HK2-driven histone H3K18 lactylation promotes stromal cell senescence and decidualization deficiency in URSA via CUX1-mediated SASP factor transcription

Abstract

Background

Unexplained recurrent spontaneous abortion (URSA) is characterized by defective endometrial stromal cell decidualization, with cellular senescence emerging as a key contributor. However, the metabolic–epigenetic mechanisms linking glycolysis to senescence-driven decidualization failure remain unclear. This study elucidates how hexokinase 2 (HK2)-mediated glycolytic reprogramming promotes histone lactylation-dependent stromal senescence and decidualization impairment in URSA.

Methods

We employed multi-omics profiling (RNA-seq, metabolomics, and CUT&Tag) of primary stromal cells from patients with URSA and controls to map the histone H3K18 lactylation (H3K18la)–cut-like homeobox 1 (CUX1)–senescence-associated secretory phenotype (SASP) axis. Subsequently, this axis was validated both in vitro decidualization models and URSA murine models.

Results

Decidual tissues from patients with URSA exhibited stromal cell senescence and impaired decidualization. Mechanistically, HK2-driven glycolysis elevated lactate production, which in turn promoted H3K18la at the CUX1 promoter. CUX1 then directly activated the transcription of key SASP factors, thereby propagating the senescence state. Critically, CUX1 depletion or glycolysis inhibition rescued these senescence and decidualization deficiency in vitro. Furthermore, CUX1 knockdown in the URSA murine model reduced stromal senescence and improved decidualization.

Conclusions

Our findings define a novel HK2–H3K18la–CUX1–SASP signaling axis that drives URSA pathogenesis by linking metabolic reprogramming with epigenetic regulation. This work highlights CUX1 as a potential therapeutic target for correcting decidualization deficiency in URSA.

Graphical abstract

Supplementary Information

The online version contains supplementary material available at 10.1186/s11658-026-00879-y.

Introduction

Approximately 2.5% of women trying to conceive experience recurrent spontaneous abortion (RSA), a condition characterized by two or more clinically confirmed pregnancy losses occurring prior to the 24th week of gestation [1, 2]. While known causes include chromosomal errors, infection, aberrant uterine anatomy, immune system issues, and endocrine abnormalities, the etiology remains unclear in nearly 50% of RSA cases. These are defined as unexplained recurrent spontaneous abortion (URSA) [2, 3]. Emerging evidence implicates uterine microenvironmental dysregulation, particularly defective decidualization processes, as a key pathogenic mechanism underlying idiopathic recurrent pregnancy loss [4]. Decidualization is the post-ovulatory remodeling of the endometrium, characterized by the coordinated proliferation and differentiation of endometrial stromal cells (ESCs) from a fibroblast-like to an phenotype, forming decidual stromal cells (DSCs) [5, 6]. DSCs can promptly envelop the conceptus, participate in embryo biosensoring, and afterward generate a decidual matrix to facilitate trophoblast invasion [7]. In contrast, decidualization disruption can trigger a cascade of adverse effects that impair pregnancy outcomes [8]. A growing body of experimental and clinical evidence indicates that decidualization deficiency is a key contributor to URSA [9, 10]. However, the molecular mechanisms underlying this pathology remain poorly understood.

Cellular senescence manifests as an irreversible proliferative cessation, induced by genotoxic and metabolic stressors, concurrently exhibiting multidimensional changes in cell morphology, secretory profile, and epigenetic alterations, among others [11, 12]. Aberrant cellular senescence is implicated in tissue dysfunction and various pregnancy complications [13, 14]. Senescence of endometrium stromal cells limit their potential to differentiate into DSCs [14, 15]. Furthermore, senescent cells adopt a senescence-associated secretory phenotype (SASP), releasing a complex array of pro-inflammatory factors that perpetuate pathological microenvironments at the maternal–fetal interface, critically compromising key gestational events such as embryo apposition and placentation [16]. Recent mechanistic studies delineate premature decidual senescence as a pathognomonic feature of URSA cases, particularly manifesting during the embryo-endometrial crosstalk phase that is critical for successful implantation [17, 18]. Thus, stromal cell senescence emerges as a pivotal mechanism driving decidualization deficiency in URSA, yet the specific molecular pathways governing this process remain a critical knowledge gap.

Cell fate is tightly tailored by the interplay between metabolism and epigenetics. Previous studies based on single-cell sequencing and metabolomic profiling have unraveled increased glycolysis in DSCs and high level of lactate in peripheral blood in URSA [19, 20]. Furthermore, lactate has emerged as a key signaling molecule to drive SASP [21]. Emerging research delineates that glycolytic flux-derived lactate serves as a critical mediator of senescence acceleration through histone lactylation [22]. This recently discovered post-translational modification is a specific subtype of the broader lysine acylation family, involving the covalent attachment of a lactate-derived lactyl group to lysine residues on histone tails. Histone lactylation relaxes chromatin structure and promotes gene transcription, thus directly linking cellular metabolic states to epigenetic regulation (Supplementary Fig. 1). More importantly, histone lactylation is recognized as a new participant to confer cell senescence and mediates disease development [23]. However, whether histone lactylation is implicated in the signaling pathways that link glycolysis to cellular senescence and decidualization deficiency in URSA has not been determined. Consequently, exploring the complex interplay between metabolic and epigenetic mechanisms during cellular senescence represents a promising research direction.

Here, we identify stromal cell senescence as a key driver of decidualization deficiency in URSA. We demonstrate that this senescent phenotype is driven by enhanced hexokinase 2 (HK2) mediated glycolysis, which promotes H3K18 lactylation (H3K18la). Notably, the cleavage under targets and tagmentation (CUT&Tag) epigenomic profiling revealed pronounced H3K18 lactylation enrichment at transcriptional start sites of cut-like homeobox 1 (CUX1). As a homeodomain transcription factor, CUX1 acts as a vital switch for specific SASP factor transcription, thereby leading to stromal cell senescence and decidualization deficiency in URSA. Collectively, this study elucidates a novel pathogenic axis (HK2–H3K18la–CUX1) that integrates metabolic reprogramming with epigenetic regulation to drive SASP activation and stromal cell senescence in URSA. This leads to speculation that targeting this pathogenic metabolism–epigenetic signaling cascade may represent a promising therapeutic approach for treating decidualization deficiency in URSA.

Materials and methods

Reagent or resource	Source	Identifier
Antibodies
For CUT&Tag or ChIP
Rabbit anti-H3K18la	PTM BIO	Cat. no. PTM-1427RM
Rabbit anti-CUX1	Abcam	Cat. no. ab307821
For immunofluorescence and Western blot
Rabbit anti-HK2	Abcam	Cat. no. ab209847
Rabbit anti-l-lactyl lysine	PTM BIO	Cat. no. PTM-1401
Rabbit anti-H3K18la	PTM BIO	Cat. no. PTM-1427RM
Rabbit anti-H3K9la	PTM BIO	Cat. no. PTM-1419RM
Rabbit anti-H3K23la	PTM BIO	Cat. no. PTM-1413RM
Rabbit anti-H3K56la	PTM BIO	Cat. no. PTM-1421RM
Rabbit anti-histone H3	Abcam	Cat. no. ab1791
Rabbit anti-p16	Abcam	Cat. no. ab51243
Rabbit anti-p21	Abcam	Cat. no. ab109199
Rabbit anti-CUX1	Abcam	Cat. no. ab307821
Mouse anti-vimentin	Abcam	Cat. no. ab8978
Rabbit anti-β-actin	Cell Signaling Technology	Cat. no. 4970
Rabbit anti-phospho-histone H2A.X (Ser139)	Cell Signaling Technology	Cat. no. 9718S
IgG control antibody	Millipore	Cat. no. 12–370
Goat anti-rabbit IgG antibody	Millipore	Cat. no. AP132
Goat anti-rabbit IgG (H&L)	LI-COR	Cat. no. 926–32211
Goat anti-mouse IgG (H&L)	LI-COR	Cat. no. 926–68070
Goat anti-rabbit IgG H&L (Alexa Fluor® 488)	Abcam	Cat. no. ab150077
Goat anti-rabbit IgG H&L (Cy3®)	Abcam	Cat. no. ab6939
Goat anti-rabbit IgG H&L (Alexa Fluor® 647)	Abcam	Cat. no. ab150083
Goat anti-mouse IgG H&L (Alexa Fluor® 488)	Abcam	Cat. no. ab150113
Chemicals, oligonucleotides, and critical commercial assays
DMEM/F12	Thermo Fisher	Cat. no. 12634010
Fetal bovine serum (FBS)	Gibco	Cat. no. 10099141
Penicillin	Beyotime	Cat. no. ST488-1
Streptomycin	Beyotime	Cat. no. ST488-2
8-Br-cAMP	Sigma	Cat. no. B5386
MPA	APExBIO	Cat. no. B1510
Triton X-100	Solarbio	Cat. no. T8200
PBS	Cienry	Cat. no. CR20012
RIPA	Beyotime	Cat. no. P0013B
Protease inhibitor cocktail	Bimake	Cat. no. B14002
4% polyformaldehyde	Biosharp	Cat. no. BL539A
Sodium l-lactate	Sigma-Aldrich	Cat. no. 71718
HiFiScript cDNA synthesis kit	CWBIO	Cat. no. CW2569M
SYBR Green Pro Taq HS premix	AGBIO	Cat. no. AG11701
TRIzol reagent	Invitrogen	Cat. no. 15596026
Lactic acid (LA) content assay kit	Solarbio	Cat. no. BC2230
Human IL-1β ELISA kit	Youke Life Sciences Technology	Cat. no. EH0006
Human IL-6 ELISA kit	Youke Life Sciences Technology	Cat. no. EH0001
Human TNF-α ELISA kit	Youke Life Sciences Technology	Cat. no. EH0002
Mouse CXCL1 ELISA kit	BYabscience	Cat. no. BY-EM220048
Mouse CXCL3 ELISA kit	BYabscience	Cat. no. BY-EM223880
Mouse TNF-α ELISA kit	BYabscience	Cat. no. BY-EM220852
Senescence β-galactosidase staining kit	Beyotime	Cat. no. C0602
Seahorse XF glycolysis stress test kit	Agilent	Cat. no. 103020–100
Seahorse XF Cell Mito Stress test kit	Agilent	Cat. no. 103015–100
Mitochondrial membrane potential assay kit with JC-1	Beyotime	Cat. no. C2006
CoraLite® 594-Phalloidin (red)	Proteintech	Cat. No. PF00003
Isoflurane	RWD	Cat. no. R510-22
H2O2	Sigma-Aldrich	Cat. no. 88597
d-galactose	Sigma-Aldrich	Cat. no. 3455
CoCl2	Sigma-Aldrich	Cat. no. 232696
A-485	MedChemExpress	Cat. no. HY-107455
Progesterone	Solarbio	Cat. no. P9060
Estradiol	Aladdin	Cat. no. E110145
Sesame oil	Aladdin	Cat. no. S304679
Dasatinib	Sigma-Aldrich	Cat. no. SML2589
Quercetin	Sigma-Aldrich	Cat. no. Q4951
Luria–Bertani medium	Yuanye Bio-Technology	Cat. no. R20125
Trelief® 5α chemically competent cell	TSINGKE	Cat. no. TSC01
E.Z.N.A. Endo-free Plasmid Midi Kit	Omega	Cat. no. D6915-03
riboFECT™ CP transfection reagent	RiboBio	Cat. no. C10511
LipoFiter™ 3.0 liposomal transfection reagent	HANBIO	Cat. no. HB-TRLF3-1000
Negative control plasmid (NC: GL119 pSLenti-CMV-MCS-3xFLAG-PGK-Puro-WPRE)	Obio Technology	N/A
HK2 overexpression plasmid (OE: pSLenti-CMV-HK2-3xFLAG-PGK-Puro-WPRE)	Obio Technology	N/A
CUX1 overexpression plasmid (OE: pSLenti-CMV-CUX1-3xFLAG-PGK-Puro-WPRE)	Obio Technology	N/A
CUX1, HK2, and scramble siRNA	Obio Technology	N/A
rAAV-U6-shRNA(CUX1)-CMV-EGFP-pA and rAAV- U6-shRNA (scramble)-CMV-EGFP-SV40polyA	Obio Technology	N/A
Experimental models: cell lines
Human: telomerase-immortalized human endometrial stromal cells (T-hESC)	ATCC	Cat. no. CRL-4003
Software and microscope
ImageJ (version 1.53v)	National Institutes of Health	https://imagej.nih.gov/ij/download.html
GraphPad Prism version 9.0	GraphPad Software	https://www.graphpad.com/
Confocal laser scanning microscope	Zeiss	LSM880
Fluorescence microscope	OLYMPUS	VS120-S6-W
Inveon MM platform	Siemens Preclinical Solutions, Knoxville	N/A
Odyssey fluorescence imaging system	Odyssey	N/A

Patients and tissue samples

Decidual specimens were prospectively collected from 20 clinically defined patients with URSA and 20 age-matched controls at Hangzhou Hospital of TCM Affiliated to ZCMU. All participants provided written informed consent. To ensure the validity of our comparative analysis and minimize confounding factors, we established strict inclusion and exclusion criteria for both the URSA and control cohorts. The detailed criteria were as follows:

For the URSA group

Inclusion criteria: Women with a history of two or more consecutive, unexplained pregnancy losses were recruited for this study. To specifically investigate the mechanisms of early decidualization failure, inclusion for tissue donation was limited to a current pregnancy loss occurring prior to 12 weeks of gestation.

Exclusion criteria: Patients were excluded if they had any of the following: parental chromosomal abnormalities; uterine anatomical abnormalities (e.g., septate uterus, Asherman’s syndrome) confirmed by hysteroscopy or sonohysterography; endocrine disorders (e.g., uncontrolled thyroid dysfunction, diabetes mellitus, polycystic ovary syndrome); diagnosed autoimmune diseases (e.g., antiphospholipid syndrome, systemic lupus erythematosus); a history of using hormonal contraceptives or an intrauterine device within the 3 months preceding the index pregnancy; cervical incompetence; acute or chronic infections; or a history of smoking, alcohol consumption, or use of illicit drugs.

For the control group

Inclusion criteria: Women with at least one prior uncomplicated, full-term delivery and no history of pregnancy loss or other obstetric complications (e.g., preeclampsia, fetal growth restriction), who were undergoing termination of a confirmed, normally developing pregnancy for nonmedical reasons (e.g., social reasons).

Exclusion criteria: The same exclusion criteria applied to the URSA group were also applied to the control group to ensure a healthy baseline.

Human stromal primary cell isolation and culture

Stromal cells from the decidual tissue of control and URSA groups were isolated as previously described [24]. Decidual tissues were minced, washed, and subsequently digested by incubation with 0.5 mg/mL collagenase in HBSS at 37 °C for 30 min. The resulting suspension was filtered through a 70-μm nylon filter and centrifuged at 400g for 10 min. Stromal cells were then purified by centrifugation on a discontinuous gradient of 20%, 40%, and 60% Percoll gradient for 20 min at 800g. The stromal cell fraction was collected from the 20%/40% interface. The isolated cells were then resuspended, washed, and cultured in DMEM medium with 10% FBS.

Cell culture and in vitro decidualization of telomerase-immortalized human endometrial stromal cells (T-hESCs)

T-hESCs were used for this study as they are a well-established in vitro model for decidualization, retaining the key physiological capacity to differentiate in response to hormonal stimuli in a manner comparable to primary endometrial stromal cells [25]. Following established protocols [26, 27], T-hESCs were decidualized in vitro over a 6-day period using 0.5 mM 8-Br-cAMP and 1 μM MPA, with medium renewal every 48 h. Successful decidualization was confirmed by morphological changes and by quantifying the upregulation of decidual marker mRNAs, such as prolactin (PRL) and insulin-like growth factor binding protein 1 (IGFBP1) [28].

Plasmid transfection

T-hESCs were seeded at a density of 2 × 10⁵ cells per well in 24-well plate. For transfection, plasmid DNA (0.8 μg) was diluted in 50 μL of DMEM. In a separate tube, 1.6 μL of LipoFiter™ 3 reagent was mixed with 50 μL of RPMI 1640 and incubated for 5 min. The DNA and LipoFiter™ 3 mixtures were then combined and incubated at room temperature for 20 min. The resulting transfection complexes were added dropwise to the cells. The plate was then centrifuged at 200g for 1.5 h and subsequently incubated under standard conditions (e.g., at 37 °C in a 5% CO₂ incubator).

siRNA transfection

For gene knockdown experiments, T-hESCs undergoing decidualization were transfected with 50 nM of either target-specific siRNA (for HK2 or CUX1) or a nontargeting scramble control. Transfections were performed on day 0 and day 4 of the decidualization protocol using the riboFECT™ CP transfection kit according to the manufacturer’s instructions. All siRNA sequences are listed in Supplementary Table 1.

Cell processing and grouping

Interventions on primary human stromal cells

To investigate the role of histone lactylation in primary cells, decidual stromal cells isolated from patients with URSA were treated with either the p300/CBP inhibitor A-485 (10 nM) [29] or an equivalent volume of DMSO as vehicle control. Cells were incubated for 48 h before being harvested for subsequent analyses.

Interventions on T-hESCs during in vitro decidualization

To dissect the molecular pathways in a controlled system, T-hESCs undergoing in vitro decidualization were subjected to the following interventions:

1. To explore the effect of histone lactylation on cellular senescence and decidualization, T-hESCs were cultured in standard decidualization medium (Control group) or medium supplemented with 25 mM sodium lactate (NaLa) (Con + NaLa group) [22].

2. To identify that HK2 can affect cell senescence and decidualization by regulating H3K18la expression, the cells were transfected with 50 nM HK2 siRNA, HK2 scramble, HK2 NC, and HK2 overexpression (OE) plasmid at induction initiation (day 0) and mid-phase (day 4) of the decidualization protocol. Besides, cells were treated with 25 mM NaLa to induce H3K18 lactylation, with 250 µM CoCl₂ (a hypoxia-mimetic agent that stabilizes hypoxia-inducible factor 1-alpha (HIF-1α) and induces glycolysis) to induce cellular hypoxia [30], and with 20 mM oxamate (a lactate dehydrogenase inhibitor that suppresses glycolytic lactate production) to inhibit glycolysis [31], respectively. Thus, the induced decidualized T-hESCs were divided into NC plasmid group, HK2 OE plasmid group, HK2 OE plasmid + oxamate group; CoCl₂ + scramble siRNA group, CoCl₂ + HK2 siRNA group, CoCl₂ + HK2 siRNA + NaLa group.

3. To identify that H3K18la can affect cell senescence and decidualization by regulating CUX1 expression, the cells were transfected with 50 nM CUX1 siRNA, CUX1 scramble, CUX1 NC, and CUX1 OE plasmid at induction initiation (day 0) and mid-phase (day 4) of the decidualization protocol. Besides, we treated cells with 25 mM NaLa to induce H3K18 lactylation, with 250 µM CoCl₂ to induce cellular hypoxia, with 20 mM oxamate to interfere glycosis, respectively. Thus, the induced decidualized T-hESCs were divided into CUX1 scramble group, NaLa + CUX1 scramble group, NaLa + CUX1 siRNA group; CoCl₂ + CUX1 NC plasmid, CoCl₂ + oxamate + CUX1 NC plasmid, CoCl₂ + oxamate + CUX1 OE plasmid group.

SA-β-gal staining

Cellular senescence was assessed by a senescence β-galactosidase staining kit. Briefly, cells were fixed with 1 mL of fixative solution. After fixation, 1 mL of staining solution was added, and the cells were incubated overnight at 37 ℃. The samples were observed under the microscope (VS120-S6-W, Olympus), and analyzed by using ImageJ. Semiquantitative assessment was carried out by randomly choosing three sections, with five distinct areas within each section being targeted for evaluation.

Morphological evaluation of decidualization

To evaluate morphological decidualization, the cytoskeleton was visualized by staining for F-actin. T-hESCs were fixed, permeabilized, and then incubated with a fluorescently labeled phalloidin working solution for 20 min at room temperature. After washing, coverslips were mounted and images were acquired using a confocal laser scanning microscope (CLSM, LSM880; Zeiss).

Artificial induction of in vivo decidualization

Decidualization was induced as previously mentioned [32, 33]. Ovariectomized mice (6–8 weeks old) were anesthestized and administered exogenous estrogen and progesterone after endogenous hormone regression within 2 weeks; i.e., mice were subcutaneously injected with estradiol (E2) 100 ng/day for 2 days, followed by 1 mg progesterone (P4) and 10 ng E2 for 3 days. After that, the uterine horn was exposed under anesthesia and 100 µL of corn oil was injected into one side to simulate the destruction of endometrial epithelial cells during implantation and to initiate decidualization. The opposite side was not treated for control. Subsequently, P4 s.c. 1 mg/ day and E2 s.c. 10 ng/day were given consecutively for 5 days to simulate the hormonal changes during pregnancy. Finally, the uterus horn tissues were extracted for further experiments 6 h after the last administration.

CUX1 gene interference sequence screening

Subsequently, HEK293T cells underwent transfection with a designated plasmid, and three technical replicates were established for each experimental combination. After transfection, cells were placed in a 37℃, 5% CO₂ incubator for 72 h. We removed the cells, extracted the total RNA of the cells, and then performed real-time PCR to screen plasmids with the most significant interference on target genes. In this study, shRNA2 exhibited the most significant effect on the downregulation of CUX1 compared with other interference sequences (Supplementary Fig. 2).

rAAV-U6-shRNA (CUX1)-CMV-EGFP-pA uterine local injection

We anesthetized mice with isoflurane, and disinfected the abdomen with 75% alcohol. We made a 1-cm longitudinal incision in the lower abdomen. We pulled the uterus out of the incision and injected the adeno-associated virus (AAV) into the uterus at multiple points according to the grouping situation, with 5 × 10¹² vg/mL and 3 μL/point. After injection, we repositioned the uterus and sutured the wound. Considering the onset time and duration of action of AAV, this study conducted local uterine injection of AAV at 3 weeks before the preparation of the URSA model. The decidual tissues were harvested on the 8th day of gestation.

URSA mouse models

To model spontaneous abortion, CBA/J female mice were paired with DBA/2 males, mimicking human URSA, while CBA/J × BALB/c crosses served as controls with lower embryo resorption rates [34]. Eight-week CBA/J female (n = 84), DBA/2 male (n = 30), and BALB/c male mice (n = 12) (Huafukang Biotechnology, China) were kept under standard laboratory conditions. Females were cohabited with males (2:1 ratio) to establish URSA or normal pregnancy groups. Gestational timing was determined by daily vaginal plug checks (day 1 = plug detection).

To investigate the impact of aging inhibition on the process of decidualization in URSA, every 12 pregnant CBA/J × DBA/2 mice received intraperitoneal (i.p.) injections of dasatinib (5 mg/kg) and quercetin (50 mg/kg) or sterile PBS once a day from day 1 of pregnancy. As a result, 36 pregnant mice were divided into three groups on the basis of their mating combinations: Control group (n = 12), URSA group (n = 12), and URSA + dasatinib + quercetin group (n = 12).

To observe the regulation of CUX1 on stromal cell senescence and decidualization in URSA, every 12 CBA/J mice mating with DBA/2 mice were transfected with AAV-CUX1 shRNA or AAV-scramble shRNA. Consequently, 48 pregnant CBA/J mice were allocated into four distinct groups on the basis of their respective mating crosses: Control group (n = 12), URSA group (n = 12), URSA + AAV-shRNA (scramble) group (n = 12), and URSA + AAV-shRNA (CUX1) group (n = 12). All pregnant mice were monitored on day 8 of pregnancy.

Analysis of fetal resorption rate

According to previous studies, the embryo resorption rate (R) can be calculated using the formula R = R_e/(R_e + F), wherein R_e represents the count of resorbed embryos and F denotes the count of viable embryos [35].

qRT-PCR

qRT-PCR was performed using a LightCycler 96 system (Roche) according to standardized protocols. Relative mRNA expression levels were quantified via the 2^−ΔΔCt method with the primer sequences provided in Supplementary Table 2.

ELISA

The concentrations of specific SASP factors (e.g., interleukin (IL)-6, IL-1β, and tumor necrosis factor (TNF)-α) in human stromal cells and mouse stromal cells were detected according to ELISA kit instructions.

Lactate content analysis

Lactate acid content in stromal cells was detected by the lactate acid assay kit, and its absorbance was measured at a wavelength of 570 nm. Next, the lactate calibration curve was plotted and the lactate concentration calculated.

Immunofluorescence

For T-hESCs, the cells were fixed using 4% paraformaldehyde and subsequently blocked with immunostaining blocking buffer for 2 h. Then, they were incubated with CUX1, H3K18la, p16, p21, and γ-H2AX antibodies (all at 1:200 dilution), then incubated with secondary antibody for 2 h. Finally, CLSM was used to collect the image.

For decidua samples, the frozen sections of decidual tissue were first prepared and then blocked with immunol staining blocking buffer for 2 h, stained with primary antibody overnight (rabbit anti-HK2, H3K18la, CUX1, p16, and p21, 1:200 dilution; mouse anti-vimentin, 1:200 dilution), and incubated with secondary antibody for 2 h. Finally, the tissues were examined using a fluorescence microscope (VS120-S6-W, OLYMPUS). ImageJ was employed for the analysis of the cells and tissues. For the purpose of semiquantitative assessment, three sections were chosen at random, and within each section, five distinct fields were selected.

Immunohistochemistry

The decidual tissues paraffin sections (5 μm) were prepared. Following antigen retrieval procedures, the tissue slides were treated with a blocking solution containing 3% hydrogen peroxide and 5% BSA. Subsequently, the decidual tissue sections underwent incubation with the primary antibody at 4 °C overnight, followed by an additional hour of incubation with goat anti-rabbit immunoglobulins. Finally, the decidual tissue section images were captured with an optical microscope.

Western blot

The expression of pan-Kla, H3K9la, H3K18la, H3K23la, H3K56la, H3, p16, and p21 in decidual tissues or cells was detected by Western blot. Decidua tissue or cells lysates were processed, then electrophoresis and membrane transfer were performed (300 mA, 80 min). Then, the membranes were incubated with pan-Kla, H3K9la, H3K18la, H3K23la, H3K56la, H3, p16, and p21 antibodies and the secondary antibodies. Membranes were scanned using the Odyssey fluorescence imaging system, and band intensities were quantified using ImageJ software.

¹⁸F-FDG PET image reconstruction and quantitative evaluation

PET/CT imaging was employed to monitor abortion events in mice. More precisely, the imaging was conducted utilizing an Inveon MM system. A computer-operated bed featuring an 8.5-cm transverse and 5.7-cm axial field of view was utilized for the procedure. The animals were placed under anesthesia induced by 2% isoflurane. A single dose of 150 uci [¹⁸F]-FDG was injected into the tail vein, and the animals were awakened and return to the cage immediately after anesthesia. Thirty minutes following the administration of the tracer, the animals were subjected to anesthesia using isoflurane and positioned in a prone posture on a PET scanning bed, where anesthesia was continued for the duration of the study, with 1.5% isoflurane supplied at a rate of 2 L/min for oxygen. We processed the scan data using Inveon Acquisition Workplace (IAW) version 1.5.0.28. Prior to PET scanning, 10-min CT x-rays were scanned at 80 kV, 500 µA power, and 1100 ms exposure time for attenuation correction. Thereafter, a static PET scan lasting for 10 min was obtained, and the resulting images underwent reconstruction through the application of the three-dimensional ordered subset expectation maximization algorithm. Thereafter, the images were reconstructed by IAW. Finally, the region of the uterus and embryo was mapped under CT image guidance.

Global untargeted metabolomics

For metabolomics analysis and data processing, cell samples were transferred to 2-mL centrifuge tube and ground with liquid nitrogen at 55 Hz for 60 s. The homogenates were incubated on ice for 5 min and then centrifuged at 12,000 rpm for 10 min at 4 ℃ after a 5-min incubation in an ice bath. Supernatants were filtered through 0.22-μm membranes prior to LC–MS analysis using a UPLC Ultimate 3000 system coupled to a Q Exactive Orbitrap mass spectrometer.

Data collection and processing were executed with the assistance of SIEVE software (Thermo Fisher Scientific). The effectiveness of principal component analysis (PCA) was evaluated by visually inspecting the clustering of pooled samples and replicate samples. Additionally, the variable importance in projection (VIP) score for each metabolite was determined. For a more in-depth analysis, a univariate approach was employed. This involved calculating the statistical significance and fold change (FC) of each metabolite between two distinct groups, utilizing the t-test as the statistical method. The criteria for identifying differential metabolites were P < 0.05 and log FC > 1.5. Following this, hierarchical cluster analysis was carried out using the R package Pheatmap. To gain insights into the biological significance of the identified metabolites, The Kyoto Encyclopedia of Genes and Genomes (KEGG) database served as the primary resource for this exploration. P < 0.05 was considered to indicate statistical significance.

Seahorse XF assays for mitochondrial and glycolytic function

Mitochondrial and glycolytic functions in primary stromal cells were assessed using the Seahorse XFe96 Analyzer (Agilent Technologies). Briefly, cells were seeded in cell culture plates and allowed to adhere. On the day of the assay, the growth medium was replaced with Seahorse XF assay medium (pH 7.4), and cells were incubated at 37 °C in a non-CO₂ incubator for 1 h. For the mitochondrial stress test, oligomycin (1.5 µM), FCCP (1.0 µM), and a mix of rotenone/antimycin A (0.5 µM each) were sequentially injected into the wells. Concurrently, the glycolysis stress test was performed on the same preparation by sequential injection of glucose (10 mM), oligomycin (1 µM), and 2-deoxy-d-glucose (50 mM). The oxygen consumption rate (OCR) and extracellular acidification rate (ECAR) were measured in real time, allowing for the calculation of key parameters of mitochondrial respiration and glycolytic function.

Assessment of mitochondrial membrane potential

The mitochondrial membrane potential (MMP) was assessed using the JC-1 fluorescent probe. Cells were incubated with the JC-1 working solution according to the manufacturer’s instructions. For flow cytometric analysis, the cells were trypsinized, resuspended in JC-1 staining buffer, and analyzed. Flow cytometry was performed on a Beckman Coulter (CytoFlex S).

RNA-seq experiment and high-throughput sequencing

Total RNA from stromal cells of patients with URSA and women with normal pregnancy and T-hHSCs was isolated using TRIzol. RNA quality (purity, concentration, and integrity) was verified via an Agilent TapeStation 4200. RNA-seq libraries were prepared using Illumina-compatible kits and sequenced on a DNBSEQ-T7 platform. Reads were aligned to the CHOK1GS genome (Ensembl) with STAR (version 2.5.3a), followed by exon-level quantification (featureCounts, Subread version 1.5.1) for reads per kilobase million (RPKM) calculations. To visualize the overall variance and clustering between sample groups, principal component analysis (PCA) was performed. Differential expression (edgeR, P < 0.05, |log₂FC|≥ 1) and functional enrichment, such as Gene Ontology (GO) and KEGG, were analyzed using Metascape (https://metascape.org/) [36].

CUT&Tag

CUT&Tag assays were conducted by SeqHealth (Wuhan, China). Stromal cells from women with URSA and healthy controls were harvested. Briefly, cells were bound to concanavalin A-coated beads, incubated overnight at 4 °C with anti-H3K18la or IgG antibodies, then treated with secondary antibody and pG-Tn5 transposase. Tagmented DNA was amplified, purified, and sequenced on an Illumina Novaseq 6000 (PE150).

Data analysis

Raw sequencing data was first filtered by fastp (version 0.23.1). DeepTools (version 2.4.1) was used to visualize the distribution of reads on upstream and downstream of transcription start sites (TSS). Then, peak calling, annotation, and peak distribution were performed by MACS2 (version 2.2.7.1) and bedtools (version 2.30.0) software packages. The different peaks were identified by csaw (version 1.24.3) and used for motifs analysis by Homer (version 4.10). Peaks with |log₂FC|> 0.585 and false discovery rate (FDR) < 0.05 were classified as statistically significant differential peaks (DPs). The GO enrichment analysis of the annotated genes was performed using Metascape (https://metascape.org/).

Chromatin immunoprecipitation (ChIP)

Formaldehyde (4 °C, 12 min) and glycine (0.125 mol/L) were applied on day 6 of induced decidualization in vitro. Chromatin was sonicated to generate short fragments. The target protein was immunoprecipitated using anti-H3K18la, anti-CUX1 antibodies. The immunoprecipitated DNA was purified and eluted in a small volume of elution buffer (or water), and the resulting solution was used as a template for qPCR analysis. qPCR was conducted on immunoprecipitated DNA using SYBR Green Pro Taq HS Premix (AGBIO) with primer sequences detailed in Supplementary Table 3.

Acquisition of cell senescence associated genes

Cell senescence-associated genes were retrieved from the CellAge Database [37], Aging Atlas [38], Genecards [39], Genage [40], and Digital Ageing Atlas [41]. A total of 4696 cell senescence-associated genes were identified. The Venn online mapping tool [42] was used to determine the intersection of URSA-DEGs, CUT&Tag-identified genes, and cell senescence-associated genes.

CUX1 ChIP-seq binding sites at CXCL1, CXCL3, and TNF promoters

Following the prediction of CUX1 binding sites in promoter regions using the JASPAR database, ChIP-qPCR primers were designed. The CUX1 ChIP-seq datasets for MCF7 (GSE91415) and K562 (GSE105363) cells were retrieved from the Gene Expression Omnibus (GEO) [43]. These signals were then visualized using integrative genomics viewer (IGV) (version 2.16.2) for validation.

Statistical analysis

Data analysis utilized R (version 4.3.0) and GraphPad Prism 9.0, with results expressed as mean ± SEM. Statistical significance (P < 0.05) was determined using two-tailed t-tests or one-way analysis of variance (ANOVA) with Tukey or Dunnett’s post hoc tests.

Results

Clinical characteristics of the study cohort

To ensure the validity of our comparative analysis, we enrolled a total of 40 participants, comprising 20 women with URSA and 20 gestational age-matched healthy controls. As summarized in Supplementary Table 4, the two groups were well-matched with no significant differences in age, body mass index (BMI), gestation period, and menstrual cycle at sample collection. This confirms that the observed molecular differences are unlikely to be attributed to these basic clinical parameters.

Patients with URSA present decidualization deficiency accompanied by stromal cell senescence

To characterize the decidualization state in URSA, we evaluated decidual tissues from patients and healthy controls on the basis of pathohistological, morphological, and molecular biological analyses. Histological examination revealed that decidual tissue from the URSA group had fewer decidual stromal cells, along with increased inflammatory cell infiltration and vascular congestion compared with the control group (Fig. 1A). Successful decidualization involves significant morphological alterations, as ESCs transform into rounded, enlarged, secretory decidual cells, a process accompanied by intricate rearrangements. In the URSA group, however, F-actin polymerization and stress fiber production were disrupted, resulting in fewer organized, elongated stress fibers (Fig. 1B). Furthermore, the mRNA levels of the key decidualization markers PRL and IGFBP1 were significantly lower in patients with URSA than in healthy controls (Fig. 1C). Collectively, these data suggest that patients with URSA exhibit deficient decidualization.

Fig. 1.

Patients with URSA present decidualization deficiency accompanied by stromal cell senescence. A HE staining of decidual tissue in control and URSA group. The decidual tissue of the control group showing abundant decidual stromal cells, clear uterine glands, with no obvious congestion of blood vessels and rare inflammatory cell infiltration. Decidual tissues derived from the URSA group showing a small number of decidual stromal cells, accompany with significant inflammatory cell infiltration and congested blood vessels. Scale bar: 100 µm (left), 50 µm (right); B Double-immunofluorescence staining of F-actin and vimentin. Scale bar: 100 µm (left), 20 µm (right); C Relative mRNA expression of PRL and IGFBP1 in control and URSA groups verified by qRT-PCR (n = 12 per group); D, E Levels of p16 and p21 were verified by Western blot; Blot quantification and statistics are shown in the right panel (n = 5 per group); F Double-immunofluorescence staining of p16 and vimentin in the two groups. Scale bar: 100 µm (left), 20 µm (right) (n = 20 per group); G Double-immunofluorescence staining of p21 and vimentin in the two groups. Scale bar: 100 µm (left), 20 µm (right) (n = 20 per group); H–J Levels of IL-6, IL-1β, and TNF-α in stromal cells derived from control and URSA groups verified by ELISA assay (n = 20 per group). *P < 0.05, **P < 0.01, by two-tailed Student’s t-test

Previous studies have confirmed that stromal cell senescence can compromise endometrial decidualization [44]. Here, we preliminarily observed cell senescence in the decidual tissue of URSA. Senescent cells are characterized by irreversible replicative arrest and the secretion of a diverse array of pro-inflammatory factors known as SASP [45]. Western blot demonstrated a significant upregulation of the senescence markers p16 and p21 in the decidual tissue from patients with URSA (Fig. 1D, E), which could restrain cyclin-dependent kinase (CDK) 4, CDK6, and CDK2 (in the case of p21^CIP1) and halt the cell cycle [46]. Consistently, double-immunofluorescence staining confirmed increased co-localization of p16 and p21 with the stromal cell marker vimentin in the URSA group (Fig. 1F, G). In addition, ELISA assays revealed that stromal cells derived from URSA secreted significantly higher levels of SASP factors, including IL-6, IL-1β, and TNF-α (Fig. 1H–J). Herein, these data collectively indicate that decidualization deficiency observed in URSA is related to stromal cell senescence.

Stromal cell senescence contributes to decidualization deficiency in URSA

Previous works have indicated that stress-induced stromal cell senescence could lead to decidualization deficiency in vivo [44]. In this part, we first investigated the role of cellular senescence in decidualization through artificial decidualization models [32]. In this model, exogenous addition of estradiol and progesterone mimics hormonal changes in pregnancy in ovariectomized mice. In this procedure, corn oil was administered into one uterine horn to mimic the disruption of the epithelial layer caused by embryo attachment and subsequent decidualization (stimulated), while the untreated horn acted as a control (unstimulated). At the same time, cell senescence was induced with d-galactose [47], with distilled water as control (Fig. 2A). As shown in Fig. 2, d-galactose significantly induced cellular senescence within the decidua, as verified by increase co-colocalized of p16 and p21 with vimentin (Fig. 2B). Furthermore, this induction of senescence completely abrogated the decidualization response, as evidenced by a dramatic reduction in deciduoma weight (Fig. 2C). Consistent with this finding, both protein and mRNA levels of the decidualization markers PRL and IGFBP1 were significantly reduced in the d-galactose-treated mice (Fig. 2D). These result suggest that senescence in stromal cells is sufficient to impair their differentiation into decidual cells and cause decidualization failure.

Fig. 2.

Senescence of stromal cell is a critical contributor to mediating decidualization deficiency in URSA. A The experimental plan outline of cell senescence induction based on artificial decidualization model; B The double-immunofluorescence staining of p16 and p21 with vimentin in artificial decidualization mice treated with d-galactose or distilled water (n = 3 per group); C Stimulated uterine horn gross weight shown as a histogram (n = 6 per group); D Protein and mRNA levels of PRL and IGFBP1 verified by immunofluorescence and qRT-PCR in artificial decidualization mice treated with d-galactose or distilled water (n = 3–6 per group); E Experimental plan outline of cell senescence inhibition based on URSA murine model; F The double-immunofluorescence staining of p16 and p21 with vimentin. Scale bar: 50 µm (left), 10 µm (right) (n = 3 per group); G Embryo absorption of in URSA mice receiving senolytic drugs or normal saline compared with normal pregnancy mice (n = 6 per group); H, I Immunofluorescence and qRT-PCR to verify PRL and IGFBP1 levels in URSA mice receiving senolytic drugs or normal saline compared with normal pregnancy mice (scale bar: 50 µm) (n = 3–6 per group). *P < 0.05, **P < 0.01, by two-tailed Student’s t-test or repeated-measures one-way ANOVA followed by post hoc Dunnett’s multiple-comparisons test. ns: not significant

Next, to determine whether targeting senescence could rescue the URSA phenotype, we utilized the established CBA/J × DBA/2 mouse model of spontaneous abortion. Pregnant URSA-prone mice were intraperitoneally injected with senolytic cocktails (dasatinib and quercetin) or a vehicle control starting on gestation day 1 (GD1), and decidual tissues were collected on GD8 (Fig. 2E). The result showed that the heightened levels of senescence markers (p16 and p21) in the URSA group were significantly reduced by senolytic treatment, reaching levels indistinguishable from the normal controls (Fig. 2F). Remarkably, senolytic treatment significantly reduced embryo resorption numbers (Fig. 2G). Additionally, to investigate the effects of senolytic intervention on decidualization process in URSA, we quantified decidual PRL and IGFBP1 expression via immunofluorescence intensity and qPCR on GD8. Senolytic treatment restored PRL and IGFBP1 expression in URSA mice to levels indistinguishable from those in normal controls (Fig. 2H, I). Collectively, these findings indicate that stromal cell senescence is a critical and pharmacologically targetable driver of decidualization deficiency and embryo loss in URSA.

Glycolysis driven by HK2 induces stromal cell senescence in URSA

Cellular metabolism is highly dynamic and profoundly impacts cell fate decisions, including differentiation and senescence [48]. To obtain a comprehensive snapshot of the metabolic phenotypes in URSA stromal cells, we performed global untargeted metabolomics. This analysis identified 555 differential metabolites in URSA stromal cells (Fig. 3A–D; Supplementary Table 5). A volcano plot was used to visualize the distribution of all detected metabolites, highlighting those with significant alterations (Fig. 3E), while the corresponding heatmap illustrates the distinct expression patterns of the differential metabolites between the two groups (Fig. 3F). KEGG pathway analysis revealed significant enrichment of glycolysis/gluconeogenesis, pyruvate metabolism, and ferroptosis among differential metabolites (Fig. 3G). Notably, lactate levels were significantly elevated in URSA samples, as confirmed by both untargeted metabolomics (Fig. 3H) and a lactate content assay (Fig. 3I).

Fig. 3.

Glycolysis related lactate could accounts for stromal cell senescence and decidualization deficiency in URSA. A, B Total ion chromatogram of metabolites between stromal cells derived from control and URSA group (A: positive ion model; B: negative ion model); C, D PCA score plot between stromal cells derived from control and URSA group, showing that samples in the groups were closely clustered to one another (C: positive ion model; D: negative ion model); E Volcano plot of differential metabolites between stromal cells of control and URSA group. Red dots represent upregulated metabolites in the URSA group, while blue dots represent downregulated metabolites; F Heat map of differential metabolites between stromal cells of control and URSA group. The upregulated metabolites are marked in red, while the downregulated metabolites levels are presented in blue; G Bubble map of the impact values of metabolic pathways, where the horizontal coordinate is the impact value enriched into different metabolic pathways and the vertical coordinate is the enrichment pathway. The size of the dots indicates the corresponding number of metabolites on the pathway. The color of the dot reflects the P value, where the yellower the color, the smaller the P value, and the bluer the color, the larger the P value; H The levels of lactic acid between stromal cells derived from control and URSA group on the basis of global untargeted metabolomics; I The levels of lactic acid detected by lactate content assay kit in stromal cells between control and URSA group (n = 6 per group); J The correlation of the fluorescence intensity between lactate and p16, p21, respectively; K The experimental plan outline of in vitro decidualization of T-hESCs by using 8-Br-cAMP and MPA with or without NaLa incubation; L The T-hESCs were examined for cell senescence by staining for SA-β-gal activity. Scale bar: 100 µm (n = 3 per group); M, N Immunofluorescence and western blot to detect the level of p16 (M), p21 (N), with or without NaLa incubation (confocal laser fluorescence microscopy, scale bar: 20 µm) (n = 3–5 per group); O, P Immunofluorescence to detect the level of γH2AX (O) and F-actin (P), with or without NaLa incubation (confocal laser fluorescence microscopy, scale bar: 20 µm); Q mRNA levels of PRL and IGFBP1 in different groups were verified by qRT-PCR (n = 6 per group). * P < 0.05, **P < 0.01, by two-tailed Student’s t-test

To investigate the mechanistic basis for lactate accumulation, we assessed mitochondrial function. Seahorse XF analysis demonstrated significant mitochondrial dysfunction in URSA stromal cells, evidenced by impaired oxidative phosphorylation, including reduced basal, maximal, and ATP-linked respiration (Supplementary Fig. 3A). Concurrently, these cells exhibited a compensatory glycolytic shift, with enhanced glycolytic capacity and reserve (Supplementary Fig. 3B). We further visualized mitochondrial integrity using the JC-1 probe. URSA stromal cells showed a pronounced loss of mitochondrial membrane potential (MMP) (Supplementary Fig. 3C). Collectively, these functional assays support the metabolomic findings and substantiate the occurrence of a glycolytic switch in URSA, which underlies the elevated lactate production.

Given that glycolysis can regulate decidualization and affect cellular aging reprogramming [49], we next sought to directly link lactate accumulation to the pathological phenotype in URSA. First, in primary patient samples, we found that lactate levels exhibited a direct proportionality with the fluorescence intensities of p16 and p21 proteins (Fig. 3J). To test for causality, we treated T-hESCs with exogenous sodium lactate (NaLa) during in vitro decidualization. We first established the baseline morphological changes of this model: compared with vehicle-treated controls that remained fibroblast-like and spindle-shaped, cells induced with 8-Br-cAMP and MPA underwent a robust transformation into large, epithelioid cells, characterized by the reorganization of F-actin into a dense cortical ring—a hallmark of successful decidualization (Supplementary Fig. 4). Against this validated baseline, we found that NaLa treatment accelerated premature senescence, as confirmed by an increased level of the number of SA-β-gal-stained cells (Fig. 3K, L), a well-defined biomarker for cellular senescence. Consistent with this, immunofluorescence and western blotting confirmed the elevation of p16, p21, and γ-H2AX in response to lactate (Fig. 3M–O). Moreover, NaLa treatment severely resulted in decidualization deficiency, characterized by F-actin disorganization (Fig. 3P) and reduced mRNA level of PRL/IGFBP1 (Fig. 3Q). Taken together, these findings suggest that lactate, a product of glycolysis, contributes, at least in part, to stromal cell senescence and decidualization deficiency in URSA.

To investigate the molecular drivers of the glycolytic shift in URSA from gene expression level, we performed RNA-seq on primary stromal cells from patients and healthy controls. Principal component analysis (PCA) of the resulting transcriptomic data revealed a clear separation between the URSA and control groups, confirming a distinct gene expression signature for each cohort (Supplementary Fig. 5). Subsequent differential expression analysis identified 2006 upregulated genes and 2394 downregulated genes (|log FC|≥ 1, P < 0.05) in the URSA group (Fig. 4A, B). KEGG pathway analysis also revealed that these DEGs were mainly involved in the regulation of nuclear factor kappa-light-chain-enhancer of activated B cells (NF-κB) signaling pathway, glycolysis/gluconeogenesis, and HIF-1 signaling pathway (Fig. 4C). Thus, we further focus on the expression profile of 11 key glycolytic rate-limiting enzymes (Fig. 4D) in stromal cells of URSA and healthy control based on RNA-seq. As illustrated by Fig. 4E, there were nine rate-limiting enzymes differentially expressed between URSA and healthy controls (HK2, GAPDH, LDHA, ENO1, ENO2, ALDOA, PFKM, GPI, and PKM). Furthermore, qRT-PCR validation confirmed that the expression profiling of HK2 and PGK1 were significantly elevated in URSA stromal cells when compared with controls (P < 0.01) (Fig. 4F). Besides, the double-immunofluorescence staining results verified an increased level of co-localization of HK2 and vimentin in URSA (Fig. 4G). Taken together, these data provide insight that lactate derived from HK2-dependent glycolysis induced stromal cell senescence and decidualization deficiency in URSA.

Fig. 4.

Lactate derived from HK2-dependent glycolysis induced stromal cell senescence and decidualization deficiency in URSA. A Volcano plot illustrating transcriptional dysregulation of differentially expressed genes in stromal cells: Control group versus URSA group; B Heatmap depicting transcriptional disparities of differentially expressed genes (DEGs) in stromal cells derived from the control group versus the URSA cohort; Upregulated genes are marked in red, and downregulated genes levels are presented in blue; C KEGG pathway enrichment analysis of differentially expressed genes between stromal cells of control and URSA groups; D Schematic diagram of key enzymes in the glycolysis process; E Heat map of glycolysis related enzymes based on the RNA-seq; F qRT-PCR was utilized to validated the expression of differentially expressed glycolysis related enzymes in stromal cells of control and URSA group (n = 6 per group); G Double-immunofluorescence staining to verify the level of co-localization of HK2 and vimentin. Scale bar: 100 µm (left), 20 µm (right) (n = 20 per group). **P < 0.01, by two-tailed Student’s t-test. ns: not significant

H3K18 lactylation mediated stromal cell senescence and decidualization deficiency in URSA

Lactylation of histones represents a recently identified post-translational modification (PTM) that bridges cellular metabolic state and epigenetic inheritance. Since lactate can be a precursor to stimulate histone lactylation, we projected that histone lactylation may be altered in URSA. Western blot revealed a significantly higher level of global lysine lactylation (Pan-Kla) in decidual tissues from patients with URSA compared with healthy controls (Fig. 5A). Immunofluorescence assay confirmed that this increased lactylation was prominent in vimentin-positive stromal cells (Fig. 5B). Among the lactylated histones, a band at ~17 kDa, corresponding to histone H3, exhibited the highest lactylation levels. Thus, we further explored whether lactylation modification of H3 was involved in stromal cell senescence. As a result, H3K18la levels showed the same trend in stromal cells of URSA as the overall lactylation levels did, with no change in H3K9la, H3K23la, and H3K56la (Fig. 5C–F). Consistently, co-localization of H3K18 with vimentin was significantly enhanced in decidual tissue of URSA (Fig. 5G). In addition, H3K18la levels positively correlated with p16/p21 expression in patients with URSA (with p16, R = 0.72; with p21, R = 0.77, respectively) (Fig. 5H). These results suggest H3K18la as a specific epigenetic mark associated with stromal cell senescence in URSA.

Fig. 5.

H3K18la depending on HK2-mediated glycolysis was closely to stromal cell senescence in URSA. A Global lactylation levels were detected in stromal cells of control and URSA group by western blot (n = 5 per group); B Double-immunofluorescence staining to verify the level of co-localization of global lactylation and vimentin. Scale bar: 100 µm (left), 20 µm (right) (n = 20 per group); C–F H3K9la, H3K18la, H3K23la, and H3K56la levels were detected in stromal cells of control and URSA group by western blot (n = 5 per group); G Double-immunofluorescence staining to verify the level of co-localization of H3K18la and vimentin. Scale bar: 100 µm (left), 20 µm (right) (n = 20 per group); H The correlation between H3K18la expression levels and p16, p21 levels in URSA. *P < 0.05, **P < 0.01, by two-tailed Student’s t-test. ns: not significant

To functionally connect HK2-driven glycolysis with H3K18la and the pathological phenotype, we performed a series of gain- and loss-of-function experiments in T-hESCs. HK2 OE plasmid significantly increased H3K18la levels (Supplementary Fig. 6A and Fig. 6A) and induced a premature senescence phenotype, characterized by increased SA-β-gal staining and upregulation of p16, p21, and γ-H2AX during induced decidualization in vitro (Fig. 6B–D). This HK2-induced senescence was accompanied by severe decidualization defects, including a disorganized, fibroblast-like actin cytoskeleton and suppressed expression of PRL and IGFBP1 (Fig. 6E, F). Importantly, these effects were reversed by treatment with the glycolysis inhibitor oxamate, which restored normal H3K18la levels, attenuated senescence, and rescued decidualization (Fig. 6A–F). Conversely, we treated cells with CoCl₂ to induce hypoxia and glycolysis. In this model, siRNA-mediated knockdown of HK2 significantly inhibited the H3K18la level (Supplementary Fig. 6B and Fig. 6G) and rescued stromal cell s from senescence (Fig. 6H–J). This rescue of the molecular phenotype translated to a functional recovery, as HK2 knockdown restored proper cytoskeletal organization and normalized the expression of PRL and IGFBP1 (Fig. 6K, L). Collectively, these results validated that lactate derived from HK2-mediated glycolysis could promote stromal cell senescence through H3K18la, which may be the key mechanism to explain decidualization deficiency in URSA.

Fig. 6.

H3K18 lactylation mediated stromal cell senescence and decidualization deficiency in URSA. A–F T-hESCs were randomly divided into Con + NC plasmid, Con + HK2 OE, and Con + HK2 OE + oxamate group and were induced for decidualization. The confocal laser fluorescence microscopy was utilized to assess the abundance of H3K18la of H3K18la (A), p16 (B), p21 (C), γH2AX (D), and F-actin (E). And the mRNA levels of PRL and IGFBP1 were verified by qRT-PCR in the three mentioned above groups (F); G–L T-hESCs were randomly divided into CoCl₂ + scramble siRNA, CoCl₂ + HK2 siRNA, and CoCl₂ + HK2 siRNA + NaLa group, and were then induced for decidualization. The confocal laser fluorescence microscopy was utilized to assess the level of H3K18la (G), p16 (H), p21 (I), γH2AX (J), and F-actin (K). The mRNA levels of PRL and IGFBP1 in different groups were verified by qRT-PCR (L). *P < 0.05, **P < 0.01, by repeated-measures one-way ANOVA followed by post hoc Dunnett’s multiple-comparisons test (n = 3–4 per group). Scale bar: 20 µm

H3K18la promotes stromal cell senescence by promoting CUX1 transcription

Histone lactylation is known to epigenetically activate transcriptional programs by enhancing chromatin accessibility through structural relaxation of nucleosome arrays at promoter-proximal regions [50]. To investigate how lactate affect global gene expression via H3K18 lactylation in URSA, we performed genome-wide cleavage under targets and tagmentation (CUT&Tag) analysis in primary stromal cells from patients with URSA and controls to screen candidate genes regulated by H3K18la (Fig. 7A). Analysis identified 1548 differential H3K18la peaks, with 22.29% mapped to promoter–TSS regions (Fig. 7B, C). To investigate the epigenetic regulatory effects of H3K18la on stromal cells, target genes of H3K18la binding peaks at promoter were classified by Gene Ontology (GO) analysis, and these genes were mainly enriched in regulation of MAP kinase activity, regulation of DNA-templated transcription initiation, apoptotic DNA fragmentation, and so on (Fig. 7D), suggesting that H3K18la may affect various biological processes.

Fig. 7.

H3K18la promotes stromal cell senescence by promoting CUX1 transcription. A Schematic of stromal cell isolation from healthy controls and individuals with URSA; B The binding density of H3K18la visualized by deepTools: the heatmap presents the CUT&Tag tag counts on the different H3K18la binding peaks in stromal cells between controls and women with URSA, ordered by signal strength; C Genome-wide distribution of upregulated and downregulated H3K18la-binding peaks in stromal cells between controls and women with URSA; D GO-BP analysis of the H3K18la binding peaks at promoter genes; E The intersection of senescence related genes with upregulated DEGs in URSA marked with H3K18la at their promoter regions; F qRT-PCR assays monitoring the expression of the CUX1, PAM, and RAPGEF4 in stromal cells between controls and women with URSA (n = 12 per group); G Expression and co-localization of H3K18la and CUX1 in stromal cells of control and URSA groups analyzed by immunofluorescence-based three-dimensional reconstruction technique. Scale bar: 20 µm; H The correlation between of CUX1 and global lactylation (left panel) and H3K18la (right panel) levels in decidual tissue of patients with URSA; I The upper panel is an IGV snapshot of H3K18la tracks across the CUX1 gene locus. The lower panel shows a zoomed-in view of H3K18la ChIP-seq signals at the promoter region of CUX1. The red box in the lower panel indicates the peaks closest to the transcription start site (TSS) region of the CUX1 gene; J H3K18la levels at the CUX1 promoter regions in control and URSA stromal cells analyzed by ChIP-qPCR assay; Normal rabbit immunoglobulin G (IgG) was used as a negative control. Data were normalized to input and are expressed as mean fold enrichment over input relative to an IgG control (n = 3 per group); K H3K18la levels at the CUX1 promoter regions analyzed by ChIP-qPCR assay in T-hESC versus T-hESC treated with 25 mM NaLa (n = 3 per group); L qRT-PCR assays monitoring expression of the CUX1 in T-hESC treated with different concentrations of NaLa (0, 5, 10, and 25 mM) for 24 h (n = 6 per group); M Western blot to detect the level of CUX1, with or without 25 mM NaLa incubation (n = 5 per group); N Schematic proposing a mechanism of H3K18la promoting stromal cell senescence via targeting CUX1 transcription in URSA.*P < 0.05, **P < 0.01, by two-tailed Student’s t-test. L *P < 0.05, by repeated-measures one-way ANOVA followed by post hoc Tukey multiple-comparisons test

To further elucidate the potential target genes of H3K18la in regulating stromal cell senescence in URSA, we employed a multi-omics filtering strategy. We first intersected our CUT&Tag data with our RNA-seq data, identifying 13 genes that were both upregulated in URSA and marked by increased H3K18la at their promoter regions. We then cross-referenced this list with a curated database (Gencards, CellAge Database, Aging Atlas, Genage, and Digital Ageing Atlas) of 4696 known senescence-associated genes. This stringent filtering process yielded three high-confidence candidates: CUX1, PAM, and RAPGEF4 (Fig. 7E). Validation by qRT-PCR confirmed that all three genes were upregulated in URSA stromal cells, but CUX1 exhibited the most pronounced and statistically robust increase (Fig. 7F). Critically, a functional correlation analysis revealed that only CUX1 mRNA levels were significantly and positively correlated with the proportion of SA-β-gal-positive senescent cells in patient samples (Supplementary Fig. 7). Furthermore, CUX1 expression was strongly correlated with global lactylation and H3K18la levels in patients with URSA, in both primary stromal cells and decidual tissue samples (Fig. 7G, H).

To confirm a direct epigenetic link, we examined the CUX1 locus. Both CUT&Tag peak analysis and ChIP-qPCR confirmed a significant enrichment of H3K18la at the CUX1 promoter in URSA stromal cells compared with controls (Fig. 7I, J). To establish causality, we demonstrated in T-hESCs that exogenous sodium lactate (NaLa) treatment was sufficient to both increase H3K18la enrichment at the CUX1 promoter and drive CUX1 transcription in a dose-dependent manner (Fig. 7K, L). Furthermore, Western blot analysis confirmed that high-dose NaLa treatment led to a significant increase in CUX1 protein expression (Fig. 7M). Conversely, treating primary URSA stromal cells with A-485, a potent p300/CBP inhibitor that suppresses lactylation [29], reduced H3K18la levels, abolished its enrichment at the CUX1 promoter, and consequently suppressed CUX1 expression (Supplementary Fig. 8). These findings collectively suggest that CUX1 as the primary transcriptional target linking H3K18la to stromal cell senescence in URSA (Fig. 7N).

Fig. 8.

CUX1 may promote stromal cell senescence by activating promoter activity of SASP factors. A Heat map of differentially expressed genes in T-hESCs treated with CUX1 OE plasmid versus NC plasmid group during induced decidualization. The upregulated genes are marked in red, and the downregulated genes are presented in blue; B Volcano plot illustrating the differentially expressed genes in T-hESCs following treatment with the CUX1 overexpression (OE) plasmid, in comparison with the negative control (NC) plasmid group, during the process of induced decidualization; C KEGG pathway enrichment analysis on differentially expressed genes between T-hESCs treated with CUX1 OE plasmid versus NC plasmid group during induced decidualization; D–I ChIP-qPCR analysis on the enrichment status of CUX1 at the promoter of EDA (D), CXCL1 (E), CXCL3 (F), PRKCB (G), TNF-α (H), and TNFSF13B (I) in T-hESCs treated with CUX1 OE plasmid versus NC plasmid group during induced decidualization. *P < 0.05, by two-tailed Student’s t test (n = 3 per group). ns: not significant

CUX1 promotes stromal cell senescence by activating promoter activity of specific SASP factors

To elucidate the downstream mechanisms of CUX1, we performed RNA-seq on T-hESCs following its overexpression. We first confirmed that the CUX1 overexpression (CUX1 OE) plasmid significantly increased CUX1 levels (Supplementary Fig. 9A). Subsequent RNA-seq analysis revealed 2184 upregulated and 1044 downregulated genes (log FC ≥ 1.5 and P < 0.05) (Fig. 8A, B), with upregulated targets enriched in “NF-κB signaling pathway” and “Cellular senescence pathway” via KEGG analysis (Fig. 8C). Since the NF-κB pathway is critical in regulating the transcription of pro-inflammatory SASP factors [51], we focused our analysis on the 22 DEGs enriched in the NF-κB pathway (Fig. 8C). Subsequently, six genes (EDA, CXCL1, CXCL3, PRKCB, TNF, and TNFSF13B) out of the 22 DEGs were screened as potential targets of CUX1 by using the JASPAR database (Supplementary Table 6). Additionally, interrogation of CUX1 ChIP-seq datasets revealed a conserved CUX1–NF-κB composite motif within CXCL1, CXCL3, and TNF promoter regions (Supplementary Fig. 10). This is consistent with prior studies demonstrating that CUX1 and NF-κB p65 can form molecular complex to mediate CXC chemokine transcription (CXCL1, CXCL2, CXCL3, and CXCL6) [52]. Bioinformatic analysis has identified CUX1 cis-elements proximal to NF-κB binding regions in TNF-α and IL-6 promoters [53]. All of these are important components of SASP factors and play a key role in promoting the aging of neighboring cells. To confirm this direct regulation in our system, we performed ChIP-qPCR in decidualizing T-hESCs. CUX1 overexpression markedly increased the binding of CUX1 to the gene promoters of CXCL1, CXCL3, and TNF-α (Fig. 8D–I). These findings present that CUX1 is the upstream regulator of CXCL1, CXCL3, and TNF-α. Collectively, these findings establish that CUX1 is an upstream transcriptional activator of specific SASP components, such as CXCL1, CXCL3, and TNF-α.

Fig. 9.

H3K18la facilitated stromal cell senescence via CUX1-mediated SASP factor transcription in URSA. A–D T-hESCs were randomly divided into CUX1 scramble, NaLa + CUX1 scramble, and NaLa + CUX1 siRNA group, and were induced for decidualization. qRT-PCR was utilized to detect the mRNA levels of CXCL1, CXCL3, and TNF-α (A); The confocal laser fluorescence microscopy was utilized to assess the p16, p21, and γH2AX levels (B) (scale bar: 20 µm); The stromal cell senescence was assessed by SA-β-gal staining (C) (scale bar: 100 µm); The confocal laser fluorescence microscopy was utilized to assess the F-actin level (D) (scale bar: 20 µm); Besides, PRL and IGFBP1 in different groups were verified by qRT-PCR (D) (n = 3–6 per group); E–I T-hESCs were randomly divided into CoCl₂ + CUX1 NC plasmid, CoCl2 + oxamate + CUX1 NC plasmid, and CoCl₂ + oxamate + CUX1 OE plasmid group, and were induced for decidualization. The confocal laser fluorescence microscopy was utilized to detect H3K18la (E) (scale bar: 20 µm) and CUX1 (F) (scale bar: 20 µm) in the three above-mentioned groups; Besides, the mRNA levels of CXCL1, CXCL3, and TNF-α were verified by qRT-PCR (G); Confocal laser fluorescence microscopy was utilized to assess the p16, p21, and γH2AX levels (H) (scale bar: 20 µm), as well as the F-actin level (I) (scale bar: 20 µm); qRT-PCR was utilized to to assess the mRNA level of PRL and IGFBP1 (I) (n = 3–6 per group). *P < 0.05, **P < 0.01, by two-tailed Student’s t-test or repeated-measures one-way ANOVA followed by post hoc Dunnett’s multiple-comparisons test

Then, loss-of-function experiments were performed to analyze the biological function of CUX1 in H3K18la-regulated stromal cell aging. A CUX1 siRNA was transfected in T-hESC, followed by NaLa treatment to H3K18la modification. As we hypothesized, CUX1 siRNA significantly downregulated CUX1 expression (Supplementary Fig. 9B). In addition, during induced decidualization, CUX1 knockdown dramatically rescued the cells from the effects of lactate. This rescue was evident at the molecular level, with a significant reduction in the expression of SASP markers (CXCL1, CXCL3, and TNF-α) and senescence markers (p16, p21, and γ-H2AX) (Fig. 9A, B). This molecular rescue translated to a functional recovery: CUX1 knockdown significantly attenuated the number of SA-β-gal staining positive cells and restored proper decidualization, as evidenced by oval-shaped distribution of F-actin and high expression level of PRL and IGFBP1 (Fig. 9C, D).

In addition, T-hESCs were exposure to CoCl₂ to induce hypoxia and glycolysis, with or without oxamate treatment (Supplementary Fig. 11). As shown by Fig. 9E, glycolysis inhibitor oxamate significantly reduced H3K18la levels during induced decidualization. This inhibition of glycolysis had profound downstream effects: oxamate treatment suppressed the upregulation of CUX1 (Fig. 9F), reduced the expression of SASP factors (CXCL1, CXCL3, and TNF-α) (Fig. 9G) and senescence markers (p16, p21, and γ-H2AX) (Fig. 9H), and fully rescued decidualization, as evidenced by a well-organized F-actin cytoskeleton and restored PRL and IGFBP1expression (Fig. 9I). Critically, we then tested whether CUX1 could override this rescue; forced overexpression of CUX1 completely reversed the beneficial effects of glycolysis inhibition. CUX1 OE re-induced the full senescence program (Fig. 9G, H) and caused a return to an aberrant decidualization state, as verified by the long shuttle-shaped distribution of F-actin fluorescence and decreased mRNA level of PRL and IGFBP1 (Fig. 9I).

Finally, to validate the therapeutic potential of targeting CUX1 in vivo, we used the CBA/J × DBA/2 mouse model of URSA. We achieved in situ gene silencing of CUX1 in the murine uterus through local delivery of an adeno-associated virus vector expressing a CUX1 shRNA via intrauterine injection (Fig. 10A). The results showed that CUX1 shRNA effectively inhibited the expression of CUX1 in decidual tissues, especially in stromal cells in the URSA murine model (Supplementary Fig. 12). Remarkably, in vivo knockdown of CUX1 significantly alleviated the pathological features of URSA. This was evidenced by a marked decrease in SASP factor secretion (CXCL1, CXCL3, and TNF-α) (Fig. 10B) and a reduction in stromal cell senescence, as confirmed by decreased colocalization of p16/p21 with vimentin (Fig. 10C), and lower overall p16/p21 protein levels (Fig. 10D). In addition, upon injection of URSA mice with CUX1 shRNA, we observed increased level of PRL and IGFBP1 (Fig. 10E) and reduced embryo resorption (Fig. 10F, G). Collectively, these data indicate that CUX1 is a candidate target gene of H3K18la, which facilitates the expression of specific SASP factors, such as CXCL1, CXCL3, and TNF-α, thus accelerating stromal cell senescence in URSA.

Fig. 10.

CUX1 knockdown results in cell senescence, decidualization deficiency, and embryo absorption in URSA murine model. A Schematic representation of the CUX1 or scramble AAV injection for in vivo CUX1 manipulation in URSA murine model; B ELISA assay detected the CXCL1, CXCL3, and TNF-α levels in stromal cells of URSA murine model treated with CUX1 or scramble shRNA (n = 8 per group); C Double-immunofluorescence staining for the p16 and p21 levels in stromal cells derived from URSA murine model treated with CUX1 or scramble shRNA. Scale bar: 50 µm (left), 10 µm (right); D p16 and p21 expression levels verified by Western blot in URSA murine model treated with CUX1 or scramble shRNA (n = 4 per group); E PRL and IGFBP1 mRNA expression verified by qRT-PCR in URSA murine model treated with CUX1 or scramble shRNA (n = 6 per group); F Mice embryo absorption in URSA murine model treated with CUX1 or scramble shRNA (n = 6 per group); G Positron emission tomography (PET)–computed tomography (CT) imaging to identify successful pregnancy in URSA murine model treated with CUX1 or scramble shRNA. *P < 0.05, **P < 0.01, by repeated-measures one-way ANOVA followed by post hoc Tukey multiple-comparisons test. ns: not significant

Discussion

Successful pregnancy requires a series of precisely orchestrated events, among which decidualization is paramount. This transformation of the endometrium provides a nutritive and immunoprivileged matrix essential for tuneful crosstalk between fetus and mother by secreting cytokines, morphogens, and signaling molecules [54]. A growing body of evidence indicates that an aberrant response of ESCs to deciduogenic cues is a key contributor to URSA [17]. Thus, unraveling the molecular mechanisms involved in the impairment of decidualization has been a subject of interest in the field of URSA.

Recent single-cell transcriptome analysis has unraveled that stromal cells in URSA undergo abnormal development, characterized by the upregulation of genes linked to cellular senescence [55]. These findings led us to hypothesize that impaired decidualization in URSA is driven by premature senescence of stromal cells. In support of this hypothesis, our study found an increased level of p21 and p16 in the decidual tissue of URSA, especially in the stromal cell compartment. Upregulation of these proteins is a well-established trigger for the irreversible cell cycle arrest that defines cellular senescence [56]. Besides, we found that these senescent stromal cells secreted high levels of SASP factors, such as IL-1β, IL-6, and TNF-α, This pro-inflammatory secretome can reinforce and propagate senescence in adjacent cells via paracrine signaling, a phenomenon known as “bystander” senescence [45, 57]. This phenomenon significantly contributes to the deterioration of decidua functionality. To establish causality, we demonstrated in vivo that d-galactose-induced systemic senescence was sufficient to impair artificial decidualization in mice. More importantly, we showed that clearing senescent cells with senolytic drugs significantly rescued decidualization and reduced embryo absorption in URSA murine model. Collectively, these data provide strong evidence that stromal cell senescence is a key driver of decidualization deficiency in URSA. Our findings are consistent with recent studies showing that senescent stromal can impair ligand–receptor interaction with trophoblasts, convey reactive oxygen species to the trophoblasts, and hinder the dissemination of blastocyst-like spheroids [44, 58]. Thus, targeting premature stromal cell senescence holds considerable promise for effective therapies of URSA. However, the specific molecular pathways that initiate and sustain this pathological senescence in the decidua have remained poorly understood, representing a critical knowledge gap that our present study aimed to address.

Glycolysis, once considered a mere consequence of a cell’s metabolic state, is now recognized as a crucial determinant in regulating cellular processes such as proliferation, differentiation, senescence, and quiescence [59, 60]. During the first trimester, the endometrium experiences physiological hypoxia as invading extravillous cytotrophoblast cells temporarily plug the uterine spiral arteries [61, 62]. Furthermore, in patients with RSA, the decidual microenvironment appears to experience a more prolonged and pathological hypoxia, which promotes a shift toward glycolytic energy metabolism [19]. These data were in line with our findings, which verified an increased expression of HK2 and a corresponding accumulation of lactate in stromal cells from individuals with URSA. This metabolic switch from oxidative phosphorylation to glycolysis allows for rapid adenosine (ATP) production but also results in substantial lactate generation. Emerging evidence has recast lactate from a metabolic waste product to a critical signaling molecule, or “lactyl-hormone,” capable of orchestrating diverse cellular responses [63]. In this study, we provide a direct link between this metabolic state and the senescence phenotype, demonstrating a significant positive correlation between lactate concentration and the expression of senescence markers p16 and p21 in URSA stromal cells. Furthermore, it was very interesting that lactate treatment significantly induced changes associated with stromal cell senescence. Similarly, Zhang et al. revealed that high concentrations of lactate could regulate Akt kinase activity and thus induce oxidative stress and senescence [64]. In addition, lactate has been shown to drive the SASP in stromal cells by promoting reactive oxygen species production via NADPH oxidase 1 [21]. Therefore, our data strongly support a model where lactate, produced as a consequence of HK2-driven glycolysis, is a key driver of stromal cell senescence in URSA. To our knowledge, this is the first study to establish this complete mechanistic link between glycolytic reprogramming and pathological cell senescence in the context of recurrent pregnancy loss, providing novel insights into the disease’s pathogenesis.

Histone lactylation, a recently identified post-translational modification process, serves as a direct bridge between cellular metabolism and regulation. Mechanistically, lysine lactylation is conceptually similar to the well-studied lysine acetylation. Both modifications neutralize the positive charge of the lysine residue, which is thought to weaken the electrostatic interaction between the histone tail and the negatively charged DNA backbone. This promotes a more “open” or relaxed chromatin structure, facilitating transcriptional activation. Furthermore, the enzymatic machinery that governs these two marks shows significant overlap; the acetyltransferase p300/CBP has been identified as a primary “writer” for both acetylation and lactylation, while class I histone deacetylases (HDACs 1–3) can function as “erasers” for both marks [65]. Dysregulation of histone lactylation is frequently identified in several human diseases [66], highlighting the importance of histone-based gene regulation. However, to date, its role in reproductive disorders has remained unexplored. Our study addresses this gap, providing what is, to our knowledge, the first evidence of aberrant histone lactylation in URSA. Specifically, we identified a noteworthy and specific elevation of H3K18la in the stromal cells of individuals with URSA. We further established a strong link between this epigenetic mark and the pathological phenotype, demonstrating that H3K18la levels positively correlate with senescence markers and are functionally associated with the decidualization deficiency in URSA. Our findings are consistent with emerging research in other fields; for instance, Wei et al. revealed that H3K18la elevation activates NF-κB signaling through enhanced promoter binding of p65/p50 subunits, driving SASP upregulation of key factors IL-6 and IL-8 [67]. These data suggest that H3K18la, fueled by the glycolytic shift in URSA, may be responsible, at least in part, for the stromal cell senescence in URSA.

Furthermore, our study extends previous work by providing a novel mechanistic link between histone lactylation and the regulation of cellular aging. By integrating our RNA-seq, CUT&Tag, and ChIP-qPCR data, we disclosed for the first time that H3K18la stimulates the transcription of homeobox transcription factor CUX1, which in turn orchestrates the pro-senescent SASP program in URSA stromal cells. CUX1 is a highly conserved nonclustered homeobox transcription factor, with one homeodomain and three cut repeat DNA-binding domains. A post-genome-wide association studies (GWAS) functional study demonstrated that CUX1 could regulate the expression of cell senescence related genes in endothelial cells (ECs). CUX1 overexpression induces both replicative and stress-induced senescence through p16 INK4a upregulation [68]. Consistent with this, our study demonstrated elevated CUX1 levels in URSA stromal cells. Besides, our own transcriptomic analysis following CUX1 overexpression revealed an enrichment of genes involved in inflammation and cell senescence. Mechanistically, we provide direct evidence that CUX1 binds to the promoter region of multiple SASP factors (CXCL1, CXCL3, and TNF-α). This finding is corroborated by bioinformatic analyses and prior work showing that CUX1 can form a complex with NF-κB to mediate chemokine transcription. In line with our study, a study by Wu revealed that IL-1β, a SASP gene itself, was increased by upregulating CUX1 in human endothelial cells [69]. Jiang et al. demonstrate that elevated CUX1 expression in senescent endothelial cells drives aging mechanisms, with ectopic CUX1 overexpression triggering senescence through p16 pathway activation [68]. Taken together, our results establish a critical and previously unknown role for CUX1 as the pivotal transcription factor that translates the epigenetic signal of H3K18la into a functional program of stromal cell senescence and decidualization failure in URSA.

Critically, the pathophysiological relevance of this metabolism–epigenetics–senescence axis was substantiated in a well-established URSA murine model (CBA/J × DBA/2). This model recapitulates key features of the human condition, including elevated embryo resorption and impaired decidualization [34]. Our demonstration that senolytic treatment or uterine-specific knockdown of CUX1 within this model significantly improved decidualization and pregnancy outcomes confirms that stromal cell senescence is not merely a cellular phenomenon but a critical pathogenic driver in vivo, and validates the therapeutic potential of targeting this axis. To further solidify the causal links within our proposed pathway, we employed a pharmacological approach in primary URSA stromal cells. Treatment with A-485, a potent inhibitor of the p300/CBP enzymes that act as “writers” for lactylation, effectively closed the mechanistic loop. The observation that A-485-mediated reduction of H3K18la led to a concomitant downregulation of CUX1 and suppression of the SASP provides direct experimental evidence that H3K18la is functionally upstream of CUX1 activation and is required for propagating the senescent phenotype. This positions H3K18la not merely as a biomarker but as a key functional node in the pathway. Of note, our study is the first to implicate CUX1 in the context of a female reproductive disorder. And, it is of particular interest considering that stromal cell senescence in URSA can be ascribed, at least to some extent, to H3K18la/CUX1 signaling driven by HK2-dependent glycolysis. This integrative view of metabolism–epigenetics interplay complements established pathways in aging biology.

However, inevitably, there are also some shortcomings that should be pointed out. First, as a central metabolic intermediate, lactate has pleiotropic functions and could influence senescence through mechanisms beyond histone lactylation, such as modulating cellular redox balance or acting through other signaling pathways. Our study was specifically designed to dissect the novel histone lactylation-mediated pathway, and the potential contributions of these alternative lactate-driven mechanisms warrant further investigation. Second, our study focused on early gestation; future work is needed to investigate how histone lactylation and its associated molecular networks may change across different phases of the menstrual cycle in patients with URSA. Third, the scope of the current investigation was limited to the lactylation of histone H3, and it would be intriguing to further explore the significance of other histones as well as nonhistone proteins in URSA. Finally, given the central role of decidualization deficiency in other major pregnancy complications, such as preeclampsia and preterm birth, it will be of great interest to investigate the pathogenic influence of the HK2–H3K18la–CUX1 axis in these disorders.

Conclusions

This study revealed a novel molecular mechanism in which HK2-dependent glycolysis drives the H3K18la–CUX1 axis to modulate SASP components (CXCL1, CXCL3, and TNF-α), thus promoting stromal cell senescence in URSA. This work yields data for a better comprehension of the interactions among energy metabolic reprogramming, epigenetic modification, and cellular senescence, revealing a novel mechanism underlying URSA. Targeting this metabolism–epigenetic signaling cascade may represent a promising therapeutic approach for the development of drug interventions against stromal cell senescence and the treatment of decidualization deficiency in URSA.

Abbreviations

8-Br-cAMP

8-Bromoadenosine 3′:5′-cyclic monophosphate

AAV

Adeno-associated virus

ChIP

Chromatin immunoprecipitation

CUT&Tag

Cleavage under targets and tagmentation

CUX1

Cut-like homeobox 1

DSC

Decidua stromal cell

ESC

Endometrial stromal cell

H3K18la

Lactylation of H3K18

HK2

Hexokinase 2

MPA

Medroxyprogesterone acetate

RSA

Recurrent spontaneous abortion

SASP

Senescence-associated secretory phenotype

T-hESCs

Telomerase-immortalized human endometrial stromal cells

URSA

Unexplained recurrent spontaneous abortion

Author contributions

X.X.Z., H.L.Z., and X.L.F. performed the experiments; J.M. and X.L.F. conceived and designed the experiments; Y.P.J. performed bioinformatic analysis; H.L.Z. and Y.M.M. analyzed the data. X.X.Z., Y.Z., and Y.P.J. wrote the paper.

Funding

This research was funded by National Natural Science Foundation of China (no. 82305294), Zhejiang Provincial Natural Science Foundation of China (no. LQ24H270019), Zhejiang Traditional Medicine and Technology Program, China (no. 2025ZR173), Zhejiang Medical and Health Project (no. 2025KY158), Research Project of Zhejiang Chinese Medical University (no. 2024JKZKTS37) to X.X.Z. and National Natural Science Foundation of China (nos. 82174421 and 81973894) to X.L.F.

Data availability

No datasets were generated or analyzed during the current study.

Declarations

Ethics approval and consent to participate

The project was conducted in compliance with the guidelines outlined in the Declaration of Helsinki and the Basel Declaration. Studies involving human tissues were approved by the Ethics Committee of Hangzhou Hospital of TCM affiliated to ZCMU (Hangzhou, China) (approval no. 2024KLL119, date of approval: 3 May 2024). The study was performed in accordance with the principles of the Declaration of Helsinki. Written informed consent was obtained from all participants for the use of their clinical information. The animal experiments were performed in accordance with the Basel Declaration and were approved by the Institutional Animal Care and Use Committee, Zhejiang Chinese Medical University (approval no. IACUC-20230904–18, date of approval: 4 September 2023). The ethics committee follows the guidelines of the International Council for Laboratory Animal Science (ICLAS) to ensure the ethical compliance of the experiments.

Consent for publication

This manuscript is approved by all authors for publication.

Competing interests

The authors declare no competing interests.