This study recruited 20 patients with RSA and 20 healthy control participants. Inclusion criteria for the RSA group: Females aged 20–45 years who meet the criteria for an RSA diagnosis, with a current intrauterine pregnancy confirmed via ultrasound. Both partners must provide informed consent and sign the consent form. Exclusion criteria included known causes of miscarriage, anatomical abnormalities of the reproductive tract, uncontrolled endocrine or metabolic disorders, autoimmune diseases, prothrombotic states, recent acute infections and other major systemic diseases. The control group comprised healthy women in early pregnancy who had voluntarily terminated their pregnancy for non-medical reasons (the normal early pregnancy abortion group). The gestational age of this group was matched with that of the RSA group, and all the RSA group’s exclusion criteria were applied. All participants provided informed consent and signed informed consent forms.

This study involved collecting products of conception tissue and peripheral blood.

On the day of the ultrasound-guided abortion or surgical procedure, the tissue was obtained using sterile instruments and immediately divided into two portions. One portion was placed in 4% paraformaldehyde for fixation, followed by paraffin embedding, histological sectioning and immunohistochemical analysis. The other portion was placed in a sterile, enzyme-free cryopreservation tube, rapidly frozen in liquid nitrogen and transferred to –80 °C storage for subsequent molecular biology experiments, such as protein/RNA extraction, Western blot and quantitative real-time polymerase chain reaction (qPCR). Peripheral blood: Collect 5–10 mL of fasting venous blood from patients in the morning using EDTA anticoagulant tubes (Bacet, Wenzhou, Zhejiang, China). Centrifuge at 3000 rpm for 15 minutes at 4 °C. Aspirate the plasma from the supernatant, divide it into smaller portions, and store it at –80 °C. All sample tubes were labelled with unique, traceable codes (concealing patient names) and a sample information database was created. This database recorded the following information for each sample: ID, patient group, age, parity, number of miscarriages, gestational age, collection date, processing method and storage location. Biological samples were stored in –80 °C freezers or liquid nitrogen tanks, and storage equipment operation was regularly checked.

A total of forty-five 6-week-old mice, comprising thirty CBA/J, five DBA/2, ten BALB/c, were purchased from GemPharmatech™ (Nanjing, Jiangsu, China) and housed in stainless steel cages with ad libitum access to food and water. They were kept under controlled conditions of temperature (22 $\pm$ 2 °C) and humidity (50 $\pm$ 5%) with a light/dark cycle of 12/12 h. To generate a normal pregnancy model, female CBA/J mice were mated with male BALB/c mice at a 2:1 ratio. For the RSA model, female CBA/J mice were paired with male DBA/2 mice at a 2:1 ratio. The next morning, the presence of a vaginal plug was used to mark the start of gestational day 0 (G0). On G5, normal pregnant mice received an injection of either RU486 (5 mg/kg; Aladdin, Shanghai, China) or an equivalent volume of phosphate-buffered saline (PBS). All mice were humanely euthanised on G13 under sustained isoflurane anaesthesia (3.0–4.0%) by trained personnel using cervical dislocation. Pregnancy outcomes were assessed by examining the uterine horns. Necrosis and haemorrhage of the foetuses and placentas were evaluated by macroscopic inspection. The miscarriage rate was determined by dividing the number of resorption sites by the total number of implantation sites.

The EL4 cells were obtained from the Cell Bank of the Chinese Academy of Sciences (TCM41, Shanghai, China). The cells were cultured in Dulbecco’s Modified Eagle Medium (DMEM) supplemented with 10% fetal bovine serum (FBS) (Gibco, Waltham, MA, USA) at 37 °C in a humidified atmosphere containing 5% CO₂. The medium was refreshed every three days, and the cells were subcultured when they reached 80% confluence. Mycoplasma testing was negative. The EL4 cells have been authenticated by STR profiling.

The uterine tissue segments were fixed in 4% paraformaldehyde, dehydrated using a graded ethanol series and embedded in paraffin wax Thin sections were stained with haematoxylin and eosin (H&E) and examined using the Zeiss light microscope (Oberkochen, Germany). For immunohistochemistry, the placental sections were deparaffinised and treated with 3% hydrogen peroxide to block endogenous peroxidase. They were then incubated with goat serum to minimise non-specific binding. An overnight incubation of the slides was performed at 4 °C with anti-TLR4 antibody (1:500, 19811-1-AP; proteintech, Wuhan, Hubei, China), anti-p65 antibody (1:200, 3033; Cell Signaling Technology, Danvers, MA, USA), and anti-FOXP3 antibody (1:500, 22228-1-AP; proteintech, Wuhan, Hubei, China). After three washes in 0.01 M PBS, the samples were incubated with horseradish peroxidase (HRP)-conjugated goat anti-rabbit IgG (1:5000, TA373083L; OriGene, Beijing, China) for 30 min at 37 °C. Immunostaining was visualised using 3,3^′-diaminobenzidine (DAB) and the sections were counterstained with Mayer’s haematoxylin (ZSBG-BIO, Beijing, China). The sections were then dehydrated, cleared and mounted. Immunohistochemical staining for TLR4, p65 and FOXP3 was quantified using a standardised H-scoring system [19].

The frequency of Tregs was determined using flow cytometry to analyse single-cell suspensions from mouse spleens and human peripheral blood. Spleens were harvested from each group and processed into single-cell suspensions using a previously established protocol [20]. Human peripheral blood mononuclear cells were isolated using Ficoll density gradient centrifugation. To quantify the CD4+CD25+FOXP3+ T cell population, mouse and human Tregs staining kits (88-8111-40 and 88-8999-40, respectively, eBioscience, San Diego, CA, USA) were employed in accordance with the manufacturers’ instructions. Data are presented as the percentage of Tregs within the total cluster of differentiation 4 (CD4)+ T cell population [21].

Total protein was extracted from the abortion products of patients with RSA, and the protein concentration was determined by the bicinchoninic acid (BCA) method. Immunoblotting was performed using standard methods. TLR4 antibody (1:500, 19811-1-AP; proteintech, Wuhan, Hubei, China), TLR4 (1:2000, ab13556; Abcam, Cambridge, UK), p65 (1:1000, 3033; Cell Signaling Technology, Danvers, MA, USA), and FOXP3 (1:2000, ab215206; Abcam, Cambridge, UK) were used to detect the proteins. GAPDH (1:50,000, 60004-1-Ig, proteintech, Wuhan, Hubei, China) was used as an internal control to normalise protein loading. Anti-rabbit (1:5000) secondary antibodies were conjugated with HRP (P6782; Sigma-Aldrich, Shanghai, China). HRP-labeled Goat Anti-Mouse IgG (H + L) (1:2000, A0216, Beyotime, Shanghai, China) were used. Following separation, the membranes were trimmed according to the migration of a colorimetric protein marker. Immunoreactive bands were detected using an enhanced chemiluminescence system (Merck, Darmstadt, Germany) and quantified using densitometry with ImageJ software (Version 1.54; National Institutes of Health, Bethesda, MD, USA).

Total RNA was extracted from the abortion products of patients with RSA using Trizol reagent (15596026; Invitrogen, Carlsbad, CA, USA) for qPCR analysis. The expression of three target genes (TLR4, p65, FOXP3) and the housekeeping gene ($\beta$-actin) was assessed using 20 µL reactions containing 10 µL of Real-Time PCR SYBR Green Master Mix (RR036A; Takara, Kyoto, Japan), 0.3 µL of each primer and 7.4 µL of RNase-free water. Amplification was performed on a Cobas z480 System (Roche, Rochester, NY, USA) under the following conditions: Initially, the samples were heated for 30 seconds at 94 °C, followed by 40 cycles of 5 seconds at 94 °C and 10 seconds at 60 °C, then 5 seconds at 95 °C, 1 minute at 60 °C and 30 seconds at 50 °C. To assess the mRNA transcript levels, the expression of target genes was normalised to the housekeeping gene $\beta$-actin. The relative mRNA expression was calculated using the 2^{-Δ⁢Δ⁢Ct} method. The Ct values of the target genes were normalized to the $\beta$-actin ($\mathrm{\Delta }$$\mathrm{\Delta }$Ct). The $\mathrm{\Delta }$$\mathrm{\Delta }$Ct value was then determined by comparing the $\mathrm{\Delta }$Ct of each sample to the mean $\mathrm{\Delta }$Ct of the calibrator group. All qPCR reactions were conducted in triplicate. The sequences for all primers used in this study are listed in Supplementary Table 1.

Small interfering RNAs (siRNAs) were synthesised by Genepharma (Shanghai, China). A universal control siRNA was used as a non-specific control. EL4 cells were transfected with the siRNA duplex using Lipofectamine 2000 (Invitrogen) according to the manufacturer’s instructions. The p65 genes were amplified and cloned into a pcDNA3.1 vector (Novagen, Madison, WI, USA). Transfection of EL4 cells was achieved using the Amaxa Nucleofector system (Thermo Fisher Scientific, Waltham, MA, USA).

Statistical analyses were performed using GraphPad Prism 8 software (GraphPad Software, Inc., San Diego, CA, USA). Student’s t-test was used for intergroup comparisons, and one-way ANOVA was used for multiple comparisons. Tukey’s post-hoc test was then used to identify specific differences between groups. A p-value of 0.05 was considered statistically significant.

To evaluate the divergence in gestational outcomes in RSA mice, both normal control and RSA pregnancy models were established, as previously described [22]. In the normal pregnancy group, mice received intraperitoneal injections of either PBS or RU486 during early gestation (G5). All mice were euthanised at G13. In the normal pregnancy model, administration of RU486 induced an abortion rate of 97.53%, which was markedly higher than the 6.25% observed in the PBS-treated control group, confirming the efficacy of RU486 in establishing an early pregnancy loss model. However, the miscarriage rate in the RSA pregnant mouse model was 31.64% (Table 1). The RSA group experienced fetal and placental necrosis and haemorrhage, whereas control mice had almost no obvious fetal abnormalities (Fig. 1A). H&E staining was performed on uterine sections from the control, RU486-treated, and RSA model groups (Fig. 1B). The maternal–fetal interface in the RSA mouse model displayed pronounced haemorrhagic changes and inflammatory characteristics. This result confirmed that the RSA model was successfully constructed.

Fig. 1.

Reduced Treg cell expression in RSA mice. (A) Representative uterine specimens were collected from both normal and RSA-affected pregnancies at G13. (B) Histopathological evaluation using H&E staining was performed on maternal–fetal interface sections from the corresponding experimental groups. The placental structure was clearly defined in the RSA group, but disorganised in the RU486 group. Black arrows mark loose oedema, yellow arrows indicate necrotic cells, red arrows indicate bleeding. 2$\times$ Scale bar: 500 µm; 4$\times$ Scale bar: 200 µm; 10$\times$ Scale bar: 100 µm; 20$\times$ Scale bar: 50 µm. (C) Flow cytometry to detect the proportion of CD4+CD25+FOXP3+ T cells. H&E, haematoxylin and eosin.

Given the established role of CD4+CD25+FOXP3+ T cells in maternal–fetal immune tolerance, their prevalence among splenocytes was quantified using flow cytometry. The proportion of these Tregs was significantly reduced in both the RSA and RU486 groups. However, it was still higher than that of the non-pregnant group (Fig. 1C). The statistical graph for Fig. 1C is shown in Supplementary Fig. 2. These findings suggest that impaired expansion or maintenance of CD4+CD25+FOXP3+ T cells is associated with pregnancy loss in the RSA model.

FOXP3, a pivotal transcription factor governing the development and function of Tregs, plays a critical role in diverse immunological processes including immune homeostasis, autoimmune pathogenesis, infectious responses, and tumour immune evasion [23]. The expression of TLR4, p65, and FOXP3 in placenta tissue was assessed using Western blotting and immunohistochemistry. As shown in Fig. 2 (p 0.05), the RSA model exhibited increased expression levels of TLR4 and p65, while FOXP3 expression decreased. In the RU486 model, however, only a decrease in FOXP3 expression was observed. These data indicate that a reduction in Tregs is associated with TLR4/p65 activation in the RSA mouse model.

Fig. 2.

Regulation of the TLR4/p65/FOXP3 signalling pathway in RSA mice. (A) The protein expression levels of TLR4, p65, and FOXP3 in placental tissue were assessed by Western blotting. (B) TLR4, p65, and FOXP3 was detected by immunohistochemistry. Scale bar: 50 µm. *: Compared with the control group, p 0.05; **: Compared with the control group, p 0.01; ***: Compared with the control group, p 0.001; #: Compared with the control group, p $>$ 0.05. TLR4, Toll-like receptor 4; FOXP3, forkhead box protein P3.

Activation of the TLR4 signalling pathway significantly contributes to the downregulation of FOXP3 expression [24]. To verify the link between RSA and Treg reduction, Treg levels in the peripheral blood of patients with RSA were assessed using flow cytometry. Compared with the control group, the number of Tregs in patients with RSA was reduced (Fig. 3A and Supplementary Fig. 3, p 0.001). We used RT-PCR to further validate that TLR4 and p65 were highly expressed, whereas there was low expression of FOXP3 in the abortion products of patients with RSA (Fig. 3B, * for p 0.05, ** for p 0.01). Western blot analysis of TLR4, p65, and FOXP3 protein expression yielded the same results (Fig. 3C, p 0.05). These findings suggest that Treg reduction in patients with RSA is associated with TLR4/p65 activation.

Fig. 3.

Expression levels of TLR4, p65, and FOXP3 in patients with RSA. (A) Percentage of CD4+CD25+FOXP3+ T cells in peripheral blood was assessed using flow cytometry. (B) qPCR was performed to measure TLR4, p65, and FOXP3 transcript levels in decidual tissue. (C) Protein expression of TLR4, p65, and FOXP3 were detected by Western blotting. *: compared with the normal decidua, p 0.05; **: Compared with the normal decidua, p 0.01; ***: Compared with the normal peripheral blood, p 0.001.

To investigate the underlying molecular mechanisms, we employed the EL4 cell line, a model characterised by strong inducible FOXP3 expression upon stimulation, while retaining numerous intrinsic T-cell properties [25]. To assess the involvement of p65 in RSA-induced FOXP3 downregulation, we initially stimulated the TLR4/p65 pathway using the agonist LPS, and subsequently inhibited p65 expression via siRNA targeting p65 (si-p65), with a non-targeting control siRNA (si-NC) serving as the negative control. Moreover, the TLR4 inhibitor IAXO102 was chosen to inhibit the TLR4/p65 signalling pathway, followed by overexpression of p65 with pcDNA3.1-p65, the corresponding empty vector control (pcDNA-NC) was used as the negative control. Compared to the si-NC group, transfection with si-p65 significantly reduced p65 protein expression (Fig. 4A, p 0.05). Conversely, transfection with the pcDNA3.1-p65 plasmid substantially increased p65 protein expression compared to the pcDNA-NC group (Fig. 4B, p 0.05). These results verified the efficiency of p65 knockdown and overexpression. The results also demonstrated that LPS stimulation of EL4 cells significantly increased TLR4 and p65 expression while downregulating FOXP3. Furthermore, knockdown of p65 prevented LPS-induced suppression of FOXP3 expression (Fig. 4A, p 0.05). Pharmacological inhibition of TLR4 using IAXO102 inhibited TLR4 and p65 expression and significantly upregulated FOXP3. However, this enhanced FOXP3 expression induced by IAXO102 was inhibited by pcDNA3.1-p65 (Fig. 4B, p 0.05). Taken together, these results indicate that p65 is critical for the TLR4-mediated regulation of FOXP3 in EL4 cells.

Fig. 4.

Downregulation of FOXP3 by TLR4 signalling is p65-dependent. (A) EL4 cells were transfected with p65-target siRNA or control siRNA with or without LPS (TLR4 agonist). Western blotting showed TLR4, p65, and FOXP3 protein expression. *p 0.05 compared to the si-NC group. $p 0.05 compared to the si-p65 group. #p $>$ 0.05 compared to the si-p65 group. (B) Cells were transfected with pcDNA3.1-p65 or empty vector with or without IAXO102 (TLR4 inhibitor). *p 0.05 compared to the pcDNA3.1-NC group. $p 0.05 compared to the pcDNA3.1-p65 group. #p $>$ 0.05 compared to the pcDNA3.1-p65 group.