Introduction

In recent years, low fertility is a global problem (Agbaglo et al., 2022), exacerbated by pregnancy complications. Preeclampsia (PE) is a progressive systemic disease during pregnancy, with pregnancy-induced hypertension, proteinuria, and liver and kidney injury as the main diagnostic indicators (Chappell et al., 2021). With a global prevalence of 7–10%, PE is a major cause of maternal and perinatal mortality and morbidity, especially in low-income and middle-income countries. The only specific treatment is delivery (Chappell et al., 2021). However, due to the variety of symptoms and the heterogeneity of the disease, the pathogenesis of PE is still unclear (Burton et al., 2019; Grotegut, 2016).

PE is often complicated by gestational diabetes mellitus (GDM), which is characterized by insulin resistance and associated with aberrant maternal immune cell adaption (Corrêa-Silva et al., 2018; McElwain et al., 2021; McIntyre et al., 2019). Studies have shown that GDM is an independent risk factor for preeclampsia (Nerenberg et al., 2013; Ostlund et al., 2004; Schneider et al., 2012). However, whether a different pathogenesis underlies between PE and GDM is still unclear.

The fetus is equivalent to a semi-homogenous graft for the mother. Therefore, maternal-fetal immune tolerance plays an important role in the maintenance of pregnancy. Elevated levels of maternal inflammation are now considered to be an important mechanism in the pathogenesis of preeclampsia (Aneman et al., 2020; Deer et al., 2023; Jung et al., 2022), including macrophages (Faas et al., 2014; Yao et al., 2019), granulocytes (Lampé et al., 2015; Lampé et al., 2011; D. Liu et al., 2021), natural killer (NK) cells (Bachmayer et al., 2006; Fukui et al., 2011; Travis et al., 2020), innate B1 cells (Jensen et al., 2012; LaMarca et al., 2011; Zhong et al., 2007) and γδ T cells (Chatterjee et al., 2017). Moreover, adaptive immune response is also critical for the pathogenesis of PE (Deer et al., 2023). In PE, CD4⁺ T cells prefer to differentiate into pro-inflammatory Th1 and Th17 phenotypes (Eghbal-Fard et al., 2019; Fu et al., 2014; Lang et al., 2021; Lu et al., 2020); Saito et al. (2007) instead of non-inflammatory Th2 and Treg phenotypes (Care et al., 2018; Cornelius et al., 2015a; Saito et al., 2007; Santner-Nanan et al., 2009; Sasaki et al., 2007). However, a comprehensive and in-depth understanding of the maternal-fetal interface of PE is still lacking.

The immune cells crosstalk with each other at the maternal-fetal interface elaborately. The interaction between B2 cells and Th cells depends on CD40 and its ligand CD40L, which helps B2 cells provide AT1-AAs and differentiate into memory B cells (Cornelius et al., 2015b). The cytokines derived from NK cells can promote the differentiation of Th17 to co-work in recurrent miscarriage (Fu et al., 2013). Nevertheless, studies about the regulation of the immune network at the maternal-fetal interface of PE still need to be completed.

In this study, we combined cytometry by time of flight (CyTOF), single-cell RNA sequencing (scRNA-seq), and rodent experiment to identify the overall immune cell composition and their interactions at the maternal-fetal interface in PE, GDM, and GDM complicated with PE (GDM&PE). This study revealed the PE-specific immune cell network, which was regulated by pro-inflammatory macrophages, providing new ideas about the pathogenesis of PE.

Results

Overall immune cell profile in the placenta of PE, GDM, and GDM&PE

To fully characterize the immune microenvironment at the maternal-fetal interface of PE, we collected full-term placentas of PE and normal pregnancy (NP), and did a CyTOF test (Table 1), which panel was shown in Table 2. Since PE is usually complicated by GDM, the subjects of GDM and GDM complicated with PE (GDM&PE) were also included. The experimental workflow for CyTOF is shown in the schematic diagram (Figure 1A). We found that in addition to the well-known macrophages and T cells, there are also γδ T cells, B cells, NK cells, granulocytes, dendritic cells (DCs), and myeloid-derived suppressor cells (MDSCs). An overall distribution of CD45⁺ cell subsets in placentas was shown in the t-SNE maps (Figure 1B). There were no significant differences in the proportion of these large subpopulations comparing the PE, GDM, GDM&PE group to the NP group (Figure 1C). To identify the ten cell subpopulations, CD4, CD8, and γδTCR were used as markers to distinguish CD4⁺ T cells, CD8⁺ T cells, and γδ T cells. CD11b is a marker that distinguishes myeloid cells, including macrophages and granulocytes (Figure 1D). The ten cell subsets were further annotated by 41 markers in the heatmap (Figure 1E, Table 3).

Figure 1

Table 1

Table 2

Table 3

Specific altered T cell profile in the placenta of PE

To fully understand the distribution in each cell subset, we first analyzed 15 clusters of CD4⁺ T cells from the placentas of NP, PE, GDM, and GDM&PE (Figure 2A). CD4⁺ T cell clusters were defined by canonical marker set signatures (Figure 2-figure supplement 1 A). The frequencies of memory-like CD45RA^-CCR7^midCD28⁺CD127⁺CD4⁺ T cells (cluster 8) were significantly increased in individuals with PE compared with NP controls, while. At the same time, there was no significance in the GDM and GDM&PE groups compared with the NP group (Figure 2B, Table 4). We analyzed the expression of common intracellular molecules in CD4⁺ memory-like T cells by sorting CD45RO⁺CCR7⁺CD4⁺ T cells from placental samples by flow cytometry. Higher levels of IL-17A and lower levels of Foxp3 were found in memory-like CD45RO⁺CCR7⁺CD4⁺ T cells in the placentas of individuals with PE (Figure 2C). Consistent with the results of CyTOF, the higher fluorescence intensity of CD45RO⁺CD4⁺ T cells was found in the placentas of individuals with PE (Figure 2D). Moreover, lower expression of immune checkpoint molecules including T-cell immunoglobulin mucin-3 (Tim-3) and programmed cell death 1 (PD-1) was found in CD45RA^-CCR7⁺CD4⁺ memory T cells in the PE group, suggesting that these cells have a lower immunosuppressive capacity (Chen et al., 2022; Fanelli et al., 2021; Rasmussen et al., 2022; Wang et al., 2016), though there is no significant difference in the GDM or GDM&PE groups comparing with the NP group (Figure 2E). Then, 18 clusters of CD8⁺ T cells were analyzed (Figure 2F). CD8⁺ T cell clusters annotation was based on the expression of canonical marker signatures (Figure 2-figure supplement 1 B). CD45RA^-CCR7⁺CD38⁺pAKT^midCD127^low (cluster 2) memory-like CD8⁺ T cells were were significantly increased in the PE group (Figure 2G, Table 4). Cluster 2 expressed lower expression of PD-1 and TIGIT, suggesting the activation of cytotoxicity of these cells (Figure 2H) (Morita et al., 2020). Then, 22 clusters of γδT cells were analyzed and annotated (Figure 2I, Figure 2-figure supplement 1 C). Cluster 15, which expressed lower expression of Tim3, was significantly decreased in the PE group (Figure 2J, 2K). However, regardless of CD4⁺ T cells or CD8⁺ T cells, there is no significant change in the memory-like clusters in the GDM or GDM&PE group compared to the NP group.

Figure 2

Table 4

In conclusion, significant changes in placental T cell profile were found in the placentas of PE but not in those of GDM or GDM&PE, suggesting that abnormal activation of T cells in the placenta is associated with the pathogenesis of PE.

Abnormal polarization of macrophages was correlated with specific immune cell subsets in individuals with PE

Except for T cells, we also analyzed 29 clusters of CD45⁺CD3^-CD11b⁺ cells from placentas of NP, PE, GDM, and GDM&PE, including 9 clusters of macrophages, 11 clusters of granulocytes, and 5 clusters of NK/NK-like cells (Figure 3A). The clusters of CD45⁺CD3^-CD11b⁺ cells were defined by canonical marker set signatures (Figure 3 B). Moreover, significantly decreased frequencies of CD11b⁺CD15^hiHLA-DR^low granulocytes (cluster 12), which are identified as granulocyte myeloid-derived suppressor cells (gMDSCs), were found in the PE group (Figure 3C, Table 4). In addition, the frequency of anti-inflammatory macrophages (anti-inflam Macs) (CD206⁺CD163^-CD86^midCD33⁺HLA-DR⁺, cluster 25) was also significantly decreased, whereas the frequency of pro-inflammatory macrophages (pro-inflam Macs) (CD206^-CD163^-CD38^midCD107a^lowCD86^midHLA-DR^midCD14⁺, cluster 23) was significantly increased in the PE group (Figure 3C, Table 4). However, the frequencies of macrophages and gMDSCs were unchanged significantly in the GDM or GDM&PE groups.

Figure 3

Pearson correlation analysis indicated positive correlations between pro-inflam Macs (cluster 23 in CD11b⁺ cells) and CD4⁺ memory-like T cells (cluster 8 of CD4⁺ T cells), as well as CD8⁺ memory-like T cells (cluster 2 in CD8⁺ T cells). CD206^- pro-inflam Macs were negatively correlated with gMDSCs (cluster 12 in CD11b⁺ cells), but were not statistically significant. Conversely, CD206⁺ anti-inflam Macs (cluster 25 in CD11b⁺ cells) were positively correlated with gMDSCs and negatively correlated with CD8⁺ memory-like T cells (Figure 3D, 3E).

These results suggested that abnormally polarized CD206^—pro-inflam Macs are positively correlated with memory-like CD4⁺ and CD8⁺ T cells in the placentas of individuals with PE and negatively correlated with gMDSCs.

F480⁺CD206^- pro-inflam Macs induced immune imbalance at the maternal-fetal interface and PE-like symptoms

Though increased pro-inflam Macs have been reported in the placenta from individuals with PE (Faas et al., 2014), few studies have reported the interaction between macrophages and other immune cells in the placenta. CD45⁺F4/80⁺CD206^- pro-inflam Macs or CD45⁺F4/80⁺CD206⁺ anti-inflam Macs were isolated from the uterus and placentas of mice with RUPP and injected into normal pregnant mice (Figure 4A). RNA-seq was used to analyse the difference between CD45⁺F4/80⁺CD206^- pro-inflam Macs and CD45⁺F4/80⁺CD206⁺ anti-inflam Macs. Significant differences were found between the two groups of macrophages (Figure 4B). We found that the expression of 2123 genes was increased, and 2699 genes decreased significantly in the pro-inflam Macs compared to the anti-inflam Macs (Figure 4C). The expression levels of genes associated with inflammatory response (Ccl7, Ccl8, Ccl2, IL6) (He et al., 2019; Wu et al., 2023), complement system activation (C1qa, C1qb) (Chen et al., 2021), and lipid metabolism (Pltp, Apoe) (Desrumaux et al., 2016) were significantly increased in the pro-inflam Macs; whereas the expression levels of genes associated with the regulation of angiogenesis and vascular endothelial growth factor production (VEGFα, Macro) and tissue development (TGF-β1, Thbs1, Fn1, Slpi) (Jin et al., 2023; Nugteren et al., 2021; Li et al., 2022) were significantly decreased (Figure 4C, 4D). Moreover, Gene Set Enrichment Analysis (GSEA) was performed to further explore the most significantly enriched functional terms between the two groups of macrophages. We found that NF-κB signaling and interspecies interaction between organisms were enriched in the pro-inflam Macs, while tissue development and epithelial cell morphogenesis were enriched in the anti-inflam Macs (Figure 4E).

Figure 4

To further investigate the effect of the pro-inflam Macs polarization in immune imbalance at the maternal-fetal interface, PLX3397, the inhibitor of CSF1R, which is needed for macrophage development, was used to deplete the macrophages of pregnant mice (X. Chen et al., 2023;J. Chen et al., 2022). As expected, an increased embryo resorption rate, decreased fetal top-rump length and fetal weight were found in mice injected with pro-inflam Macs (Figure 4F). Increased SBP and UACR were also found in mice injected with pro-inflam Macs (Figure 4G). Moreover, it is shown that CD44⁺ memory-like Th17 cells and memory-like CD8⁺ T cells increased while CD11b⁺Ly6G⁺ gMDSCs decreased in mice injected pro-inflam Macs compared with mice injected anti-inflam Macs, which validated our findings in CyTOF (Figure 4H). Clodronate liposomes were also used to deplete the macrophages of pregnant mice (X. Liu et al., 2022) before pro-inflam Macs or anti-inflam Macs were injected into the mice. The same experimental results were obtained (Figure 4-figure supplement 1). In conclusion, the F480⁺CD206^- pro-inflam Macs induced immune imbalance at the maternal-fetal interface and PE-like symptoms.

Pro-inflam Macs and anti-inflam Macs subsets are functionally heterogeneous

To further explore the role that macrophages play in the immune imbalance at the maternal-fetal interface in PE, scRNA-seq was performed to analyze the uterine CD45⁺ immune cells from mice that were transferred with pro-inflam Macs or anti-inflam Macs from the the uterus and placentas of PE mice. An unsupervised cluster detection algorithm (SEURAT) was applied and eight types of immune cells were detected by mostly distinguishable cell type-specific genes, including macrophages, monocytes, granulocytes, T/NK cells, B cells, plasma, mast cells, and basophils (Figure 5A, 5B).

Figure 5

Macrophages were further identified into 15 clusters defined by marker set signatures. Single-cell differential expression analysis (SCDE) was performed for each population and characteristic gene expression patterns were detected for cluster 0-cluster 14 to characterize the phenotypes of these subsets in detail (Figure 5C). We found that the frequency of cluster 0 was significantly increased in mice injected with pro-inflam Macs, while the frequency of cluster 1 was significantly decreased (Figure 5D). We also found that cluster 0 highly expresses genes associated with foetal and tissue resident (Folr2) (Nalio et al., 2022; Thomas et al., 2021), complement system activation (C1qa, C1qb, C1qc) (Chen et al., 2021), and inflammatory response (Ccl7, Ccl8) (He et al., 2019; Wu et al., 2023); while cluster 1 highly expresses genes associated with tissue repair (Chil3, Slpi, Fn1) (Jin et al., 2023; Nugteren et al., 2021; Li et al., 2022), blood vessel morphogenesis (Thbs1) (Che et al., 2021) and preventing oxidative stress (Gsr, Mgst1) (Coppo et al., 2022) (Figure 5E). Changes in gene expression patterns of cluster 0 (pro-inflam Macs) and cluster 1 (anti-inflam Macs) were analyzed in mice injeced the pro-inflam Macs or anti-inflam Macs. We found that in the pro-inflam Macs group, enriched GO terms in cluster 0 and cluster1 including ‘antigen processing and presentation of exogenous antigen’ and ‘inflammatory response’, in which high expression of genes such as CCL5, CCL8, CALR, IRF7, IL10, IFI44, IFI30, OAS3, etc. could be observed (Figure 5F, 5G). The control group showed elevated expression of MMP14, Chil3, EGR1, ATF3 and VASP; and these genes were enriched in ‘vascular endothelial growth factor production’ and ‘tissue development’ (Figure 5F, 5G). These results were consistent with the results of the transcriptome RNA sequencing results of macrophages.

These data suggested that the CD45⁺F480⁺CD206^- pro-inflam Macs with a Folr2⁺Ccl7⁺Ccl8⁺C1qa⁺C1qb⁺C1qc⁺ phenotype play an improtant role in the development of PE.

Pro-inflam Macs induced the memory-like Th17 cells, which was associatied with the development and recurrence of PE

We also analyzed the scRNA-seq data of uterine T/NK cells from mice that were injected with or without pro-inflam Macs. T/NK cells were further identified into 12 clusters defined by marker set signatures (Figure 6A, 6B). We found that the frequency of cluster 0 was increased in mice injected with pro-inflam Macs, while the frequency of cluster 1 and cluster 2 were significantly decreased (Figure 6A). In addition to genes associated with Th17 cells (IL-17a, IL17f, Rora, Il1r1, TNF) (Wu et al., 2019; Leite et al., 2023), cluster 0 also exhibits high expression of genes associated with memory phenotype (Cxcr6, S100a4, CD44) (Schroeder et al., 2023; Bieberich et al., 2021), suggesting that cluster 0 was the memory-like Th17 cells (Figure 6A, 6C). Cluster 1 exhibits high expression of genes associated with immunoregulation (Lef1, Tcf7, Ccr7) (Qiu et al., 2022; Sekine et al., 2020); Cluster 2 exhibits high expression of genes associated with immunosuppression (Gata3, GITR, CD28), suggesting that cluster 1 and 2 were T cells with immunosuppressive function (Esensten et al., 2016; Pai et al., 2023) (Figure 6A, 6C).

Figure 6

Changes in gene expression patterns of cluster 0 (memory-like Th17 cells) were analyzed in the pro-inflam Macs group and Control group. We found that in the pro-inflam Macs group, enriched GO terms including ‘positive regulation of response to external stimulus’ and ‘tumor necrosis factor production’, in which highly express genes such as TNFSF11, TNF, IL27RA, IGF1R, CD226, LAMP1 (Figure 6C, 6D). The control group showed elevated expression of ASS1, EIF5, S100A9, CTLA4, S100A8 and CXCR4; these genes were enriched in GO terms including ‘regulation of programmed cell death’ and ‘cellular biosynthetic process’ (Figure 6D, 6E).

Combining the CyTOF and sc-RNA seq data, we found the frequency of memory-like IL-17A⁺CD4⁺ T cells significantly increased at the maternal-fetal interface of individuals with PE. To confirm the effect of memory-like Th17 cells in PE, we constructed an animal model of PE by reducing uterine perfusion pressure (RUPP), and mice with sham operation were considered as controls (Figure 6-figure supplement 1A). We found that mice of PE showed an increased embryo absorption rate, decreased fetal weight, and pup crown-rump length compared with the sham operation group (Figure 6-figure supplement 1B). Increased systolic blood pressure (SBP) and urine albumin creatine ratio (UACR) were also observed in mice with RUPP, which indicated a successful PE mice mode was built (Figure 6-figure supplement 1C). Our studies revealed that the frequency of memory-like Th17 cells (CD4⁺CD44⁺IL-17A⁺ T cells) was significantly increased in the PE mice (Figure 6F).

To confirm the importance of memory-like CD4⁺ T cells in the pathogenesis of PE, the CD4⁺CD44⁺ T cells in the uterus and placentas from PE or NP mice were sorted and intravenously injected the cells into normal pregnant mice on day12.5 of gestation (Figure 6F). Increased embryo absorption rate, decreased fetal weight and pup crown-rump length were found in mice injected with PE mouse-derived CD4⁺CD44⁺ T cells (Figure 6G). The typical PE-like symptoms, including increased SBP and UACR, were also observed in those mice (Figure 6H).

It has been reported that sustained expansion of immunosuppressive memory Tregs during prior pregnancy is beneficial for maintaining a second pregnancy (Rowe et al., 2012). To verify whether memory-like Th17 cells promote the recurrence of PE, we established a second pregnant mouse model with a history of PE or NP in the first pregnancy. An increased embryo resorption rate, decreased crown-rump length, and fetal weight were found in mice with a history of PE pregnancy compared with those with a history of NP pregnancy (Figure 6—figure supplement 1D). Moreover, mice with a history of PE in the first pregnancy showed increased SBP and UACR during the second pregnancy (Figure 6—figure supplement 1E). Consistently, higher levels of IL-17A were also found in memory-like CD4⁺ T cells in mice with a history of PE in the first pregnancy (Figure 6—figure supplement 1F).

Therefore, pro-inflam Macs induced the IL-17a⁺IL17f⁺Rora⁺Il1r1⁺TNF⁺Cxcr6⁺S100a4⁺ CD44⁺ memory-like Th17 cells, which was associatied with the development and recurrence of PE.

Pro-inflam Macs lead to a reduced proportion of gDMSCs

The uterine granulocytes from mice that were transferred with pro-inflam Macs from the uterus and placentas of PE mice were further analyzed using sc RNA-seq. Granulocytes were identified into 12 clusters defined by marker set signatures (Figure 7A, 7B).We found that cluster 0-cluster 3 were different between the two groups: the frequency of cluster 0 and cluster 2 were increased in mice injected with pro-inflam Macs, while the frequency of cluster 1 and cluster 3 were decreased (Figure 7A). Cluster 0 has the signature of a mature granulocyte (FOS, FOSB) (Montaldo et al., 2022), and also highly expresses IL-1R2 and CSF3R; Cluster 1 highly express genes associated with immunoregulation genes (Hmox1, CCL4, CCL3, Lcn2) (Wenzel et al., 2022; Montaldo et al., 2022); Cluster 2 highly expressed pro-inflammatory genes (IFITM1, IL1B, Ptgs2, Cxcl3) (Ulff-Møller et al., 2018; Drummond al., 2022); while cluster 3 highly expressed gMDSCs associated genes (Ly6g, S100a8, S100a9, Retnlg, Wfdc21) (von Wulffen et al., 2023; Kao al., 2023) (Figure 7C). Our animal experiments above (Fig 4) have identified a reduced proportion of CD11b⁺Ly6G⁺ gMDSCs in the mice injected with pro-inflam Macs. Changes in gene expression patterns of cluster 3 (gMDSCs) were analyzed in the pro-inflam Macs group and the control group. Enriched GO terms including ‘tumor necrosis factor production’ and ‘leukocyte mediated immunity’ were found in the pro-inflam Macs group, which highly express the genes such as IRF7, EGR1, C1QA, C1QB, CCL2, CSF3R, TNFRSF1B (Figure 7D, 7E). The control group showed elevated expression of CCL4, CXCL10, CD177, MMP8, S100A13, LCN2 and SPP1; these genes were enriched in ‘granulocyte migration’ and ‘localization’ ((Figure 7D, 7E).

Figure 7

In conclusion, from the rodent experiments and the data of scRNA-seq, we demonstrated that pro-inflam Macs suppressed the production of Ly6g⁺S100a8⁺S100a9⁺Retnlg⁺Wfdc21⁺ gMDSCs.

Pro-inflam Macs induced the production of memory-like Th17 cells via IGF1-IGF1R

Our results above showed that pro-inflam Macs induced memory-like Th17 cells in PE, however, the underlying molecular mechanisms were still unknown. CellPhoneDB analysis indicated that increased communication counts and signaling pathways numbers between macrophages and T/NK cells were observed in mice injected with pro-inflam Macs (Figure 8—figure supplement 1A, 1B). Then we identified the interacting ligand-receptor pairs between different types of immune cells and macrophages. Insulin-like growth factor 1 (IGF1) -IGF1 receptor (IGF1R) ligand-receptor pair was significantly enhanced between macrophages and T/NK cells in mice injected with pro-inflam Macs (Figure 8A). IGF1 has significant effects on immune function maintenance, and signaling through IGF1R could cause increased aerobic glycolysis, favoring Th17 cell differentiation over that of Treg cells (Bekkering et al., 2018; DiToro et al., 2020). For further demonstration, we analyzed the frequencies of IGF1⁺CD14⁺ cells and IGF1R⁺CD4⁺ cells in placentas of individuals with PE and NP and found them significantly increased in the PE group (Figure 8B).

Figure 8

Trophoblast-derived extracellular vesicles from the placenta of PE (PE-EVs), which carry the fetal antigen, could induce M1 macrophage polarization to participate in the development of PE according to our previous study (X. Liu et al., 2022). To confirm whether pro-inflammatory macrophages induced the production of memory-like Th17 cells via IGF1-IGF1R, macrophages from human peripheral blood, after incubating with PBS, trophoblast-derived extracellular vesicles from NP (NP-EVs) or PE-EVs in vitro, were co-cultured with CD4⁺ naïve T cells (Figure 8C). The frequencies of memory-like CD45RO⁺ CCR7⁺ IL-17A⁺ CD4⁺ cells significantly increased in the PE-EVs-induced macrophages group rather than in the NP-EVs-induced macrophages group. However, the frequencies of memory-like Th17 cells significantly decreased after CD4⁺ naïve T cells were treated with the IGF1R inhibitor BMS-754807, an inhibitor of IGF1R (Figure 8D).

To investigate the role of IGF1-IGF1R in the development of PE in vivo, CD45⁺ CD4⁺ T cells from NP mice were isolated and treated with BMS-754807 or PBS, then injected into NP mice at day 11.5 of gestation (Figure 8E). Lipopolysaccharide (LPS)-induced PE mice model was constructed by intraperitoneal injection of LPS on day12.5 and day15.5 of gestation (Han et al., 2021), and anti-CD4 antibody was used to deplete the CD4⁺ T cells in pregnant mice. The mice injected with CD4⁺ T cells treated by BMS-754807 presented decreased embryo resorption rate and increased pup crown-rump length and fetal weight (Figure 8F), decreased SBP and UACR (Figure 8G), and decreased frequency of memory-like Th17 cells at the maternal-fetal interface (Figure 8H). These data suggested that pro-inflam Macs could induce the production of memory-like Th17 via the IGF1-IGF1R, leading to the development of PE.

Discussion

Preeclampsia, a progressive systemic disease during pregnancy, is closely related to the alterations in the immune environment at the maternal-fetal interface. In this study, the overall immune cell profile in the placenta of individuals with NP, PE, GDM and GDM&PE were detected by CyTOF and a PE-specific change in immune microenvironment of maternal-fetal interface were described for the first time. Further, we provided a novel insight that pro-inflam Macs induced memory-like Th17 cells, memory-like CD8⁺ T cells, and suppressing the production of gMDSCs in mice maternal-fetal interface by scRNA-seq. In addition, we first validated that IGF1-IGF1R was involved in producing memory-like Th17 cells induced by pro-inflammatory macrophages in vitro and in vivo, thus leading to the development of PE.

Various studies represented the effect of a particular subset of immune cells in the pathogenesis of PE or GDM (Bachmayer et al., 2006; Care et al., 2018; Cornelius et al., 2015a; Eghbal-Fard et al., 2019; Faas et al., 2014; Fu et al., 2014; Fukui et al., 2011; Lampé et al., 2015; Lampé et al., 2011; Lang et al., 2021; D. Liu et al., 2021; Lu et al., 2020; Santner-Nanan et al., 2009; Sasaki et al., 2007; Travis et al., 2020; Yao et al., 2019). However, only one report has provided an overall analysis of immune cells in the human placental villi in the presence and absence of spontaneous labor at term by scRNA-seq (Miller et al., 2022). Here we used CyTOF and scRNA-seq to comprehensively analyze the immune cell profile and explore the interaction of immune cells in PE. Additionally, most previous studies have studied immune cells from maternal systemic circulation rather than directly from the tissue of maternal-fetal interface (Cornelius et al., 2015b; Deer et al., 2021; Santner-Nanan et al., 2009; Shields et al., 2018; Wallace et al., 2012). In this study, cells were directly isolated from the tissue of maternal-fetal interface, which is considered as a more accurate approach to mimic the real immune environment of the maternal-fetal interface.

Abnormal inflammatory responses were reported to play an important role in the pathogenesis of both PE and GDM (Aneman et al., 2020; Corrêa-Silva et al., 2018; Deer et al., 2023; Jung et al., 2022; McElwain et al., 2021). However, It is unclear how the maternal-fetal interface immune microenvironment differs between the two diseases. We analyzed the overall placental immune cell profile in NP, PE, GDM and GDM&PE by CyTOF and revealed a PE-specific immune cell profile: The frequencies of memory-like Th17 cells (CD206^-CD163^-CD38^midCD107a^lowCD86^midHLA-DR^midCD14⁺) were increased, while the decreased frequencies of CD69^hiHelios ^midCD127^mid γδT cells, anti-inflam Macs (CD206⁺CD163^-CD86^midCD33⁺HLA-DR⁺) and granulocyte myeloid-derived suppressor cells (gMDSCs, CD11b⁺CD15^hiHLA-DR^low) were observed in the placenta of PE, but not in that of GDM or GDM&PE compared with that of NP.

PE is often divided into two subtypes based on the time of onset. Late-onset PE, occurring at 34 weeks of gestation or later, accounts for most preeclampsia cases, while early-onset PE, occurring before or at 33 weeks of gestation(Lisonkova and Joseph, 2013). However, our previous study observed no significant difference in the M1/M2 macrophage polarization in placentas with early-onset or late-onset PE (X. Liu et al., 2022). Therefore, our research was conducted regardless of the preeclampsia-onset type to ensure consistency in macrophage polarization of the two subtypes.

There have been studies indicating that macrophages, Th17 cells, Treg cells, and CD8⁺ T cells play an essential role in PE development (Care et al., 2018; Eghbal-Fard et al., 2019; Lager et al., 2020; Yao et al., 2019). Macrophages was indicated with a pro-inflammatory phenotype in the placenta with PE (Faas et al., 2014; Yao et al., 2019). Care et al. reported that placental Tregs helped suppress inflammation at the maternal-fetal interface (Care et al., 2018). Similarly, Lu et al. found an abnormal increase of IL-17A in the placenta in the development of PE (Lu et al., 2020). Meanwhile, the abnormal infiltration of CD8⁺ T cells in the placenta may contribute to PE (Lager et al., 2020). Consistent with the previous findings, we found that memory-like CD4⁺ and memory-like CD8⁺ T cells significantly increased in the PE group and that IL-17A was significantly higher expressed in CD4⁺ memory-like T cells from the PE group, which is vital in vascular inflammation, leading to the development of PE (Amador et al., 2014; Madhur et al., 2021). Though memory T cells play a key role in fetal-maternal tolerance in normal pregnancy (Kieffer et al., 2019), few studies have reported the characteristics of memory-like T cells in the development of PE.

We constructed an adoptive transferred mouse model and confirmed that pregnant mice developed PE symptoms after transferring PE mouse-derived memory-like CD4⁺ T cells. It is worth noting that we isolated CD4⁺ memory-like T cells directly from the maternal-fetal interface, which is different from previous studies that the CD4⁺ T cells for adoptive transfer were isolated from the spleen and induced in vitro (Cornelius et al., 2015b; Deer et al., 2021; Shields et al., 2018; Wallace et al., 2012). Recurrence is an important concern in individuals with PE (Bernardes et al., 2019). We found that a prior PE pregnancy increased the probability of PE in the next pregnancy in mice, accompanied by an increased frequency of memory-like Th17 cells at the maternal-fetal interface, suggesting that increased memory-like Th17 cells are positively related to the recurrence of PE. However, additional animal experiments are needed to confirm the function of the fetal-specific memory Th17 cells in the recurrence of PE. Activation of CD8⁺ T cells and Th1 cells can be induced by abnormal polarized macrophages in recurrent miscarriage (Li et al., 2022). gMDSCs have been reported to have immune suppressive activity and are necessary for maintaining maternal-fetal tolerance (Köstlin-Gille et al., 2019; Zhou et al., 2018). However, the correlation between macrophages and other immune cells in the development of PE remained unclear. We verified that pro-inflam Macs induced memory-like Th17 cells and memory-like CD8⁺ T cells and inhibited the production of gMDSCs in PE by correlation analysis and adoptive transfer in animal experiments for the first time, indicating that the immune cells orchestrate a network in the maternal-fetal interface elaborately.

The function of IGR1 and IGF1R in PE is controversial. IGF1 plays a significant role in maintaining immune function, and IGF1R facilitates the differentiation of naïve CD4⁺ T cells into Th17 cells (DiToro et al., 2020). A study showed that IGF-1R was downregulated in the serum of individuals with PE (Liao et al., 2021). Decreased amounts of IGF1R were also found in the placentas of individuals with PE (Robajac et al., 2015). However, another study found no significance in the affinity and the number of IGF-1R between placentas from individuals with PE or not (Díaz et al., 2005). We found that mice exhibiting PE symptoms had significantly enhanced IGF1-IGF1R interaction between inflammatory macrophages and memory-like Th17 cells by analyzing the data of scRNA-seq. Then, this finding was further confirmed in human placentas from individuals with PE and NP, for the frequencies of IGF1⁺ CD14⁺ cells and IGF1R⁺ CD4⁺ cells were found to significantly increase in the PE group. In vitro, we found that PE-EVs-treated macrophages secreted more IGF1 than NP-EV-treated macrophages and induced a higher frequency of memory-like Th17 cells. The inhibition of IGF1R on CD4⁺ naïve T cells resulted in a decreased frequency of memory-like Th17 cells, implying that the IGF1-IGF1R is critical for the production of memory-like Th17 cells induced by inflammatory macrophages. Animal experiments confirmed that IGF1R inhibition on CD4⁺ T cells at the maternal-fetal interface reduced the frequency of memory-like Th17 cells, improving PE.

In addition, placentas from different groups for CyTOF analysis and flow cytometry were matched by age, preterm pregnancies, and previous abortions; however, there are differences in body mass index (BMI), gestational age, term pregnancies, and the number of living children. Obesity can increase inflammatory and oxidative stress markers in the placental environment (Spradley et al., 2015). Also, it has been reported that the placental immune state shifts with gestational age (Lewis et al., 2018). However, as PE is often accompanied by obesity and early termination of pregnancy, it is difficult to exclude these factors in sample collection. Moreover, limited placental samples in the GDM&PE group are the shortage of this study, for it is hard to collect enough clean samples that exclude interference factors because the number of pregnant women exposed to COVID-19 has increased sharply since December 2022 in China.

In summary, this study demonstrated the statistically distinguished placental immune microenvironment in individuals with PE, but not in those with GDM or GDM&PE. More importantly, macrophages orchestrate a network in the maternal-fetal interface elaborately. These findings provide novel insights into PE pathogenesis and a potential immune target for the clinical prevention and treatment of PE.

Materials and Methods

Clinical sample collection

The samples used in this study were collected from Sir Run Run Shaw Hospital between October 2020 and August 2023. Informed consent was obtained from all volunteers, and the Ethics Committee of Sir Run Run Shaw Hospital, Zhejiang University School of Medicine, approved the study. Human placentas were obtained from women with NP, PE, GDM, or GDM&PE who underwent elective cesarean delivery. The diagnostic criteria for PE included new-onset hypertension after 20 weeks of gestation with SBP ≥140 mmHg and/or diastolic blood pressure ≥90 mmHg and proteinuria (≥300 mg) on at least two occasions. A positive glucose tolerance test diagnoses gestational diabetes mellitus. Women with normal blood pressure, full-term pregnancies, and no complications were designated as controls. The detailed clinical characteristics of the pregnant women in this study are presented in Table 5 and Table 6.

Table 5

Table 6

Mice

Eight-week-old female C57 mice and ten-week-old male BALB/c mice were purchased from Hangzhou Ziyuan Laboratory Animal Technology Co., Ltd (Hangzhou, China) and Shanghai Jihui Experimental Animal Breeding Co., Ltd (Shanghai, China), respectively. All animals were maintained under pathogen-free conditions. The Guide for the Care and Use of Laboratory Animals (China) conducted all experimental procedures involving animals, and the Animal Research Ethics Committee of the Sir Run Run Shaw Hospital of Zhejiang University approved the protocols.

Female C57 mice were mated with male BALB/c mice to establish an allogeneic pregnancy model (Rowe et al., 2012). The day of the vaginal plug detection was considered day 0.5 of pregnancy. SBP was measured using a noninvasive mouse tailcuff BP analyzer (BP-2010A, Softron, Japan) at 12.5, and 16.5 days of gestation. Three random urine samples were collected after day 16.5 of gestation, and UACR was measured by urinary microalbumin (CH0101060, Maccura, China). Mice were euthanized on day 18.5 of gestation.

Reduction in uterine perfusion pressure mouse model

To construct a mouse model of PE, we ligate uterine arteries in pregnant mice on day 12.5 of gestation. Briefly, after 4% chloral hydrate was intraperitoneal injected for anesthesia, bilateral incisions were made on the back of the pregnant mice, and surgical sutures were used to reduce the blood flow of the bilateral uterine arcades.

Isolation of single cells from the mouse uterus and human placenta

Uterus from pregnant mice and placentas from volunteers were washed twice with ice-cold PBS and cut into small pieces. The tissues were digested with collagenase type IV (1 mg/ml, Sigma-Aldrich, USA) and DNase I (0.01 mg/ml, Sigma-Aldrich, U.S.A) in RPMI 1640 medium (Thermo Fisher Scientific) for 40 min at 200 rpm and 37 °C. The suspensions were strained through 70-μm nylon mesh and centrifuged at 500×g for 5 minutes. Leaving the supernatants, the cell pellets from human placentas need an extra purification by Ficoll (P4350, Solarbio, China) according to the manufacturer’s instructions. Human CD4⁺ memory T cells were isolated using a Human Central and Effector Memory CD4⁺ T Cell Isolation Kit (17865, STEMCELL, Canada).

Adoptive transferred mouse model

Endogenous macrophages or CD4⁺ T cells were depleted by injecting PLX3397 (S7818, Selleck, USA), clodronate liposomes (40337ES08, YEASEN, China) or anti-CD4 antibody (BE0003-1, BioXcell, USA) intraperitoneally every three days starting from day 10.5 of gestation.

For the adoptive transferred mouse model, phycoerythrin-conjugated anti-mouse CD4 (12-0041-82, eBioscience, USA) and Pc5.5-conjugated anti-mouse CD44 (45-0441-82, eBioscience, USA) were used to label CD4⁺ memory-like T cells from uterus and placentas. PE-conjugated anti-mouse CD45 (E-AB-F1136D, Elabscience, China), FITC-conjugated anti-mouse F480 (11-4801-82, eBioscience, USA), and PE-Cyanine7-conjugated anti-mouse CD206 (141719, Biolegend, USA) were used to label anti-inflam Macs from uterus and placentas, and the remaining CD45⁺F480⁺CD206^- cells were considered pro-inflam Macs. The MoFlo Astrios EQ FACS Cell Sorter at Core Facilities, ZheJiang University School of Medicine (China), was used to obtain specific cells with a purity of > 96%. Briefly, 2–5 × 10⁵ CD4⁺ CD44⁺ T cells, 5×10⁵ pro-inflam Macs, or 5×10⁵ anti-inflam Macs were injected to pregnant C57 mice intravenously at day 12.5 of gestation.

2–5 × 10⁵ CD45⁺CD4⁺ T cells from uterus and placentas of pregnant mice were sorted and transferred to pregnant C57 mice at day 11.5 of gestation after treating with 100nM BMS-754809 (HY10200; MedChemExpress; USA) or PBS for three days in vitro.

Mass cytometry by time of flight

7 × 10⁵ cells from NP (n=9), PE (n=8), GDM (n=8), GDM&PE (n=7) presented in Table 1 were stained for cell surface markers for 30 min at 4°C in staining medium (PBS containing 1% BSA and 0.02% NaN₃). Cells were washed with protein-free PBS, stained with 2.5 μM cisplatin for 5 min at 25°C and fixed using transcription factor buffer set (BD Biosciences), followed by intracellular staining for 60 min at 4 °C. Then the cells were washed and stored at 4 °C until acquisition. Finally, cells were analyzed on a CyTOF mass spectrometer (Fluidigm, South San Francisco, CA, USA). The data were normalized for each experiment using EQ four-element calibration beads (Fluidigm) and the reference EQ passport P13H2302.

RNA sequencing

5×10⁵ CD45⁺F4/80⁺CD206^- pro-inflam Macs (n=3) and 5×10⁵ CD45⁺F4/80⁺CD206⁺ anti-inflam Macs (n=3) were isolated from the uterus and placentas of mice with RUPP and high throughput sequencing and bioinformatics analyses were conducted at Shenzhen Huada Gene Technology Service Co. Ltd.(Shenzhen, China). Only differential genes with more than two-fold change and a corrected P value less than 0.05 were considered statistically significant.

Single-cell RNA sequencing (scRNA-seq)

Mice were injected with 5×10⁵ CD45⁺F4/80⁺CD206^- pro-inflam Macs or 5×10⁵ CD45⁺F4/80⁺CD206⁺ anti-inflam Macs at 12.5 days of gestation and euthanized on day 18.5 of gestation. CD45⁺ immune cells from the uterus and placenta were isolated from mice transferred pro-inflam Macs and anti-inflam Macs. scRNA-seq was conducted at PLTTECH Service Co. Ltd. (Hangzhou, China). Differential genes with a corrected P < 0.05 were considered statistically significant.

Isolation of NP-EVs and PE-EVs

NP-EVs and PE-EVs were isolated and identified using our published protocols (Jiang et al., 2021; X. Liu et al., 2022). Briefly, after digesting the placental tissues from NP or PE, the suspensions were filtered through 100-μm nylon mesh and centrifugated at 3,000×g for 15 minutes. Then the supernatants were filtered with a 0.22-μm filter and centrifuged at 100,000×g for 1h at 4 °C. Then the pellets of EVs were resuspended and centrifuged at 100,000×g once again. A BCA assay kit conducted Protein quantitation of EVs (23235, Thermo Fisher Scientific). To isolate the T-EVs, 500 μg of the EVs were incubated with 1 µg placental alkaline phosphatase antibody (SC-47691, Santa Cruz Biotechnology) at 4 °C overnight, then washed in the recommended buffer (PBS containing 2% exosome-free FBS and 1 mmol/L ethylenediaminetetraacetic acid), and centrifuged at 100 000×g for 1 hour at 4 °C. After resuspended, the total EVs were sorted by EasySep Mouse PE Positive Selection Kit II (17666, STEMCELL) to collect the final NP-EVs or PE-EVs.

Induction of CD4⁺ memory-like T cells

Macrophages were obtained after five days of culture of mononuclear cells isolated from human peripheral blood following previous protocols (X. Liu et al., 2022). NP-EVs or PE-EVs in a 50 μg/mL concentration were added to macrophages and co-cultured at 37 °C for 8 h. The human CD4⁺ naïve T cells Isolation Kit (19555, STEMCELL, Canada) was used to isolate purified CD4⁺ naïve T cells from human peripheral blood cells according to the manufacturer’s instructions. Then, CD4⁺ naïve T cells were co-cultured with EV-treated macrophages for six days. Flow cytometry was performed to measure the frequency of memory-like Th17 cells. For animal experiments, CD4⁺ naïve T cells sorted from the uterus and placentas of pregnant mice were cultured with the IGF1R inhibitor BMS-754807 at a concentration of 10 μM for three days before co-cultured with macrophages.

Flow cytometry

Single immune cells from the mice uterus were obtained following the method described above. After incubation with Cell Stimulation Cocktail (00-4975-93, Invitrogen, USA) for five hours at 37 °C, cells were collected and surface staining for either phycoerythrin-conjugated anti-mouse CD4 (12-0041-82, eBioscience, USA), APC-conjugated anti-mouse CD8 (100711, BioLegend, USA), phycoerythrin-Cy5.5-conjugated anti-mouse CD44 (45-0441-82, eBioscience, USA), APC-conjugated anti-mouse CD11b (101212, BioLegend, USA), FITC-conjugated anti-mouse Gr-1 (108406, BioLegend, USA), PE-Cyanine7-conjugated anti-mouse Ly6G (E-AB-F1108H, elabscience, China) or intracellular staining for 488-conjugated anti-mouse IL-17A (506910, BioLegend, USA) was performed according to the manufacturer’s instructions.

Single placental and peripheral lymphocytes and were obtained following the method described above. The Human Central and Effector Memory CD4⁺ T cell Isolation Kit (17865, STEMCELL, Canada) was used to obtain purified memory CD4⁺ T cells, according to the manufacturer’s instructions. After incubation with Cell Stimulation Cocktail (00-4975-93, Invitrogen, USA) for five hours at 37 °C, cells were collected and surface staining for phycoerythrin-conjugated anti-human CD4 (2384240, eBioscience, USA), FITC-conjugated anti-human CD45RO (304204, BioLegend, USA), and phycoerythrin-Cy7-conjugated anti-human CCR7 (353227, BioLegend, USA) or intracellular staining for APC-conjugated anti-human IL-17A (17-7179-42, eBioscience, USA) and Foxp3 (320014, BioLegend, USA) was performed according to the manufacturer’s instructions.

Immunofluorescence

Frozen sections of the placentas were permeabilized with PBS containing 0.5% Triton X-100 (PBST) for 20 min and incubated for 1 h with a blocking buffer in PBST. The sections were then incubated with anti-CD4 (sc-1176, Santa Cruz Biotechnology, China) and FITC-conjugated anti-human CD45RO (304204, BioLegend, USA) at 4 °C overnight, followed by Alexa Fluor 568-conjugated secondary antibodies (dilution: 1:200, Yeasen, China) for 1 h. The slides were counterstained with 4,6-diamidino-2-phenylindole (DAPI, 1 μg/ml; Roche, Switzerland) for 20 min. Digital images were obtained using confocal fluorescence microscopy (ZEISS LSM 800, Germany). ImageJ software was used to quantify the fluorescence intensity from immunofluorescence (IF) images.

Statistical analysis

Data were analysed using SPSS version 20. After the Shapiro-wilk test, the data were confirmed to be non-normal distribution. Kruskal-wallis test was used to compare the results of experiments with multiple groups. All data are presented as mean ± SEM (*P < 0.05, ** P < 0.01, *** P < 0.001; NS, not significant).

Acknowledgements

The authors would like to express their heartfelt gratitude to the participants for their contributions. This work was financially supported by the National Natural Science Foundation of China (82271694), the Zhejiang Province Natural Science Foundation key project fund (LZ24H040002), the Zhejiang Medicine and Health Science and Technology Plan Project (2023KY800).