, p  < 0.001) at 16 h after microinjection (Fig.  2 B). Subsequent in vitro maturation and fertilization assays demonstrated that Mlh3 depletion markedly impaired oocyte developmental competence, manifesting as: (1) 38% reduction in MII maturation rate (control:72% vs. Mlh3-depletion:34.67%, p  < 0.01); (2) 52% decrease in fertilization rate (control:82.4% vs. Mlh3-depletion: 30.76%, p  < 0.01); (3) 36.8% decline in 8-cell formation (control:61.8% vs. Mlh3-depletion: 25%, p  < 0.001); (4) 38.63% decline in blastocyst formation (control: 63.38% vs. Mlh3-depletion: 25%) (Fig.  2 C-D). To investigate whether Mlh3 loss compromises genomic integrity in early embryos, we collected 2-cell-stage embryos from both control and Mlh3-depleted groups for whole-genome amplification using MALBAC, followed by low-pass whole-genome sequencing. Copy number variation (CNV) analysis revealed focal chromosomal imbalances in the Mlh3-depletion group that were absent in controls, notably on chromosomes 6 and 9, suggesting that Mlh3 deficiency perturbs genome stability during early embryonic cleavage (Fig.  2 E). Furthermore, we assessed the quality of blastocysts that managed to develop despite Mlh3 depletion by performing immunofluorescence staining for cell nuclear. Although blastocyst morphology appeared grossly normal, quantitative analysis revealed a significantly reduced total cell number in Mlh3-depleted blastocysts, indicating compromised developmental potential even in embryos that successfully reached the blastocyst stage (Fig.  2 F, Supp. Fig. 3 ). Fig. 2 Mlh3 depletion impairs oocyte and early embryos developmental potential. A  Trim-away workflow. B  After 16 h of in vitro culture post microinjection, Mlh3 fluorescence intensity was significantly reduced. C - D  Decreased rates of maturation, fertilization, and blastocyst formation after Mlh3 depletion. E  CNV analysis of 2-cell-stage embryos from both control and Mlh3-depleted groups. F  Blastocyst cell numbers in control versus Mlh3-depleted groups. Pb = polar body. CNV=Copy number variation. p  < 0.05, p  < 0.01. bar = 20 μm Mlh3 depletion impairs oocyte and early embryos developmental potential. A  Trim-away workflow. B  After 16 h of in vitro culture post microinjection, Mlh3 fluorescence intensity was significantly reduced. C - D  Decreased rates of maturation, fertilization, and blastocyst formation after Mlh3 depletion. E  CNV analysis of 2-cell-stage embryos from both control and Mlh3-depleted groups. F  Blastocyst cell numbers in control versus Mlh3-depleted groups. Pb = polar body. CNV=Copy number variation. p  < 0.05, p  < 0.01. bar = 20 μm Collectively, these results show that MLH3 is essential for oocyte quality, and its loss compromises subsequent embryonic development, leading to chromosomal instability and reduced blastocyst formation. This highlights MLH3’s critical role in ensuring reproductive success by supporting both oocyte competence and early embryo viability. Furthermore, to assess the in vivo developmental competence of Mlh3-deficient embryos, we performed embryo transfer experiments into pseudopregnant surrogate females. An average of 15 blastocysts from either the Mlh3-depleted or control group were transferred per recipient, with each group assigned to separate surrogates (Fig.  3 A). All recipients were dissected at E13.5, and implanted embryos, viable fetuses, and placentas were quantified. In the control group, 28 embryos implanted successfully, all of which developed into viable fetuses, yielding an abortion rate of 0%. In contrast, although 36 embryos implanted in the Mlh3-deficient group, only 13 progressed to viable fetuses, corresponding to a significantly elevated abortion rate of 63.89% (Fig.  3 B-D). Moreover, crown-rump length and fetal weight were markedly reduced in the Mlh3-deficient group compared to controls, while placental size and weight showed no significant differences (Fig.  3 E-F; Supp. Table 4 ). These results demonstrate that Mlh3 depletion compromises embryonic developmental potential beyond implantation, leading to pregnancy loss, a phenotype consistent with our clinical observations in URPL patients carrying MLH3 variants. Fig. 3 in vivo developmental competence of Mlh3-deficient embryos. A  A schematic graph for delivery of Mlh3-deficient and control blastocysts into the mouse uterus. B  Images of the uterus, fetus and placentas in the two groups at E13.5. Fetus number ( C ) and abortion rate ( D ) of the two groups. Crown-rump length ( E ), placenta length ( F ), fetus weight ( G ) and placenta weight ( H ) of the two groups. * p  < 0.05, ** p  < 0.01, *** p  < 0.001 in vivo developmental competence of Mlh3-deficient embryos. A  A schematic graph for delivery of Mlh3-deficient and control blastocysts into the mouse uterus. B  Images of the uterus, fetus and placentas in the two groups at E13.5. Fetus number ( C ) and abortion rate ( D ) of the two groups. Crown-rump length ( E ), placenta length ( F ), fetus weight ( G ) and placenta weight ( H ) of the two groups. * p  < 0.05, ** p  < 0.01, *** p  < 0.001 Given our observation that Mlh3 localizes to the meiotic spindle in mouse oocytes, with notable enrichment along chromosomes (Fig.  1 D), we hypothesized that Mlh3 contributes to spindle integrity and chromosome alignment. To test this, we acutely depleted Mlh3 in germinal vesicle (GV)-stage oocytes using Trim-Away technology. Mlh3 degradation resulted in severe spindle abnormalities, including multipolar configurations, loss of bipolar fusiform structure, and disorganized chromatin failing to congress at the metaphase plate (Fig.  4 A), indicating a critical role for Mlh3 in meiotic spindle assembly and stability. To definitively establish Mlh3’s role in spindle dynamics, we performed co-immunoprecipitation of Mlh3-interacting proteins from mouse MII oocytes followed by mass spectrometry analysis. The Mlh3 interactome showed significant enrichment (FDR < 0.05) in three key biological processes, including regulation of DNA metabolic process, chromatin remodeling, and microtubule cytoskeleton organization. Notably, cyclin-dependent kinase 1 (CDK1), a master regulator of meiotic spindle organization, was uniquely identified as a high-confidence interacting protein, showing functional convergence across all three enriched pathways (Fig.  4 B). Immunofluorescence imaging showed that CDK1 is broadly distributed in the ooplasm and prominently enriched on the meiotic spindle at the MII stage. Quantitative fluorescence intensity profiling further revealed substantial co-localization between CDK1 and MLH3 in both the cytoplasm and spindle apparatus (Fig.  4 C). Reciprocal Co-IP experiments confirmed this physical interaction: anti-MLH3 antibodies successfully pulled down CDK1, and conversely, anti-CDK1 antibodies co-precipitated MLH3 (Fig.  4 D). Together, these data provide direct biochemical and spatial evidence that MLH3 forms a complex with CDK1, positioning MLH3 as a potential modulator of spindle assembly through CDK1-mediated regulatory networks during oocyte meiosis. Moreover, mechanistic investigation focused on BubR1, the SAC component governing kinetochore-microtubule (K-MT) attachment stability. Immunofluorescence analysis demonstrated that Mlh3 depletion caused BubR1 mislocalization, increased kinetochore-microtubule distances and failure to establish stable K-MT connections. These defects led to impaired chromosome alignment at the metaphase plate, indicating compromised SAC function (Fig.  4 E). Fig. 4 Mlh3 regulates meiotic spindle integrity through BubR1-mediated kinetochore-microtubule attachments in mouse oocytes ( A ) Acute Mlh3 depletion via Trim-Away technology induced severe spindle defects. B  Gene ontology analysis of Mlh3-interacting proteins revealed significant enrichment in regulation of DNA metabolic process, chromatin remodeling and microtubule cytoskeleton organization. CDK1 emerged as a top-ranked interactor across all functional categories. C  Co-localization of Mlh3 and CDK1 during meiosis II and fluorescence intensities of Mlh3 and CDK1 were measured. The distance is shown in pixel. D  Co-IP was performed to prove the interaction between Mlh3 and CDK1. IP eluates were used for immunoblot with anti-MLH3 and anti-CDK1, respectively. E  Mlh3 depletion disrupted BubR1 localization, increasing kinetochore-microtubule distances; Failed kinetochore-microtubule attachments prevented proper chromosome alignment at the metaphase plate. * p  < 0.05, ** p  < 0.01, *** p  < 0.001. bar = 20 μm Mlh3 regulates meiotic spindle integrity through BubR1-mediated kinetochore-microtubule attachments in mouse oocytes ( A ) Acute Mlh3 depletion via Trim-Away technology induced severe spindle defects. B  Gene ontology analysis of Mlh3-interacting proteins revealed significant enrichment in regulation of DNA metabolic process, chromatin remodeling and microtubule cytoskeleton organization. CDK1 emerged as a top-ranked interactor across all functional categories. C  Co-localization of Mlh3 and CDK1 during meiosis II and fluorescence intensities of Mlh3 and CDK1 were measured. The distance is shown in pixel. D  Co-IP was performed to prove the interaction between Mlh3 and CDK1. IP eluates were used for immunoblot with anti-MLH3 and anti-CDK1, respectively. E  Mlh3 depletion disrupted BubR1 localization, increasing kinetochore-microtubule distances; Failed kinetochore-microtubule attachments prevented proper chromosome alignment at the metaphase plate. * p  < 0.05, ** p  < 0.01, *** p  < 0.001. bar = 20 μm While the MLH1–MLH3 heterodimer is a well-established effector of meiotic crossover formation and mismatch repair (MMR) during gametogenesis, its involvement in somatic DNA damage repair remains unclear. Given our observation that Mlh3 depletion in oocytes compromises spindle assembly, and subsequent embryonic genome integrity, manifested by chromosomal copy number variations and reduced blastocyst quality, we sought to determine whether MLH3 also contributes to DNA damage repair in somatic contexts, which could underlie the developmental defects observed in post-fertilization embryos. To address this, we used human embryonic stem cell line (hESCs) as a model of somatic DNA damage response and treated them with 4-hydroperoxy cyclophosphamide (4-HC), a DNA alkylating agent that induces double-strand breaks (DSBs). Quantitative RT-PCR revealed transient upregulation of MLH3 mRNA following 6-hour treatment with 5–10 µM 4-HC (5–6 fold increase, p  < 0.01), which subsequently declined with prolonged exposure (Fig.  5 A). Notably, this pattern mirrored but did not strictly correlate with MLH1 expression (Fig.  5 A). At the protein level, immunoblotting demonstrated no significant MLH3 induction post-4-HC treatment (10 µM), but rather a progressive decline (0.4-fold at 12 h, p  < 0.05; Fig.  5 B). This attenuation persisted even after 4-hour recovery periods following drug withdrawal (Fig.  5 C). Immunofluorescence further confirmed sustained MLH3 loss following DNA damage (Fig.  5 D). These results indicate that, unlike its critical meiotic role, MLH3 is not actively engaged in the somatic DNA damage response in pluripotent cells. Its lack of induction and progressive loss after DNA damage suggests that the genomic instability and developmental defects in Mlh3-depleted embryos originate primarily from oocyte-intrinsic meiotic errors, rather than impaired post-zygotic DNA repair. Thus, MLH3 functions predominantly in the germline to ensure embryonic genome integrity, with minimal involvement in somatic repair during early development. Fig. 5 MLH3 expression in ESC in response to DNA damage stress. ( A ) mRNA levels after 4-HC treatment. ( B - C ) Protein levels after 4-HC treatment and recovery. ( D ) Immunofluorescence analysis of MLH3 expression. * p  < 0.05, ** p  < 0.01, *** p  < 0.001. bar = 20 μm MLH3 expression in ESC in response to DNA damage stress. ( A ) mRNA levels after 4-HC treatment. ( B - C ) Protein levels after 4-HC treatment and recovery. ( D ) Immunofluorescence analysis of MLH3 expression. * p  < 0.05, ** p  < 0.01, *** p  < 0.001. bar = 20 μm While our previous findings have firmly established MLH3 as a pivotal regulator of meiotic fidelity in oocytes, highlighting its role in pregnancy loss, it is increasingly evident that recurrent miscarriage involves a complex interplay of embryonic and maternal factors, with endometrial receptivity playing a crucial role. However, whether MLH3 also exerts an influence on the functionality of endometrial stromal cells (ESCs), particularly in decidualization, remains unexplored. In this study, we endeavored to elucidate the impact of MLH3 on human ESCs using an in vitro decidualization model. Our results demonstrate that MLH3 is dynamically upregulated during in vitro decidualization of human ESCs, with qPCR showing a progressive increase from day 0 to day 4 (Fig.  6 ). Efficient knockdown of MLH3 was confirmed at both mRNA (Fig.  6 A) and protein levels (Fig.  6 B-C). Functionally, MLH3 deficiency led to enhanced ESC proliferation (Fig.  6 F-G) and reduced apoptosis (Fig.  6 I). Upon induction of decidualization, control ESCs (si-NC) underwent characteristic morphological changes, transitioning from spindle-shaped to rounded cells (Fig.  6 D), and exhibited robust upregulation of decidual markers IGFBP1 and PRL (Fig.  6 E). In contrast, MLH3-knockdown cells showed markedly blunted induction of both markers (Fig.  6 E), indicating impaired decidual response. Together, these findings reveal that MLH3 is not only induced during decidualization but also necessary for an effective decidual response in ESCs, implicating its dysfunction in impaired endometrial preparation for pregnancy. Fig. 6 MLH3 deficiency compromises ESC functionality and decidualization ( A - C ) mRNA and protein levels after siRNA treatment. ( D ) Cellular morphology before and after decidualization in siRNA-treated cells. ( E ) PRL and IGFBP1 mRNA levels in siRNA-treated cells before and after decidualization induction. ESC proliferation (Fig. 6F–G) and apoptosis (Fig. 6I) following siRNA treatment. * p  < 0.05, ** p  < 0.01, *** p  < 0.001. bar = 200 μm MLH3 deficiency compromises ESC functionality and decidualization ( A - C ) mRNA and protein levels after siRNA treatment. ( D ) Cellular morphology before and after decidualization in siRNA-treated cells. ( E ) PRL and IGFBP1 mRNA levels in siRNA-treated cells before and after decidualization induction. ESC proliferation (Fig. 6F–G) and apoptosis (Fig. 6I) following siRNA treatment. * p  < 0.05, ** p  < 0.01, *** p  < 0.001. bar = 200 μm

Patients with URPL were recruited from Sun Yat-Sen Memorial Hospital between December 2019 and August 2025. URPL was defined as the occurrence of ≥ 3 consecutive pregnancy losses before the 28th gestational week. In accordance with the hospital’s standardized protocol for women with URPL, all enrolled patients underwent a comprehensive systematic assessment for potential etiologies of URPL at their initial visit. Participants were excluded if they met any of the following criteria: Documented parental or embryonic chromosomal abnormalities, uterine anomalies, endocrine disorders, or antiphospholipid antibody syndrome associated with prior pregnancy losses; 2) Suffering from autoimmune diseases, particularly rheumatic immune disorders; 3) Being Rh-negative; 4) Having partners with abnormal semen parameters. The clinical characteristics, as well as levels of immune and biochemical markers, were documented. Whole-exome sequencing (WES) was performed on peripheral blood samples from 94 women meeting either of the following criteria: (1) ≥ 3 spontaneous pregnancy losses or (2) ≥ 2 consecutive losses with the same partner. WES analysis was extended to include cases where chorionic villus sampling material was obtainable. This study was approved by the medical ethics committee of Sun Yat-sen Memorial Hospital, Sun Yat-sen University (approval number: SYSEC-KY-2015-05). Written informed consent was obtained from all participants. Documented parental or embryonic chromosomal abnormalities, uterine anomalies, endocrine disorders, or antiphospholipid antibody syndrome associated with prior pregnancy losses; 2) Suffering from autoimmune diseases, particularly rheumatic immune disorders; 3) Being Rh-negative; 4) Having partners with abnormal semen parameters. The clinical characteristics, as well as levels of immune and biochemical markers, were documented. Whole-exome sequencing (WES) was performed on peripheral blood samples from 94 women meeting either of the following criteria: (1) ≥ 3 spontaneous pregnancy losses or (2) ≥ 2 consecutive losses with the same partner. WES analysis was extended to include cases where chorionic villus sampling material was obtainable. This study was approved by the medical ethics committee of Sun Yat-sen Memorial Hospital, Sun Yat-sen University (approval number: SYSEC-KY-2015-05). Written informed consent was obtained from all participants. Female BALB/c mice (3–4 weeks old, n  = 45) were purchased from the Laboratory Animal Center of Sun Yat-sen University. The mice were housed 5 per cage under a 12-hour light/dark cycle. All experimental procedures were conducted in accordance with the guide for the Care and Use of Laboratory Animals published by the National Research Council, and were approved by the Institutional Animal Care and Use Committee (IACUC) of Sun Yat-sen University. (SYSU-IACUC-2023-001880). 3-week old female mice were superovulated by injecting 5 units of pregnant mares serum gonadotropin (PMSG, Ningbo Second hormone Factory, China) followed by 5 units of human chorionic gonadotropin (hCG, Ningbo Second hormone Factory) 48 h after PMSG priming. Mice were sacrificed by cervical dislocation 0, 3, or 12 h post-hCG injection. To collect GV and GVBD oocytes, cumulus-oocyte complexes (COCs) were retrieved by manual rupturing of antral ovarian follicles, and cumulus cells were removed by repeatedly pipetting. To collect MII oocytes, COCs were isolated from oviduct ampullae, and cumulus masses were removed in medium containing 0.3mg /ml hyaluronidase at 37℃. For in vitro maturation, fully grown GV oocytes were cultured in M16 medium under mineral oil at 37℃ in a 5% CO2 incubator. For Trim-Away experiments, purified TRIM21 mRNA (synthesized by Suzhou Hongxun Biotechnology) was complexed with anti-Mlh3 antibody (Abcam ab4834) at final concentrations of 1 mg/ml and 1.25 mg/ml, respectively. The mRNA-antibody complexes were stored at −80 °C and used within one month. GV-stage oocytes were microinjected with 5 pL of the complex using a FemtoJet system, then cultured in M16 medium under mineral oil at 37 °C with 5% CO2 for subsequent developmental analysis or immediate experimentation. IVF was performed following standard procedures [ 26 ]. Cumulus-oocyte complexes were collected from the ampullae 14 h after the hCG injection. Spermatozoa from 6-week-old male mouse epididymides were capacitated for 1 h in human tubal fluid (HTF) medium (Sage, Bedminster, NJ, USA) at 37 °C and 5% CO2. The gametes were co-incubated in HTF medium for 4 h. Fertilized oocytes were washed several times in potassium simplex optimization medium containing amino acids (KSOM + AA, Millipore, Billerica, MA, USA) and transferred to 60 µl drops of KSOM + AA medium covered with paraffin oil. The embryos were cultured at 37 °C with 5% CO2. MII oocytes and embryos at each stage of preimplantation development were collected at defined time points after hCG administration: 14 h (MII oocyte), 38–42 h (2-cell), 58–62 h (4-cell), 78–80 h (8-cell), 98–100 h (early blastocyst), and 114–116 h (late blastocyst). Oestrus-induced female ICR mice in the 8–12-week age rang were paired overnight with vasectomized male ICR mice (10–24 weeks old) for mating purposes. The occurrence of successful mating was confirmed the next morning by detecting the presence of a vaginal plug, marked as 0.5 days post coitum (dpc). Three days following mating (2.5dpc), embryos that had been pre-warmed to an appropriate temperature were precisely transferred into the uterine horns of the recipient female ICR mice via the pasteur pipet (Brand, 747715). After a gestation period of roughly 13.5 days, the mice were humanely sacrificed, and the total number of gestational sacs within the uterus was quantified to assess transplantation outcomes. For spindle and blastocysts immunofluorescence analysis, oocytes or blastocysts were fixed in 4% paraformaldehyde (PFA) in PBS for 1 h at room temperature, permeabilized with 0.1% Triton X-100 for 15 min, and blocked with 10% normal goat serum. Primary antibody incubations were performed overnight at 4 °C using anti-Mlh3 (1:500), anti-CDK1 (Abcam ab18; 1:500) and anti-BubR1 (Abcam ab8982; 1:400) for oocytes staining while using DAPI for blastocyst staining. Microtubules were labeled with CoraLite488-conjugated α-tubulin antibody (Proteintech) for 1 h at 37 °C. Nuclei were counterstained with DAPI (1 µg/mL). Fluorescence images were acquired using a Zeiss LSM 800 confocal microscope equipped with Airyscan super-resolution detection. For immunostained blastocysts, Z-stacks were acquired with a step size of 1 μm using a Leica SP8 STED confocal microscope. To identify proteins that physically interact with Mlh3 in mouse oocytes, we performed co-immunoprecipitation followed by liquid chromatography–tandem mass spectrometry (LC-MS/MS). Protein extracts from approximately 500 germinal vesicle (GV)-stage mouse oocytes were prepared in lysis buffer supplemented with protease inhibitors. Mlh3-interacting proteins were enriched using the Pierce Crosslink Magnetic Bead Co-IP Kit (Thermo Scientific, #88805) with a specific anti-Mlh3 antibody covalently coupled to beads; normal IgG was used as a negative control. After stringent washing, bound proteins were eluted, digested with trypsin, and analyzed by LC-MS/MS. Raw data were processed with MaxQuant against the Mus musculus UniProt database, and high-confidence interactors were defined as proteins detected in the Mlh3 pull-down but not in the IgG control. KGN cells grown to ~ 90% confluence in a 90-mm dish were lysed on ice in 600 µL of IP lysis buffer (20 mM Tris-HCl pH 7.5, 10 mM EDTA, 1 mM EGTA, 150 mM NaCl, 0.05% Triton X-100, 0.05% NP-40) freshly supplemented with 1 mM PMSF, protease inhibitor cocktail (Sigma-Aldrich, P8340, 1:100), and phosphatase inhibitors. Protein A/G agarose beads (Santa Cruz Biotechnology) were pre-washed and then incubated with either anti-MLH3 or anti-CDK2 antibody (as bait) at 4 °C for 4 h; normal mouse IgG was used as a negative control. Antibody-bound beads were collected by brief centrifugation, washed, and then incubated with the clarified KGN cell lysate overnight at 4 °C with gentle rotation. Following three stringent washes with cold IP buffer, immunoprecipitated complexes were eluted by boiling the beads in 2× SDS-PAGE loading buffer for 5 min. The resulting samples were subjected to Western blotting to detect reciprocal interactions between MLH3 and CDK1. The H9 human embryonic stem cell line (hESCs) was utilized to investigate Mlh3 function in embryonic stem cells. hESCs were cultured on mitomycin C–treated mouse embryonic fibroblasts (MEFs) in knockout serum replacement (KSR) medium containing 80% DMEM/F12 (DF12, Gibco), 20% KSR (Gibco), 2 mM L-Glutamine (Sigma), 0.1 mM nonessential aa (NEAA, Gibco), 0.1 mM β-mercaptoethanol (Sigma, M3148-100ML), and 5 ng/mL bFGF (Millipore), with routine passaging every 3–4 days. For DNA damage and repair studies, 4-hydroperoxy cyclophosphamide (4-HC, Sigma) was employed to establish the experimental model. TRIzol reagent (Takara, Japan) was used to extract total RNA of GCs according to the manufacturer’s protocol. The RNA was then transformed into cDNA using the PrimeScript RT Master Mix System (Takara). SYBR Premix Ex TaqII (Takara) was used in quantitative real-time polymerase chain reaction. Primer sequences are provided in Supplementary Table 2 . The relative RNA expression was quantified by normalizing the cycle threshold (Ct) values of target genes, comparing it to GAPDH and then calculated by the 2-△△CT method; △Ct = Avg. Ct sample- Avg. Ct GAPDH; △△Ct = Avg.△Ct Endometriosis -Avg.△Ct controls). Protein expressions were analyzed by western blot. Cells were lysed with ice-cold radioimmunoprecipitation assay lysis buffer (CWBio, China). Protein concentration was then measured using BCA protein assay kit (Beyotime, China). Protein samples (20 mg) were resolved on 10% SDS-PAGE gels (Beyotime) following the manufacturer’s protocol. Membranes were blocked with 5% bovine serum albumin (BSA; 2 g in 40 mL TBST, BD Biosciences) for 1 h at room temperature, then incubated with primary antibodies overnight at 4 °C. GAPDH served as the loading control for normalization. Protein band intensities were quantified using ImageJ software. Small interfering RNAs (siRNAs) targeting MLH3 were designed and synthesized by RiboBio Co., Ltd. (Guangzhou, China). Three distinct siRNA sequences were initially evaluated for knockdown efficiency: siMLH3 #1 (5′-GAAGCACATTACAACAAGA-3′), siMLH3 #2 (5′-GCTAGAGAATGAACCTACA-3′), and siMLH3 #3 (5′-GCAGCGGCATTTCGATGTA-3′). Human endometrial stromal cells (HESCs) were transfected at approximately 70% confluence with 100 nM of each siRNA or a non-targeting negative control (NC) using Lipofectamine RNAiMAX Reagent (Invitrogen, Carlsbad, CA, USA) following the manufacturer’s instructions. Cells were harvested 24 h post-transfection, and knockdown efficiency was assessed by reverse transcription quantitative PCR (RT-qPCR). Among the three candidates, siMLH3 #1 demonstrated the most potent suppression of MLH3 mRNA expression and was therefore selected for all subsequent experiments. To assess the effect of acute Mlh3 depletion on embryonic genome integrity, we collected two groups of mouse 2-cell-stage embryos: untreated controls ( n  = 30) and Mlh3 Trim-Away–treated embryos ( n  = 30). Whole-genome amplification (WGA) was performed on embryos using a modified Multiple Annealing and Looping Based Amplification Cycles (MALBAC) protocol. Each group of embryos was lysed in 5 µL buffer containing 15 mM DTT, 10 mM Tris-EDTA, 1 mg/mL Proteinase K, 0.30% Triton X-100, 200 mM KCl, and 0.25 µM GAT3G primer, followed by incubation at 50 °C for 3 h and protease inactivation at 70 °C for 30 min. For the MALBAC pre-amplification step, 30 µL of Amplification Mix I (1×ThermoPol Reaction Buffer, 0.1 mM dNTPs, 133 µM Mg²⁺, 0.33 µM Primer 1, and 0.33 µM Primer 2; the common 27-nt MALBAC sequence: 5′-GTGAGTGATGGTTGAGGTAGTGTGGAG-3′) was added directly to each lysate. The reaction underwent an initial denaturation at 95 °C for 3.5 min, followed by 11 quasi-linear amplification cycles with a multi-step annealing ramp (4 °C → 10 °C → 20 °C → 30 °C → 40 °C → 50 °C, each for 45–50 s), extension at 65 °C for 4 min, a brief denaturation at 95 °C for 20 s, annealing at 58 °C for 20 s, and hold at 4 °C. Reactions were immediately placed on ice to halt amplification. Exponential amplification was then carried out by adding 30 µL of a second mix (1×ThermoPol buffer, 0.1 mM dNTPs, 100 µM Mg²⁺, 0.67 µM MALBAC primer, 0.067 U/µL DeepVentR exo– polymerase) and running 17 cycles of 95 °C (20 s), 58 °C (30 s), and 72 °C (3 min). Amplified DNA was purified using Zymo-Spin columns and eluted in 50 µL buffer, with concentration and quality assessed using the Qubit dsDNA HS Assay Kit. The WGA products were subsequently used to construct whole-genome sequencing libraries following standard Illumina protocols and sequenced on an Illumina Nova Xplus platform (paired-end 150 bp). Whole-genome sequencing data from control and Mlh3 Trim-Away–treated mouse embryos were analyzed using a standardized somatic copy number alteration (SCNA) detection pipeline. Paired-end reads were first subjected to quality control and adapter trimming using fastp tool, and the resulting high-quality clean reads were then aligned to the mouse reference genome (GRCm38/mm10) using BWA-MEM (v0.7.17). SAM files were converted to BAM format, sorted, and PCR duplicates were marked and removed using GATK’s MarkDuplicates tool. To infer genomic copy number profiles, we employed the Sequenza package (v3.0.0). First, a GC-content wiggle file was generated from the reference genome. Mlh3-depleted and control embryos paired BAM files were then processed using sequenza-utils bam2seqz to compute depth ratios and B-allele frequencies, followed by binning at 500-bp windows to improve signal-to-noise ratio. The resulting seqz file was loaded into R, where chromosomal regions corresponding to non-standard chromosomes (e.g., MT, unplaced scaffolds) were excluded, and only autosomes were retained for analysis. Human endometrial stromal cells (HESCs) were cultured in DMEM/F-12 supplemented with 2% Australian FBS and 1% penicillin–streptomycin at 37 °C under 5% CO₂. Decidualization was induced by treating cells with medroxyprogesterone acetate (MPA, 1 × 10⁻⁶ M) and 8-Br-cAMP (5 × 10⁻⁴ M) for 4 days. Morphological changes were documented by phase-contrast microscopy on days 0, 2, and 4. Decidualization status was evaluated by quantifying mRNA expression of canonical decidual markers IGFBP1 and PRL via quantitative real-time PCR (qRT-PCR) on day 4. To determine whether MLH3 deficiency affects cellular proliferation and survival during decidualization, we performed EdU incorporation assays to label newly synthesized DNA in proliferating cells, followed by fluorescence microscopy and quantitative analysis. Apoptosis was assessed by flow cytometry using Annexin V-FITC/propidium iodide (PI) double staining after 96 h of decidual stimulation. The data were presented as the mean ± S.D. Statistical analysis was based on the data from at least three biologically independent replicates. Stainings and western blots were repeated at least three times. Statistical analyses were performed using one-way analysis of variance (ANOVA) and an unpaired T-test with GraphPad Prism 7.0 software (San Diego, CA). Qualitative data like IVF and blastocyst rates were compared with a Chi-square test or Fisher’s exact test. Differences were considered significant when p  < 0.05. The data are expressed as the means ± SEM, unless otherwise specified. P values are designated as * p  < 0.05, ** p  < 0.01 and *** p  < 0.001.

In this study, we identified compound heterozygous mutations in the MLH3 gene in a patient with URPL. Subsequent whole-exome sequencing of 94 additional URPL patients revealed mutations at distinct MLH3 loci in five individuals. The 5% incidence of MLH3 mutations in our cohort, substantially higher than that found in the general population, strongly implicates the MLH3 gene involved in the pathogenesis of recurrent pregnancy loss. Integrating clinical genomics with functional validation across multiple biological contexts, we demonstrate that MLH3 deficiency disrupts reproductive success through dual, non-redundant pathways: (1) a germline-intrinsic defect in oocyte meiosis and (2) a maternal somatic impairment in endometrial decidualization. In oocytes, acute Mlh3 depletion disrupts spindle assembly, causes mislocalization of the spindle assembly checkpoint (SAC) protein BubR1, a core component of the mitotic checkpoint complex essential for monitoring kinetochore–microtubule attachments and preventing premature anaphase onset [ 27 – 33 ] and impairs stable kinetochore–microtubule connections, leading to defective chromosome alignment, reduced maturation, fertilization, and blastocyst formation. Mlh3-deficient embryos exhibit chromosomal copy number variations, diminished cell numbers, and a 63.89% mid-gestation loss rate upon transfer, phenotypes that closely mirror clinical URPL. Mechanistically, MLH3 physically interacts with CDK1, revealing a non-canonical role in meiotic spindle regulation. Importantly, MLH3 is dispensable for the DNA damage response in embryonic stem cells, indicating that embryonic genomic instability arises from oocyte-intrinsic meiotic errors rather than post-zygotic repair defects. Strikingly, MLH3 is dynamically upregulated during decidualization, and its knockdown impairs morphological transformation and blunts induction of key decidual markers, despite enhancing proliferation and suppressing apoptosis. Together, these findings establish maternal MLH3 as a bifunctional regulator, critical for both oocyte quality and uterine receptivity, thereby providing a unified mechanistic framework for URPL. Through whole-exome sequencing of a well-phenotyped URPL cohort, we identified rare MLH3 variants in five patients, providing novel insights into the genetic basis of this condition. Prior work by Singh et al. used genome editing to model nine human MLH1 and MLH3 variants in mice [ 34 ]. While seven hypomorphic Mlh1 alleles led to female subfertility, reduced litter sizes, and increased embryo resorption, likely due to defective meiotic crossovers and elevated univalent formation, the study did not evaluate the reproductive consequences of Mlh3 variants derived from human populations, leaving the in vivo relevance of MLH3 dysfunction in fertility unclear. Notably, Zhang et al. recently implicated MLH3 in oocyte maturation failure and early embryonic arrest through exome sequencing of 2,140 affected individuals [ 35 ]. Our study extends this association to URPL, demonstrating that MLH3 deficiency can also underlie post-implantation embryonic demise. Together, these findings converge on a unified model: MLH3 is essential for human oocyte competence and early embryogenesis, and its deficiency may manifest clinically across a spectrum of reproductive failures, from blastulation arrest to recurrent miscarriage. Consistent with this, embryos derived from Mlh3-depleted oocytes exhibited severely impaired blastocyst formation and reduced cellularity, indicating compromised developmental competence. Crucially, it remains unclear whether the associated pregnancy loss stems primarily from defective meiotic crossover formation, a specialized function of the mismatch repair pathway, or from non-canonical roles independent of DNA repair. Our discovery of a direct physical interaction between MLH3 and CDK1, a master regulator of meiotic spindle dynamics, provides a mechanistic basis for the observed convergence of impaired oocyte maturation, blastulation failure, disrupted spindle morphology, and aberrant BubR1 localization. While we have not yet mapped the specific domain(s) of MLH3 responsible for this interaction, the co-localization of MLH3 and CDK1 on the meiotic spindle (Fig.  4 C) and their co-immunoprecipitation from native MII complexes strongly support a functional association within the spindle regulatory network. Together, these findings not only establish MLH3 as a novel genetic contributor to URPL but also suggest a non-canonical role in ensuring spindle assembly checkpoint fidelity, potentially through coordination of kinetochore-microtubule dynamics. These insights open new avenues for genetic diagnosis and counseling in idiopathic pregnancy loss. While our data point to a non-canonical role of MLH3 in spindle regulation, it is important to consider its well-established function in meiotic recombination, which may also contribute to the observed genomic instability and pregnancy loss. MutL proteins, initially discovered in bacteria, mediate DNA mismatch repair and genetic recombination [ 36 , 37 ]. In mammals, the Mlh1-Mlh3 heterodimer (MutLγ) is essential for meiotic recombination [ 17 ]. During prophase I, double-strand breaks (DSBs) initiate homologous recombination, leading to non-crossover and CO formation. In mice, ~ 10% of DSBs are resolved as COs via a MutLγ-dependent pathway. The Mlh3 endonuclease domain is critical for CO resolution in yeast; Mlh3 −/− mice exhibit meiotic arrest, chromosome missegregation, and infertility. Mouse models harboring catalytically inactive Mlh3 (e.g., Mlh3 DN/DN ) demonstrate persistent accumulation of DSB repair factors and CO pathway mediators, indicating delayed repair kinetics and defective CO formation [ 38 , 39 ]. Similarly, Mlh1 point mutations in mice have been shown to cause age-dependent fertility declines and increased embryo resorption, which are attributed to reduced crossovers and elevated rates of aneuploidy. This suggests that hypomorphic alleles of meiotic recombination genes can render females susceptible to gamete aneuploidy, leading to an increased incidence of miscarriage [ 34 , 40 ]. To assess the genomic impact of Mlh3 deficiency, we performed low-pass whole-genome sequencing on 2-cell embryos derived from oocytes with Mlh3 knockdown at the GV stage. CNV analysis revealed focal imbalances, particularly on chromosomes 6 and 9. Their presence after only one post-fertilization division strongly suggests a meiotic origin rather than de novo mitotic errors. This aligns with the subfertility observed in both mouse models and our URPL patients, none of whom showed complete infertility, and with the fact that not all miscarriage samples exhibited overt karyotype abnormalities. Together, these findings support a dual-pathway model: Mlh3 dysfunction can cause either severe meiotic error leading to pre-implantation arrest, or subtler genomic lesions (e.g., focal CNVs) in euploid-appearing embryos that nonetheless impair post-implantation development and result in pregnancy loss. In our study, we found that MLH3 exerts a positive regulatory role in decidualization. MLH3 deficiency led to enhanced proliferation, reduced apoptosis, and impaired decidual response, phenotypes consistent with a failure to exit the cell cycle and initiate terminal differentiation, as further supported by downregulation of decidual markers. Several studies [ 41 – 44 ] have also identified MLH3 mutations in patients with endometrial carcinoma, and proposed that the mechanisms by which MLH3 gene mutations contribute to endometrial carcinogenesis may include the following: impaired MMR function, which fails to correct DNA replication errors, thereby increasing the gene mutation rate and facilitating tumorigenesis; in addition, MLH3 defects may render cells incapable of triggering apoptosis following DNA damage, leading to the survival of abnormal cells and the accumulation of oncogenic mutations. However, most studies focusing on MLH3 have concentrated on tumors; our findings firstly reveal that MLH3 is not only responsive to decidual cues but also essential for proper ESC differentiation, linking its dysfunction to compromised endometrial receptivity. These findings not only highlight the need for improved cancer screening but also emphasize the importance of evaluating fecundity. Limitations also should be acknowledged. The original genetic variants in clinical cases are point mutations with undetermined functions, whereas this study adopted total protein degradation of Mlh3 protein in mouse GV-stage oocytes using the Trim-away technique. This discrepancy means the current findings cannot fully explain the recurrent miscarriage phenotype observed in clinical settings. Future work should include in vitro functional assays to characterize patient-derived MLH3 variants, such as assessing protein stability, CDK1 interaction, and subcellular localization, followed by rescue experiments via mRNA or protein injection in oocytes, and ultimately in vivo modeling using patient allele knock-in mice. Together, these approaches will enable systematic evaluation of ovarian reserve, ovulation efficiency, fertilization competence, and uterine function, thereby clarifying how specific MLH3 alleles contribute to recurrent pregnancy loss.

This study establishes maternal MLH3 deficiency as a novel genetic etiology of URPL, mechanistically linking loss of MLH3 function to defective spindle assembly checkpoint function, compromised oocyte quality, embryonic developmental potential and defective endometrial decidualization. These findings expand the role of DNA mismatch repair genes beyond canonical repair functions to encompass critical roles in both gamete competence and uterine receptivity, offering new avenues for genetic diagnosis, risk stratification, and personalized management of idiopathic pregnancy loss.

Recurrent pregnancy loss (RPL) represents a prevalent obstetric complication, affecting approximately 1%−5% of women within reproductive age, and is clinically characterized by experiencing two or more pregnancy losses before 28 weeks of gestation (including biochemical pregnancies) [ 1 – 3 ]. In addition to identified genetic, anatomical, endocrine, infectious, environmental, and immune factors, about 50% of patients have unknown causes, termed unexplained recurrent spontaneous abortion (URPL) [ 4 – 9 ]. Exploring the pathogenesis of URPL is of great significance for preventing and treating pregnancy-related diseases and improving reproductive health. Recently, emerging evidence has underscored the association between maternal genetic variants that impact oogenesis, ovarian function, and endometrial receptivity, and the occurrence of unexplained recurrent pregnancy loss. Among the maternal genetic factors that have been implicated in URPL are single-base mutations, genomic structural variations, chromosomal aberrations, and aberrant DNA methylation patterns [ 10 , 11 ]. However, few pathogenic genes have been consistently validated across studies as directly influencing pregnancy maintenance mechanisms. This inconsistency likely stems from: (1) substantial genetic heterogeneity inherent to RPL; (2) variable clinical definitions and diagnostic criteria for RPL across studies, resulting in heterogeneous patient cohorts; (3) small sample sizes in underpowered studies; and (4) a predominance of candidate-gene approaches rather than comprehensive genomic investigations [ 12 , 13 ]. Therefore, systematic identification and functional validation of pathogenic genetic variants through well-designed genomic studies remain critically important. Chromosome mis-segregation during meiosis can take place in either Meiosis I (MI) or Meiosis II (MII) [ 14 , 15 ]. Notably, a deficiency in recombination on extra or missing chromosomes has been linked to inefficient crossover maturation prior to MI. When such gametes undergo fertilization, they may give rise to aneuploid embryos, which are nearly incompatible with viability and consequently result in pregnancy loss [ 16 – 18 ]. In mice, crossover (CO) recombination requires a critical subset of genes, including the mismatch repair (MMR) proteins Mlh1 and Mlh3 . These form the heterodimeric MutLγ endonuclease complex essential for resolving double Holliday junction recombination intermediates[ 19 , 20 ]. Null alleles of Mlh1 or Mlh3 induce meiotic arrest and sterility in mice, as both genes are essential for approximately 90% of all crossovers-specifically, the interference-dependent Class I crossovers [ 14 ]. The absence of COs leads to the formation of univalent chromosomes, which fail to align correctly when attached to microtubules at the metaphase plate [ 21 – 23 ]. This triggers the spindle assembly checkpoint (SAC) and inhibits the progression of anaphase [ 24 ]. However, recombination defects impairing bivalent formation in only one or a few chromosome pairs are insufficient to trigger the SAC [ 22 , 25 ]. This permits oocytes with these defects to complete meiosis, propagating aneuploid embryos [ 23 ]. Although murine models have established essential roles for Mlh1 and Mlh3 in meiotic recombination and genomic integrity during gametogenesis, clinical evidence linking these genes to human embryonic failure or recurrent pregnancy loss remains limited. Moreover, their potential involvement in uterine biology, particularly endometrial function, decidualization, and receptivity, has scarcely been explored, leaving open the possibility that MLH1/3 defects may compromise fertility not only through oocyte/embryo quality but also via impaired endometrial responses. In this study, whole exome sequencing of 94 patients with URPL identified five rare MLH3 variants, yielding a carrier frequency of 5.32% (5/94), substantially higher than the population prevalence reported in gnomAD database (0–0.011%), supporting a potential pathogenic role for MLH3 in URPL. Functional studies revealed that Mlh3 depletion in mouse oocytes disrupts spindle assembly, impairs BubR1 localization (a core SAC component), and compromises oocyte maturation, fertilization, and embryonic development, leading to increased pregnancy loss. Moreover, MLH3 deficiency in endometrial stromal cells severely impaired decidualization, accompanied by aberrant proliferation and apoptosis. Together, these findings establish maternal MLH3 as a dual-function gene essential for both oocyte quality and endometrial receptivity, providing a novel mechanistic basis for its role in URPL pathogenesis.

Below is the link to the electronic supplementary material. ESM 1 (DOCX 7.58 MB) (DOCX 7.58 MB)