A dynamic peripheral immune landscape during human pregnancy

Abstract

Extensive immune adaptations occur during pregnancy to ensure successful delivery. However, these changes can increase the risk of disease in the mother. Here, we conducted single-cell RNA sequencing on peripheral blood mononuclear cells from pregnant women at different stages of pregnancy to elucidate the dynamic transcriptional changes in the maternal immune system. Gradual reduced cytotoxicity phenotype in highly variable cytotoxic T and natural killer cell types were observed during pregnancy. Reduced T- and B-cell response-related MHC-II and CD40 signaling as well as enhanced protolerance inducible costimulator and activin signaling may underlie the pregnancy-related weakening of adaptive immunity. Conversely, pro-inflammatory genes and pathways were upregulated in monocytes, possibly to compensate for the reduced T-cell response. Moreover, the transition from adaptive immune reduction to activation in late pregnancy in dendritic cells and CD4⁺ T cells was also detected. Notably, we proposed a novel view of the pro-aging effect of pregnancy from the perspective of immunity, and this effect may be restored postpartum. This work expands our knowledge of pregnancy immunity and may provide insights into the altered disease risks during pregnancy.

Graphical abstract

1. Introduction

Pregnancy is a complex period in which a series of anatomical and physical changes occur to ensure the reproduction of a healthy baby [1]. The immune system undergoes extensive adaptive changes during pregnancy [2]. A finely tuned immune system is required to maintain tolerance to the fetal allograft while preserving protection for both the mother and fetus against microbial challenges [3,4]. Conversely, dysregulated immunity is associated with pregnancy complications and adverse pregnancy outcomes, such as recurrent pregnancy loss [5], pre-eclampsia [6], and preterm labor [7]. Moreover, under normal conditions, the unique immune status of pregnant women also renders them more or less susceptible to various diseases than the general population [8,9]. In particular, pregnant women are more vulnerable to some types of infections [10]; indeed, influenza infections elicit higher mortality and morbidity rates in pregnant women, and the symptoms are more severe during the second and third trimesters [11]. By contrast, pregnancy may act as a protective factor against some autoimmune diseases, such as multiple sclerosis, particularly in the third trimester [12,13].

Two research directions, namely immunity in the feto-maternal interface and maternal systemic immunity, have dominated studies on pregnancy immunity. Feto-maternal interface immunity, mainly occurring in the local immune environment of the placenta, is pivotal for tolerance to the semi-allogeneic fetus as well as fetal development and pregnancy success [14], [15], [16]; however, this type of immunity does not reflect the systemic condition of the mother. By contrast, maternal systemic immunity describes the general immune state of pregnant women. Maternal systemic immunity is related to successful delivery and altered disease susceptibility to infections or autoimmune diseases in the mother [9,17]. In addition, considering the limited access to placental tissue during pregnancy, assessing placental processes by measuring changes in peripheral blood could have high clinical value for predicting adverse pregnancy outcomes. Aberrant molecules or cells in the blood, such as high levels of S100A8 and tumor necrosis factor (TNF)-α and a high neutrophil-to-lymphocyte ratio, have been reported to be related to miscarriage [18], [19], [20], [21]. However, compared with the extensive studies focused on the feto-maternal interface, much less effort has been made to understand maternal systemic immunity.

With regard to the circulating immune system, pregnancy is accompanied by a reduction in adaptive immunity, as demonstrated by decreased frequencies of T cells (TCs) and B cells (BCs) [22,23], whereas immune responses of certain innate immune cells, such as monocytes (MCs) and plasmacytoid dendritic cells (pDCs), are exacerbated [24]. Moreover, disease susceptibility varies across different pregnancy stages, indicating dynamic immune states during pregnancy. Several studies have reported immune cellular changes in different pregnancy stages. The absolute number of MCs and granulocytes in the peripheral blood increases significantly in the mid-pregnant stage and gradually decreases thereafter [22]. Additionally, the number of circulating BCs decreases obviously in the third trimester [22]. TCs in mid-pregnancy exhibit upregulation of inhibitory markers such as programmed death ligand 1, whereas TCs in late pregnancy express higher levels of activation markers such as CD38 and neopterin, compared with those in nonpregnant controls [25]. However, these studies have mainly focused on specific pregnancy stages and immune cell types. Therefore, studies that comprehensively characterize the continuous changes in immune cells during pregnancy are required to elucidate the systemic immune states during pregnancy.

Single-cell RNA sequencing (scRNA-seq) is a powerful tool for dissecting cellular heterogeneity, identifying complex cellular events, and deepening our understanding of biological systems [26]. Accordingly, in this study, we utilized scRNA-seq to generate a dynamic transcriptional landscape of peripheral blood mononuclear cells (PBMCs) from pregnant women during three stages of pregnancy. In addition, we compared these data with results from aging women and women in the postpartum period to investigate the potential relationships among aging, pregnancy, and immune changes postpartum. Our results provide insights into fluctuations in disease susceptibility during pregnancy and the potential pro-aging effects of pregnancy, which may be reversed postpartum.

2. Materials and methods

2.1. Human subjects and PBMCs acquisition

To be eligible for study participation, subjects met the following inclusionary criteria: physical and psychological health; no clinically significant abnormalities in blood chemistry. Exclusion criteria included any physiological or psychiatric pathology, medication, smoking, obesity, binge drinking, and pregnancy-related complications. The experiments using human samples in this study were approved by the Ethics Committee of Zhongshan Ophthalmic Center and Xiangya Hospital (ID: 2021030123), and the informed consent was obtained from every subject. After whole blood was collected from the subject, PBMCs were isolated by using a Ficoll Hypaque density solution according to standard density gradient centrifugation methods. The cell viability was tested, and we require the cell viability of all samples to be greater than 80%.

2.2. Single-cell RNA library preparation and sequencing

A total of 5 nonpregnant young women controls (HC), 5 aged women, and 27 pregnant women at various stages (early stage, ES, n = 6; mid stage, MS, n = 6; late stage, LS, n = 12; postpartum, PP, n = 3, Table S1) were included in the scRNA-seq study. The single-cell suspensions of the scRNA-seq samples were converted to barcoded scRNA-seq libraries using the Chromium Single Cell 5’ or 3’ library, Gel Bead and Multiplex Kit, and Chip Kit (10x Genomics, Pleasanton, CA, USA). Samples of HC and aged women were processed with the 5′ library preparation kit. In addition, for the pregnant samples (ES, MS, LS, and postpartum), one sample incorporated PBMCs from three people at the same pregnancy stage to obtain a sufficient number of PBMCs for sequencing, then were prepared using a Chromium Single Cell 3′ library preparation kit. FastQC software was used to check library quality. The Cell Ranger (version 5.0.0) was used for the initial processing of the sequencing data.

2.3. Single-cell data processing

After obtaining each library gene count, the aggr pipeline in the CellRanger Software Suite was applied to demultiplex and barcode the sequences. Based on the calculation of the single-cell expression matrix by CellRanger, the output was loaded by using Seurat (version 3.2.3, https://satijalab.org/) for further analysis. For quality control, we filtered out the cells highly expressing HBB, HBA1, and several light and heavy chain transcripts, which were considered to be RBC-contaminated cell populations. Next, cells with fewer than 200 genes detected and a mitochondrial gene ratio of greater than 15% were excluded. A total of 19 libraries were sequenced and 196688 cells (ES, 22349 cells; MS, 17980 cells; LS, 45120 cells; PP, 5403 cells; HC, 48130 cells; and AA, 57706 cells) were collected in subsequent analyses.

2.4. Dimensionality reduction and clustering analysis

For analysis of scRNA-seq data with Seurat, NormalizeData was used to normalize the data. The gene-barcode matrix was analyzed using the principal component analysis (PCA). The harmony R package was used to remove the batch effects and integrated the data from different groups. FindClusters was used to cluster cells and RunTSNE was applied to visualize using a 2-dimensional t-SNE algorithm. Additionally, FindAllMarkers was used to calculate marker genes for different clusters with default parameters. Unsupervised clustering identified 8 major cell types and 25 immune subpopulations based on the previous studies. The detailed cell clustering strategies were provided in Supplementary Tables 2 and 3. The detailed cell numbers were provided in Table S4.

2.5. DEGs analysis

DEGs analysis from different clusters between different groups was performed by FindMarkers with the Wilcoxon rank-sum test. Preg- (Pregnancy/HC), Aging- (AA/HC), or PP- (postpartum/LS) related DEGs datasets were established (adjusted P value < 0.05, |Log2FC| > 0.25).

2.6. Gene ontology enrichment analysis

Gene ontology (GO) biological process and pathway analysis were conducted on the Metascape webtool (www.metascape.org), which supports statistical analysis and visualization of functional profiles for genes and gene clusters. Among the top 30 enriched GO terms or pathways across various cell types, 10 biological process and pathway terms were selected and drawn using the ggplot2 R package.

2.7. Cell-cell communication analysis

To analyze the cell-cell communication between different clusters and the differences in cell-cell communication between different groups, CellChat R package and іTALK (https://github.com/Coolgenome/iTALK) were applied to the data. The signaling pathway networks were analyzed and visualized with default parameters.

2.8. Pseudotime analysis

Pseudotime was conducted with monocle2 R package to determine the developmental pseudotime of specific cell population affected by pregnant stages. Following the monocle vignette, we used UMI count data as input and selected DEGs identified in the last procedure to order the cells. We plotted DEGs related to specific bioprocesses along the inferred developmental pseudotime. Group annotation of cells was also aligned along the pseudotime axis. The structure of the trajectory was plotted in 2-dimensional space using the DDRTree dimensionality reduction algorithm, and the cells were ordered in pseudotime with default parameters.

2.9. Flow cytometry analysis

To verify some results, we collected PBMCs from 36 healthy subjects (Table S5). The serums were collected and PBMCs were isolated and stored in a liquid nitrogen container. Then PBMCs were resuscitated and washed for the subsequent flow cytometry analysis. LIVE/DEAD (APC-Cy7, #423105, Thermo Fisher Scientific) was first labeled to distinguish the living cells. For the surface staining, PBMCs were simultaneously stained with anti-human fluorochrome-labeled antibodies specific for CD45 (BV785, #304048), CD3 (BV421, #300434), CD14 (PerCP/Cy5.5, #301824), CD4 (APC, #317416 or PE, #300508), CD8 (PerCP/Cy5.5, #301031), CD45RA (BV785, #304140), CD45RO (BV605, #304238), CD1C (PerCP/Cy5.5, #331514), CD19 (BV605, #302244), CD20 (APC, #302310), CD38 (BV785, #303530), HLA-DQ (PE, #318105), BCMA (PE-Cy7, #357508), CD74 (APC, #326811), CD69 (BV650, #310934), CD83 (PE, #305307) (BioLegend, San Diego, USA) and CD18 (APC, #E-AB-F1057E) (Elabscience). For the PIM1 staining, cells were stained with surface antibodies, fixed, permeabilized, stained with PIM1 antibody (#3247S), then stained with FITC-labeled antibody (#4412S) (Cell Signaling Technology, Danvers, USA). For the aging markers staining, cells were stimulated with the cell stimulation cocktails including phorbol myristate acetate (50 ng/mL), ionomycin (500 ng/mL) and brefeldin A (1 µg/ml) (Sigma) under a 5% CO₂ environment at 37 °C for 4 h. Then, cells were stained with surface antibodies, fixed, permeabilized, and stained with P21 (FITC, #ab282187) (Abcam), STAT3 (PE, #560391) (BD Bioscience). Flow cytometric analysis was performed on a flow cytometer BD LSR Fortessa (BD Biosciences, San Jose, CA, USA). Finally, FlowJo (version 10.0.7, Tree Star, Ashland, OR, USA) was employed to analyze the results. Antibody sets and staining strategies to identify cells include CD45+CD14-CD56-CD3+ (TCs), CD45+CD3-CD19-CD14+ (cMC), CD3+CD8-CD4+CD45RA-CD45RO+ (CD4Tem), CD3+CD8+CD4-CD45RA-CD45RO+ (CD8Tem), CD19-CD1C+ (cDC2), CD3-CD19+CD20-CD38+ (PC), CD3+CD4+CD8+, CD19+HLA-DQ+, CD19+BCMA+, CD3+CD4+CD74+, CD3+CD4+CD18+, CD3+CD4+CD69+, CD3+CD4+PIM1+, CD3-CD14+CD83+, CD3-CD14+P21+, CD3-CD14+STAT3+.

2.10. ELISA analysis

After PBMCs isolation, the serums were collected from the samples and stored at −80 °C for further analysis. IL-1β levels in the serums were detected using enzyme-linked immunosorbent assay (ELISA) kit (Invitrogen #88-7261-88).

2.11. Statistical analysis

GraphPad Prism Software (version 8.0.2; GraphPad Software Inc., La Jolla, CA) was used for data analysis and presentation. Statistical analysis was performed with an unpaired, two-tailed Student's t-test for analyzing cluster abundance. For comparing the level of genes between groups, P value was calculated using two-sided Wilcoxon test as implemented in the function “compare_means” of ggpubr R package with default parameters. In addition, P values were derived by a hypergeometric test with the default parameters in Metascape webtool in the functional analysis of DEGs. P values above 0.05 were considered as not significant, ns; *P < 0.05, **P < 0.01, ***P < 0.001, ****P < 0.0001.

2.12. Data availability statement

The sequencing data used in this study are available from the corresponding author upon request. The scRNA-seq data of AA4, AA5, and all samples related to pregnancy is deposited in the Genome Sequence Archive in BIG Data Center, Beijing Institute of Genomics (BIG, https://bigd.big.ac.cn/gsa-human/), Chinese Academy of Sciences, under the Project Accession No. PRJCA007692 and GSA Accession No. HRA001727. In addition, all 5 young HC and AA1-AA3 data were obtained from the National Genomic Data Center (GSA Accession No. HRA000624).

3. Results

3.1. Study design and single-cell analysis

To identify alterations in the circulating immune system during pregnancy and across pregnancy stages, we generated an immune cell atlas of peripheral blood from 27 pregnant women in three pregnancy stages, including six in early-stage pregnancy (ES, before 14 weeks of pregnancy), six in mid-stage pregnancy (MS, between 14 and 28 weeks of pregnancy), and 12 in late-stage pregnancy (LS, after 28 weeks of pregnancy), as well as three women at six-week postpartum (Fig. 1a). In addition, data from five age-matched nonpregnant healthy controls (HCs) and five healthy aged women were also evaluated in this study to act as the control and explore the potential relationships between pregnancy and aging (Fig. 1a). Detailed information on the patients enrolled in this study is provided in Supplementary Table 1. After unbiased clustering, megakaryocytes, CD34⁺ cells, and five major immune cell lineages (TCs, natural killer [NK] cells, BCs, MCs, and DCs) were identified based on their expression of canonical lineage markers (Fig. S1a). Another cluster of cells that highly expressed neutrophil markers (CXCL8 and CSF3R), but did not express other neutrophil markers such as CEACAM8, was named neutrophil-like cells (Fig. S1a). We then subclustered these five immune cell types into 25 transcriptionally classical subsets based on published studies [27], [28], [29] (Fig. S1; Table S2, 3).

Fig. 1.

scRNA-seq reveals the altered immune ecosystem influenced by pregnancy. (a) Schematic of the experimental design for single-cell RNA sequencing (scRNA-seq). The box plots showing the ratio of TC (b), MC (c), cMC (d), cDC2 (e), CD4Tem (f), CD8Tem (g), PC (h) in immune cell between HC and Preg groups. (i) The flow cytometry histogram (left) and box plots (right) showing the ratio of TC in immune cells between HC and Preg groups. (j) The flow cytometry histogram (left) and box plots (right) showing the ratio of cMC in immune cells between HC and Preg groups. (k) The flow cytometry histogram (left) and box plots (right) showing the ratio of CD4Tmem between HC and Preg groups. (l) The flow cytometry histogram (left) and box plots (right) showing the ratio of CD8Tmem between HC and Preg groups. Significance in B-L was calculated using an unpaired, two-tailed Student's t-test; *P < 0.05, **P < 0.01, ***P < 0.001, ****P < 0.0001. The full names of the cell subsets see Fig. S1.

3.2. Reconstitution of the circulating immune ecosystem by pregnancy

To demonstrate the impact of pregnancy on the circulating immune cell composition, we first compared the proportions of immune cells between the HC and pregnancy groups. The TCs proportion decreased, whereas myeloid cells (MCs and DCs) increased during pregnancy (Figs. 1b, c and S2a). The increased proportions of MCs and DCs were mainly related to increased classical MCs (cMCs), intermediate MCs, conventional dendritic cells 2 (cDC2s), and pDCs (Figs. 1d, e and S2b, c). Regarding TC subtypes, we found that pregnancy led to decreased proportions of CD4⁺ effector memory TCs, CD4⁺ cytotoxic TCs, and CD8⁺ effector memory TCs as well as an increased proportion of CD4⁺ central memory TCs in blood immune cells (Figs. 1f, g and S2d, e). The frequency of plasma cells also decreased (Fig. 1h). These trends were observed on the subset composition alterations of TCs and BCs (Fig. S2f, g). We further verified the decreased proportions of TCs, CD4⁺ and CD8⁺ memory TCs, as well as the increased proportions of cMCs and cDC2s by flow cytometry analysis (Figs. 1i–l and S2h, i). Decreased proportions of plasma cells among BCs were also validated (Fig. S2j). These results suggested a state of lymphopenia and myeloid cell expansion during pregnancy.

3.3. Reduction in lymphocyte functional genes during pregnancy

Reduced lymphocytes were detected in pregnancy. Therefore, we next explored their functional changes during pregnancy. We constructed a pregnancy-related differentially expressed genes (DEGs) database for various TC and NK cell subsets via DEGs analysis between pregnancy and HC groups (designated Preg DEGs). We observed a greater number of downregulated DEGs than upregulated DEGs in nearly all NK and TC subsets (Fig. 2a). Notably, unsupervised analysis revealed that the six TC and NK cell subclusters with the largest numbers of DEGs were characterized by high expression of cytotoxicity-related genes, such as PRF1, GNLY, GZMB, and GZMK, compared with the other TC and NK cell subclusters (Fig. 2a). We thus termed these six subclusters showing high transcriptomic changes as variable cytotoxic cells (VCCs). VCCs contained CD4⁺CD8⁺ double-positive TCs, CD8⁺ effector memory TCs, CD8⁺ cytotoxic TCs, and three NK subgroups, including a CD16 (FCGR3A)⁻ CD56 (NCAM1)^bright immature NK population (NK1), a CD16⁺ CD56^dim CD57 (B3GAT1)⁻ cytotoxic NK subset (NK2), and a CD16⁺ CD56^dim CD57⁺ late NK population (NK3) (Figs. S1c and S3a). Next, we conducted GO analysis on the downregulated DEGs among VCC subsets. We found reductions in various pathways, including the type I interferon signaling pathway, leukocyte migration, leukocyte differentiation, leukocyte-mediated cytotoxicity, and cellular defense response, in these cell subsets (Fig. S3b). In addition, genes involved in the above pathways showed a gradual decrease in expression as pregnancy progressed, as revealed by pseudo-time analysis (Fig. 2b), a useful tool to computationally order cells based on their gradual transition of gene expression [30]. These results suggested that the functions of cytotoxic TCs and NK cells in cell migration, differentiation, cytotoxicity-related bioprocesses, and defense response were gradually reduced during pregnancy.

Fig. 2.

Preg led to the downregulation of lymphocyte activation and adaptive immune response. (a) Identification of variable cytotoxic cells (VCCs). The dot plot showing the number of upregulated and downregulated DEGs in NK&TC subsets (left). The most affected subsets highly express the cytotoxicity-related genes (right). (b) Heat map showing DEGs of VCCs across pregnant stages pseudotime. Gene expression levels across the pseudotime axis are normalized, and grouped by their functional categories (top). Group annotation of cells aligned along the pseudotime axis (bottom). (c) Dot plot showing the representative GO biological process and pathways enriched in downregulated DEGs in CD4⁺ TC subsets. (d) The expression changes of Th17-related (top) and TCR-related (bottom) genes along the pseudotime axis. (e) UpSet plot showing the integrated comparative analysis of downregulated DEGs among BC subsets. (f) Venn plot showing the integrated comparative analysis of downregulated DEGs of BC among 3 pregnancy stages.

CD4⁺ TCs and BCs play important roles in autoimmunity [31]. During pregnancy, CD4⁺ TC subsets showed commonly downregulated DEGs enriched in Th17 cell differentiation, TC receptor signaling, and TC proliferation, which were driven by IL27RA, IL2RG, NFATC2, CD40LG, LAT, and ZAP70 (Figs. 2c and S3c). Notably, the expression levels of these genes decreased as pregnancy progressed (Figs. 2d and S3d). In the CD4⁺ naïve TC subset, pregnancy downregulated the leukocyte transendothelial migration pathway, which was driven by the downregulation of RHOH and NCF1 (Figs. 2c and S3c). In addition, pregnancy decreased the levels of GZMK, GZMA, and CCR2 in the CD4⁺ cytotoxic TC subset (Fig. S3b). In BC, pregnancy decreased BC-related bioprocess pathways (e.g., activation, proliferation, and BC receptor pathway) in naïve, memory, and autoimmune-related BCs (Fig. S3e). The decrease in protein processing was more pronounced in plasma cells (Fig. S3e). These processes and pathways are closely associated with BC-related autoimmune activation. Using UpSet plots (Fig. 2e), we found that all BC subsets showed decreased expression of BC-related genes, including JCHAIN and SSR2. Moreover, we identified subtype-specific expression patterns, including the antibody production-related genes SSR3 and XBP1 in plasma cells; and BC differentiation-related genes, i.e., TCL1A, BACH2, STMN1, and CCR6, in highly proliferating BCs, including naïve and memory BCs (Figs. 2e and S3f). To explore the effects of pregnancy stages on BC states, we analyzed the downregulated DEGs between the ES, MS, LS, and HC groups. The number of downregulated DEGs gradually increased with pregnancy stage (Fig. 2f). Compared with the HC group, all three pregnancy stages exhibited a common pattern of pregnancy-repressed DEGs, which were related to BC activation (Fig. 2f). In addition, we identified some stage-specific downregulated genes, including decreased XBP1 in the ES group; CD47, CD7, and EIF3H in the MS group; and BANK1, IL4R, IGHG1, and IGHA1 in the LS group, indicating stage-specific transcriptional modulation in BCs (Fig. 2f).

Collectively, the transcriptomes of lymphocytes were altered by pregnancy and varied across cell types and pregnancy stages. The reduction in functional genes in cytotoxic TC and NK cell types may account for increased infection risk, whereas that in CD4⁺ TCs and BCs may be associated with increased infections and reduced autoimmune response.

3.4. Aberrant cell-cell communication patterns may contribute to the decreased adaptive immune response during pregnancy

Our data suggested multiple weakened functions of adaptive immune cells during pregnancy. Myeloid cells are important antigen-presenting cells (APCs) that assist in adaptive immunity. Thus, we next explored the gene signatures of the circulating myeloid cells and the intercellular interactions associated with the adaptive immune response. GO analysis of the DEGs of each myeloid cell subset identified pregnancy-induced downregulation of antigen processing and presentation, MHC class II (MHC-II) antigen presentation, and lymphocyte activation and proliferation pathways (Fig. 3a). These pathways were particularly downregulated in the cDC2 subset, which is a bridge between innate and adaptive immunity [32]. Pregnancy also decreased the levels of protein processing and maturation genes in pDCs, which are robust interferon I-producing cells [33] (Fig. 3a). Next, we generated a Venn plot of downregulated DEGs from myeloid cell subsets and found that genes related to antigen processing and presentation were repressed in the cDC2 subset (Fig. 3b). In addition, pregnancy decreased the expression of the activation genes IRF7 and IRF8 in the pDC subset (Fig. 3b). These results suggested that pregnancy decreased the functions of myeloid cells in adaptive immunity, particularly in the cDC2 subset. The interferon I-producing function of pDCs may also be dampened by pregnancy.

Fig. 3.

Aberrant cell-cell communication supported the decreased adaptive immune response after pregnancy. (a) Dot plot showing the representative GO biological process and pathways enriched in downregulated DEGs in DC and MC subsets. (b) Venn plot showing the integrated comparative analysis of downregulated DEGs among DC and MC subsets. (c) Circle plot showing the inferred MHC-II signaling networks. (d) Violin plot showing the expression of genes related to MHC-II signaling among immune populations between HC (blue) and Preg (red) groups. (e) Circle plot showing the inferred CD40 signaling networks. (f) Circle plot showing the inferred ICOS signaling networks in Preg group but absent in HC group. (g) Violin plot showing the expression of genes related to ICOS signaling among immune populations between HC (blue) and Preg (red) groups. (h) Circle plot showing the inferred MIF signaling networks. (i) Violin plot showing the expression of genes related to MIF signaling among immune populations between HC (blue) and Preg (red) groups. (j) The flow cytometry histogram showing the expression of CD74 in CD4⁺ TCs. Box plots showing the ratio of CD74-expressing CD4⁺ TCs between HC and Preg groups. (k) Violin plot showing the expression of genes related to ICAM signaling among immune populations between HC (blue) and Preg (red) groups. (l) The flow cytometry histogram showing the expression of CD18 in CD4⁺ TCs (left). Box plots showing the Mean Fluorescence Intensity (MFI) of CD18 in CD4⁺ TCs between HC and Preg groups (right). Significance in j, l was calculated using an unpaired, two-tailed Student's t-test; *P < 0.05.

Cell-cell interactions orchestrate homeostasis and single-cell functions [34]. Thus, we explored intercellular communication signaling in pregnancy based on the expression of ligand-receptor pairs. Using CellChat, a tool inferring intercellular communication via ligand-receptor pairs [35], we found that the MHC-II signaling pathway was decreased in the pregnancy group compared with that in the HC group (Fig. 3c). cDC2s and BCs, which can function as APCs, were the primary ligand sources, acting in an autocrine manner or in a paracrine manner with other lymphocytes, including CD4⁺ TCs (Fig. 3c, d). The decreased expression of HLA-DQ by BCs was validated by flow cytometry (Fig. S4a, b). Our data revealed that the signaling involved in the BC activation function of CD4⁺ TCs was also reduced in the pregnancy group, driven by CD40LG and CD40 in CD4⁺ TCs and BCs, respectively (Figs. 3e and S4c). Inducible costimulator (ICOS) and activin signaling pathways are important for the establishment of a tolerogenic immune response in circulation and in the placenta [36], [37], [38]. These two pathways were only present in the pregnancy group, not in the HC group (Figs. 3f, g and S4d). These results suggested that pregnancy may reduce adaptive immune responses by dampening intercellular antigen presentation and immune cell activation signaling and enhancing immune tolerance-related signaling.

We also explored the intercellular interactions among lymphocytes. MIF signaling has multiple regulatory functions in various immune cells and promotes autoimmune and inflammatory diseases [39]. ICAM is an adhesion molecule that mediates immune cell migration [40]. CD4⁺ and CD8⁺ cells were the primary MIF source, which acted in a paracrine manner to BCs, with the MIF/CD74 ligand/receptor pair being the major driver of signaling (Fig. 3h, i). We found that pregnancy suppressed the MIF signaling pathway mainly in CD4⁺ and CD8⁺ cells, as confirmed by the decreased expression of CD74 in CD3⁺CD4⁺ TCs observed by flow cytometry (Figs. 3h–j and S4e). The signaling communication network for ICAM was expressed differently from that of the MIF pathway. TCs and NK cells were the primary ligand sources in the ICAM signaling network, acting in both autocrine and paracrine manners (Figs. 3k, S4f). ICAM2/ITGB2 and ICAM2/SPN were the dominant contributors to the ICAM network (Fig. 3k), and pregnancy suppressed the ICAM signaling pathway mainly among CD4⁺ TCs, CD8⁺ TCs, and NK cells (Fig. S4f). Decreased expression of CD18 (encoded by ITGB2) by CD4⁺ TCs was validated by flow cytometry (Figs. 3l and S4g). BAFF is a prosurvival factor expressed by BCs [41]. BCs, and to a lesser extent, cDC2s were the sources of BAFF ligands and acted on BCs (Fig. S4h, i). We found that the BAFF receptor BCMA (encoded by TNFRSF17) was downregulated in BCs from the pregnancy group at both the RNA and protein levels (Fig. S4i–k).

Taken together, these findings revealed that specific intercellular interactions were affected by pregnancy. The reduced antigen presentation, antigen presentation-related signaling, immune cell activation signaling between APCs and TCs, and the decreased migration and multiple function-related signaling among lymphocytes may be the basis for reduction in the adaptive immune response during pregnancy.

3.5. Enhanced pro-inflammatory innate immune responses in pregnancy

In addition to the downregulated DEGs, we also analyzed the genes and pathways that were enhanced by pregnancy. We found that blood immune cells showed heterogeneous transcriptional changes during pregnancy based on the number of DEGs. VCCs were the cell type that was most affected by pregnancy, followed by BCs and MCs (Fig. 4a). Strikingly, we found that MCs had the most upregulated DEGs, whereas CD4⁺ TCs had the most downregulated DEGs. The number of upregulated DEGs in myeloid cell subsets was higher than that of downregulated DEGs, whereas the opposite was true for lymphocyte subsets (Figs. 4b and S5a). To explore the biological functions enhanced by pregnancy, we performed GO analysis on the upregulated DEGs for each immune cell population. Several pathways were upregulated during pregnancy (Fig. 4c). Among these pathways, inflammatory cytokine signaling, AP-1 signaling, aging, and cellular response to hormone stimulus pathways were more extensively enhanced in myeloid cells, whereas the regulation of gene silencing, chromatin assembly, and apoptosis pathways were more extensively enhanced in lymphocytes (Fig. 4c).

Fig. 4.

Preg enhanced the pro-inflammatory innate immune response in blood. (a) Rose diagram showing the numbers of DEGs in 5 functional immune populations. (b) The histogram showing the ratio of the upregulated number to the downregulated number among immune populations. (c) Dot plot showing the representative GO biological process and pathways enriched in upregulated DEGs in 5 immune populations. (d) Heat map showing DEGs of MCs across pregnant stages pseudotime. Gene expression levels across the pseudotime axis are normalized, and grouped by their functional categories (top). Group annotation of cells aligned along the pseudotime axis (bottom). (e) Violin plot showing the expression of genes related to apoptosis among immune populations between HC (blue) and Preg (red) groups. (f) Circle plot showing the upregulation in cellular interaction of IL-1 signaling predicted in iTALK tool (top). Violin plot showing the expression of IL1B among immune populations between HC (blue) and Preg (red) groups (bottom). (g) The histogram showing the representative GO biological process and pathways enriched in the DEGs both decreased in CD4⁺ TC and increased in MC.

We generated a Venn plot of upregulated DEGs and found that the expression of AP-1 pathway genes was increased in all immune cell types (Fig. S5b). We next studied pregnancy-upregulated DEGs in MCs and found upregulated genes associated with inflammation, aging, and hormone stimuli, such as IL1B, CXCL8, CDKN1A, and ANXA1 (Fig. S5b). Moreover, the expression of these genes gradually increased during pregnancy (Fig. 4d). As for circulating lymphocytes (particularly VCCs), we found increased expression of genes related to apoptosis (CYCS, CASP8) and gene silencing (ZFP36) during pregnancy (Figs. 4e and S5c). TCs apoptosis in the feto-maternal interface is vital for normal implantation [42], indicating a relationship between maternal circulating immunity and placental local immunity. In our analysis of cellular interactions among populations, we found that pregnancy induced an increase in the IL-1 signaling pathway function mainly through IL1B, which plays important roles in myeloid cell activation and inflammatory responses (Figs. 4f and S5d).

Given the different manifestations of pregnancy-associated DEGs in lymphocytes and myeloid cells, we next investigated the underlying mechanisms. The dampened TCs immune response in pregnancy may be compensated for by increased circulating inflammatory mediators, such as activated MCs and the complement system [43]. Thus, we analyzed the intersection of downregulated DEGs in CD4⁺ TCs and upregulated DEGs in MCs to study the potential mechanisms (Fig. S5e, f). GO analysis revealed that these genes were enriched in cytokine signaling, leukocyte proliferation, and leukocyte chemotaxis, driven by PIM1, ANXA1, CALR, LITAF, and CCL5, among others (Fig. 4g). PIM1 plays a critical role in the regulation of pro-inflammatory and prolabor mediators in fetal membranes [44,45]. Lipopolysaccharide-induced TNF-α factor (LITAF) promotes inflammation, obesity, and insulin resistance [46,47]. The other genes are involved in cytokine production (ANXA1) and migration (CALR and CCL5). These results supported the previously accepted view of the compensatory mechanisms between myeloid cells and lymphocytes and further revealed specific compensatory genes and pathways.

Collectively, these results suggested that although lymphocytes may undergo apoptosis and reduced function during pregnancy, an enhanced inflammation-associated phenotype was observed in myeloid cells. These divergent changes in the two classes of immune cells may result from compensatory mechanisms.

3.6. Identification of transcriptional dynamics across pregnancy stages

During different pregnancy stages, disease susceptibility varies in pregnant women, indicating the dynamic nature of the immune state during pregnancy. Thus, we explored transcriptional divergence among pregnancy stages. We first identified genes whose expression showed an increasing or decreasing trend with pregnancy stage. These genes were identified via Venn plots, which captured the simultaneously upregulated or downregulated DEGs in the ES group compared with the HC group and in the LS group compared with the ES group. We designated these genes as “trend genes” (Fig. 5a). MCs harbored the most upregulated trend genes, whereas VCCs and BCs harbored the most downregulated trend genes (Fig. S6a). Three upregulated trend genes, i.e., IRF1, JUND, and FOSB, which are all involved in the activation of inflammatory pathways, were shared by the five immune cell populations (Fig. S6b). We then analyzed trends in genes specific to different cell populations. In MCs, we found that almost all trend genes showed increased expression as pregnancy progressed. These genes were involved in the processes of inflammation (IL1B, NLRP3, HIF1A, PIM1), migration (CCL3, CCL3L1), and DNA damage (GADD45B, DDIT4; Fig. 5b). Consistent with these findings, high serum levels of IL-1β in pregnant women were amplified with increasing pregnancy stage (Fig. S6c). In VCCs, inflammation-related genes showed increased expression as pregnancy progressed, whereas cytotoxic genes showed decreased expression (Fig. 5c). Interestingly, the level of the activation marker CD69 was increased as pregnancy progressed in CD4⁺ TCs, as verified by flow cytometry (Figs. 5d, e and S6d, e). In BCs, TCL1A, SYK, and HLA-DQA1 levels gradually decreased as pregnancy progressed (Fig. S6f). Overall, these results suggested an increase in inflammation and decrease in cytotoxicity during pregnancy.

Fig. 5.

scRNA-seq showed the dynamic changes across pregnant stages. (a) Venn plot showing the analysis of “trend DEGs”. The overlapping regions indicate the upregulated trend DEGs (top) and downregulated trend DEGs (bottom). (b) Heat map showing the relative expression of trend DEGs of MCs across pregnant stages. (c) Heat map showing the relative expression of trend DEGs of VCCs across pregnant stages. (d) The flow cytometry histogram showing the expression of CD69 in CD4⁺ TCs. (e) Box plots showing the Mean Fluorescence Intensity (MFI) of CD69 in CD4⁺ TCs across pregnant stages. (f) Venn plot showing the analysis of “stage-related DEGs” in the overlapping regions. The histogram showing the number of stage-related DEGs in immune populations. Dot plot showing the relative expression of stage-related DEGs of BCs (g), cDC2s (h) and CD4⁺ TCs (i) across pregnant stages. Significance in e was calculated using an unpaired, two-tailed Student's t-test; ns, not significant, **P < 0.01, ***P < 0.001, ****P < 0.0001.

We next studied potential biomarkers or immune adaptations specific to different pregnancy stages. We identified stage-related genes that exhibited stage-specific expression patterns (Fig. 5f). According to the number of DEGs, MCs were most affected by the pregnancy stage, followed by VCCs and BCs. In VCCs, we found that DUSP2 and AP-1 family genes (JUND, FOS) were overexpressed in the MS group (Fig. S6g). In BCs, several BC signature genes (IGLC, IGHM) and ETS1 were overexpressed in the ES group (Fig. 5g). ETS1 is a key transcription factor that controls the development of BC autoimmune responses [48,49]. Notably, several other BC signature genes that were reduced in the ES group were recovered in the MS group. These genes were involved in BC activation (XBP1, BANK1, JCHAIN) and antigen presentation (HLA-DQB1 and HLA-DQA1; Fig. 5G). In cDC2s, most of the downregulated DEGs remained low in mid-term pregnant women, but increased in the LS group. Interestingly, these genes were primarily involved in antigen presentation and lymphocyte activation (Fig. 5h). In CD4⁺ TCs, several stage-related genes showed similar expression patterns. The decreases in PIM1, CD74, ANXA1, and CALR expression were partly increased in the LS group (Fig. 5i), suggesting that the reduced migration and activation of CD4⁺ TCs were partly recovered in late pregnancy. By contrast, DDX5, which is involved in gene silencing, decreased in the LS group. TXNIP, a vital component of cellular redox homeostasis and suppression of the effector TCs response [50], was increased in the ES and MS groups but decreased in the LS group (Fig. 5i). In addition, we found that the percentages of CD4⁺ effector memory TCs and CD4⁺ cytotoxic TCs were increased in the LS group (Fig. S6h). These results suggested that antigen processing and presentation, lymphocyte-activating functions of cDC2, and TC functions started to recover in the LS group.

Thus, our data suggested that inflammation was enhanced during pregnancy progression, whereas cytotoxicity tended to decrease. Notably, genes involved in adaptive immunity in DCs and CD4⁺ TCs declined in the early and middle stages of pregnancy and partially recovered in the late stages of pregnancy, indicative of a shift from immune tolerance to immune activation in late pregnancy.

3.7. Pregnancy showed immune changes similar to those of aging

Our data suggested that pregnancy not only attenuated adaptive immunity but also enhanced inflammatory activation and aging. We previously performed a substantial amount of work investigating the immune states of aging and found intriguing similarities in the immune alterations caused by pregnancy and aging. The relationship between aging and pregnancy is a topic of considerable interest; thus, we incorporated transcriptional data for peripheral blood from aged women in our research. Interestingly, both the pregnancy and aged groups showed decreased TCs and increased cMCs frequencies compared with the HC group (Figs. 6a, b and S7a).

Fig. 6.

Pregnancy induced transcriptional changes similar with aging. (a) The histogram showing the relative ratio of immune population among HC, Preg and AA groups. (b) The flow cytometry histogram (left) and box plots (right) showing the ratio of cMCs in immune cells among HC, Preg and AA groups. (c) Venn plot showing the analysis of “shared DEGs” shared by Preg and aging. The overlapping regions indicate the upregulated shared DEGs (top) and downregulated shared DEGs (bottom). (d) The histogram showing the number of aging DEGs and shared DEGs in immune populations. (e) Dot plot showing the representative GO biological process and pathways enriched in upregulated shared DEGs in 5 immune populations. (f) Venn plot showing the integrated comparative analysis of upregulated shared DEGs among 5 immune populations. (g) Box plots showing the protein level of IL-1β in serum of HC, Preg and AA groups. (h) The flow cytometry histogram showing the expression of CD83 in cMCs (left). Box plots showing the ratio of CD83-expressing cMCs among HC, Preg and AA groups (right). (i) Heat map showing the relative expression of aging-related genes across pregnant stages. (j) The flow cytometry histogram showing the expression of p21 in cMCs (left). Box plots showing the ratio of p21-expressing cMCs among HC, Preg and AA groups (right). Significance in b, g, h and j was calculated using an unpaired, two-tailed Student's t-test; **P < 0.01, ***P < 0.001, ****P < 0.0001.

Next, we identified DEGs between aged and HC groups (aging DEGs). To comprehensively interpret the relationships between pregnancy and aging, commonly upregulated or downregulated genes in aging and pregnancy were identified (shared DEGs; Fig. 6c). We then attributed the shared DEGs to each cell population and found that MCs and NK cells were among the populations most strongly affected by both pregnancy and aging based on the number of DEGs (Fig. 6d). Next, we conducted functional analysis of the shared DEGs and found that in MCs, both pregnancy and aging induced the AP-1 pathway, inflammation pathway, aging pathway, and response to oxidative stress, driven by genes such as FOS, IL1B, CD83, RETN, and DUSP1 (Fig. 6e, f). Consistent with these findings, high IL-1β levels in the serum and high CD83 expression in MCs of pregnant women were also observed in aged women (Figs. 6g, h and S7b). We then compared the expression of some aging-related markers in MCs across different groups. We found that, as in the aged group, pregnancy increased the expression of CDKN1A (p21), CDKN2D (p19), CDKN2A (p16), CDKN1C (p57), and STAT3 in MCs (Fig. 6i). The upregulation of p21 and STAT3 was validated by flow cytometry (Figs. 6j and S7c, d). For the downregulated bioprocesses, we found that pregnancy and aging shared similar patterns associated with decreased cellular defense response and leukocyte migration, mainly in NK cells (Fig. S7e). By generating a Venn diagram, we found that NK cells had the most downregulated DEGs shared between pregnancy and aging, including IL2RB, IFI30, CX3CR1, and KLRC2 (Fig. S7f).

Collectively, our functional comparative analysis of pregnancy and aging DEGs revealed that pregnancy and aging could induce many similar immune changes, including increased inflammation (mainly in MCs) and decreased cellular defense response and leukocyte migration (mainly in NK cells).

3.8. Partial restoration of the blood immune ecosystem in the postpartum period

During pregnancy, many autoimmune diseases go into remission and flare again in the early postpartum period. This phenomenon prompted us to hypothesize that restoration of pregnancy-induced immune changes may occur during the postpartum period. To test our hypothesis, we compared the proportions of each cell type between the HC, LS, and postpartum groups. The proportion of TCs decreased during pregnancy and increased postpartum (Figs. 7a, b and S8a). Conversely, the increase in CD14⁺ cMCs in pregnant women was partially rescued postpartum (Fig. S8a, b). These results were consistent with those of previous reports indicating that the postpartum period leads to a reversal of lymphopenia and myeloid cell expansion induced by pregnancy [22]. In addition, we found that the increased frequency of CD4⁺CD8⁺ double-positive TCs during pregnancy was reversed postpartum, which was confirmed by flow cytometry (Fig. S8c and d).

Fig. 7.

The immune adaptations in pregnancy were partially restored in postpartum period. (a) The histogram showing the relative ratio of immune population among HC, LS and PP groups. (b) The flow cytometry histogram (left) and box plots (right) showing the ratio of TC in immune cells among HC, LS and PP groups. (c) Violin plot showing the expression of inflammation-related genes among immune populations between LS (blue) and PP (red) groups. (d) Venn plot showing the analysis of “rescue DEGs”. The overlapping regions indicate the downregulated rescue DEGs (top) and upregulated rescue DEGs (bottom). The histogram showing the number of rescue DEGs in immune populations. (e) Dot plot showing the representative GO biological process and pathways enriched in upregulated rescue DEGs in 5 immune populations. (f) Dot plot showing the representative GO biological process and pathways enriched in downregulated rescue DEGs in 5 immune populations. (g) Venn plot showing the integrated comparative analysis of upregulated rescue DEGs among 5 immune populations. (h) The flow cytometry histogram showing the expression of PIM1 in CD4⁺ TCs (left). Box plots showing the ratio of PIM1-expressing CD4⁺ TCs among HC, LS and PP groups (right). (i) The flow cytometry histogram showing the expression of CD83 in cMCs (left). Box plots showing the ratio of CD83-expressing cMCs among HC, LS and PP groups (right). (j) Heat map showing the relative expression of aging-related genes among HC, LS and PP groups. For the box plot within each violin plot, middle lines indicate median values, boxes range from the 25th to 75th percentiles. Significance in c was calculated using two-sided Wilcoxon test as implemented in the function “compare_means” with default parameters; significance in b, h and i was calculated using an unpaired, two-tailed Student's t-test; **P < 0.01, ***P < 0.001, ****P < 0.0001.

To elucidate the molecular events occurring during the postpartum period, PP DEGs analysis was conducted between postpartum and LS groups for each major cell type. The expression levels of inflammation-related genes (FOS, DUSP1, IFITM2, TXNIP, PSME1, and CYBA) were decreased in all cell types in the postpartum group compared with those in the LS group (Figs. 7c and S8e). We further identified “rescue DEGs” as those whose expression was altered in the LS group compared with the HC group and showed the opposite changes postpartum (Fig. 7d). We found that CD4⁺ TCs had the most upregulated rescue DEGs, whereas MCs had the most downregulated rescue DEGs (Fig. 7d). Functional analysis of the PP DEGs and rescue DEGs revealed that multiple pathways decreased by pregnancy were upregulated postpartum, including antigen processing and presentation, lymphocyte activation, cytokine signaling, and the adaptive immune system, particularly in CD4⁺ TCs (Figs. 7e and S8f). Conversely, multiple pathways enhanced by pregnancy decreased postpartum, including the Janus kinase pathway, inflammatory pathway, aging and response to oxidative stress, mainly in MCs (Figs. 7f and S8g). These results suggested that some immune alterations during pregnancy, such as reduced adaptive immunity and enhanced inflammation and aging, were reversed postpartum.

To identify the postpartum rescue effects in specific cell types, we generated Venn diagrams of rescue DEGs. In BCs, we found that genes that were downregulated in pregnancy and upregulated postpartum were related to cell activation and antigen presentation. In CD4⁺ TCs, we found that the pregnancy-induced downregulation of some compensatory genes, such as CCL5 and ANXA1, were reversed postpartum. In addition, PIM1 and IL7R were rescued in the two populations (CD4⁺ TCs and VCCs) postpartum (Fig. 7g). Consistent with this observation, we found that the frequency of PIM1-expressing TCs was decreased in the LS group and increased postpartum (Figs. 7h and S8h). For the downregulated rescue DEGs, all populations had decreased levels of FOS and DUSP1 after delivery (Fig. S8i). The downregulated DEGs in MCs were associated with the Janus kinase pathway (PIM1), inflammation (CD83), aging (STAT3), and response to oxidative stress (Fig. S8i). The partly restored CD83 expression in MCs was validated by flow cytometry (Fig. 7i). Next, we further compared the expression of genes related to aging and oxidative stress in MCs among groups to explore the gene-specific rescued effects in postpartum period. Some of aging-related genes were downregulated after pregnancy, whereas others maintained high expression levels, like CDKN1B, JUN, JUND, and STAT3 (Fig. 7j). For the genes related to oxidative stress, several genes (DUSP1, JUN, MCL1, TNFAIP3) had similar trends (Fig. S8j). In addition, we also found some cell-specific patterns in other populations, such as the rescue of JUND and CDKN2D in CD4⁺ TCs and NFKBIA, FOSL2, and DDIT4 in VCCs (Fig. S8i). These analyses highlight the effects of pregnancy on specific immune cell types and showed that these changes may be partly reversed postpartum.

4. Discussion

The unique systemic immune states during pregnancy underlie an increased risk for some infections and mitigation of several autoimmune diseases in pregnant women [10], [11], [12], [13]. Here, we evaluated the immune states during pregnancy and across the three pregnancy stages. Data from aging adults and women in the postpartum period were also incorporated to provide insights into the maternal systemic immune system. The primary findings of this study were as follows: (1) VCC, namely the highly variable cytotoxic cells, were the T & NK cell types most affected by pregnancy and exhibited reduced expression of genes of cytotoxicity and cellular defense response in pregnancy. (2) Pro-autoimmune genes and processes of CD4⁺ TC and plasma cells were also reduced by pregnancy. (3) The decrease in MHC-II, CD40, ICAM, MIF, and BAFF signaling, which facilitates lymphocyte function, combined with increase in tolerogenic signaling (ICOS and activin) identified in cell-cell communication analysis, may underlie the above alterations in lymphocyte function. (4) We identified the genes and pathways related to inflammation and leukocyte migration upregulated in MCs, which may compensate for the dampened TC function in pregnancy. (5) As pregnancy progressed, we observed gradually decreased expression of cytotoxicity-related genes in VCCs, gradually increased expression of inflammation-related genes in MCs, and a shift from immune tolerance to immune activation in late pregnancy in DCs and CD4⁺ TCs. (6) Interestingly, pregnancy shared multiple immune changes with aging, indicating its potential pro-aging effect, which may not extend to the postpartum period.

Previous studies have reported altered systemic immunity during pregnancy [22,24,51,52]. However, these studies have mainly focused on one cell type, and many of the results have been contradictory. Utilizing scRNA-seq, we provided a comprehensive and integrative evaluation of the divergent changes in systemic adaptive and innate immunity. Our data supported that TCs are decreased [52], whereas MCs are increased [53] during pregnancy. In terms of gene expression, we observed various weakened functions in different types of lymphocytes. Notably, we showed that VCCs, which are characterized by high expression of cytotoxicity-related genes and play a critical role in the clearance of intracellular pathogen infections [54,55], were the cell type most affected by pregnancy and exhibited decreased expression of cytotoxicity- and defense response-related gene signatures. In this context, VCCs included CD8⁺ TCs, NK cells, and a small number of double-positive TCs. Double-positive TCs have been described in cancer, autoimmune diseases, and infections [56]; however, their roles in pregnancy are largely unclear. In our study, double-positive TCs were the most affected TC subtype based on their DEGs number, and their proportion was increased in pregnancy and restored postpartum, indicating potential value in pregnancy studies. By contrast, CD4⁺ TCs and plasma cells exhibited multiple downregulated gene signatures related to the autoimmune response, e.g., Th17 cell differentiation in TCs and protein processing in plasma cells, suggesting their weakened pro-autoimmune activity in pregnancy. These two aspects of dampened lymphocyte function, namely reduced cytotoxicity and pro-autoimmune activity may account for the enhanced risk of infections and decreased autoimmune attack during pregnancy.

Cell-cell interactions orchestrate single-cell functions [34], and this study was the first to identify the aberrant intercellular signaling underlying the abnormal changes in innate and adaptive immunity during pregnancy. Specifically, pregnancy reduced MHC-II, CD40, ICAM, MIF, and BAFF signaling, which is closely related to TC and BC function, and increased ICOS and activin signaling, two signaling pathways that have been reported to induce regulatory TCs via DCs [57,58]. The aberrant expression of these genes may contribute to pregnancy-induced regression of adaptive immunity. In addition, we found that pregnancy also induced the IL-1 signaling pathway among MCs and other cell populations. During pregnancy, MCs harbored the most upregulated genes and showed enhancement of the inflammatory phenotype during pregnancy. This activation of MCs has been reported to be induced by placental factors [51] and to act as a compensatory mechanism for the weakened TC response to strengthen the body's defense against pathogen attack [43]. However, a mechanistic explanation for this observation has not yet been reported. Integrative analysis of the DEGs simultaneously upregulated in MCs and downregulated in CD4⁺ TCs revealed the genes and enriched pathways that contribute to this compensatory mechanism. These genes were shown to be involved in cytokine signaling, inflammation, cell migration, and proliferation. Among these genes, PIM1 has been reported to regulate pro-inflammatory and prolabor mediators in fetal membranes [44,45], indicating that compensatory mechanisms are not limited to protection against pathogens but can extend to normal labor processes. Collectively, the above findings revealed the cooperative changes in gene expression and intercellular signaling in adaptive and innate immunity in the maternal systemic immune system, rather than a broad maternal immune suppression during pregnancy.

Pregnancy can be artificially divided into three stages [59]. Divergent physiological or pathological processes tend to occur in different stages, such as embryo implantation and placentation in ES [60], more severe symptoms of influenza in MS and LS [11], and parturition and preterm labor in LS [61]. These phenomena suggest that dynamic immune changes may also occur in these three pregnancy stages. In our study, we observed gradual upregulation of inflammation-related genes (PIM1, IL1B, NLRP3), particularly in MCs, and gradual downregulation of cytotoxicity-related genes in highly variable cytotoxic TC and NK cell subsets. A pro-inflammatory state is important for parturition [62], when circulating maternal leukocytes are recruited to the uterus to induce labor [51]. During this process, the number of CD14⁺ MCs in placental blood is significantly increased [63]. However, overactivation of MCs is related to pre-eclampsia [64,65]. Thus, circulating MCs and their phenotype may be a potential target or predictive factor for successful or pathological labor. In addition to the enhanced inflammation observed primarily in MCs during pregnancy, there is a shift from immune tolerance to immune activation in LS, as indicated by our scRNA-seq data. Analysis of the genes showing downregulation in ES and MS and upregulation in LS suggested specific functional recovery of cDC2s and CD4⁺ TCs. The DEGs identified in cDC2s indicated that the antigen-presentation function of these cells was partly recovered in LS, whereas CD4⁺ TCs exhibited restoration of cell activation. These findings highlighted the dynamic immune adaptations across the three pregnancy stages and may provide insights into divergent physiological or pathological changes at different stages.

Due to limited access to placental tissues during pregnancy, it is sometimes difficult to predict adverse pregnancy outcomes by monitoring local changes in the placenta. By contrast, peripheral blood is accessible even during very early pregnancy. Abnormal protein expression and cell abundances in serum samples are associated with specific pregnancy outcomes. Additionally, maternal serum samples can be used to predict gestational age based on the precisely timed transcriptional and proteinic changes that occur during pregnancy [66,67]. In our study, we observed the upregulation of apoptosis-related genes, including CYCS and CASP8, in TCs. Intriguingly, TC apoptosis has also been reported to occur at the feto-maternal interface and to contribute to normal implantation [42,68]. Similarities between immunity in the periphery and placenta were also observed in cell-cell communication signaling. ICOS and activin, two signaling molecules expressed by the systemic immune response during pregnancy, have also been reported to mediate immune tolerance in the placenta [37,38]. Inhibition of the ICOS pathway increases fetal resorption and decreases fetal survival in pregnant mice [37]. Moreover, consistent with the enhanced inflammation and the shift from adaptive immune regression to activation shown in systemic immunity in the LS, physiologic uterine inflammation and immune activation are critical for labor onset [69]. These findings suggest specific links and crosstalk between the maternal peripheral immune system and the local immune system of the feto-maternal interface, which may enable the identification of placental processes and pathology based on changes in peripheral blood. Future studies may integrate analyses of systemic immunity and local immunity to reveal the detailed peripheral-local immune connection in pregnancy.

Aging causes broad changes in the immune system and increases susceptibility to infections, cancer, and other diseases [70,71]. Conversely, many autoimmune diseases exhibit relatively low incidence rates in the elderly population [72]. We have demonstrated enhanced inflammation and reduced Th17 cell responses to autoimmune antigens in aging [28,29,73]. In this study, we elucidated the immunological mechanisms through which pregnancy promotes aging by comparing the immunological characteristics of pregnancy and aging. Further investigation of the genes upregulated both during pregnancy and aging revealed that MCs were the common and dominant mediators of inflammation in these two complex physiological processes. Several aging-related markers were also upregulated in MCs. Similarly, analysis of co-downregulated genes showed that pregnancy and aging shared similar decreases in cellular defense response and leukocyte migration, mainly in NK cells. Aging and pregnancy may also involve similar processes with regard to induction of and response to oxidative stress, inflammation, and telomere degradation [74]. These similarities may result from stressors and damage-induced metabolic changes in both aging and pregnancy [74]. Specific mechanisms of how pregnancy induced aging-related immune changes require further researches. Notably, we observed that the enrichment of aging-related signaling was rescued postpartum and that normal expression of upregulated genes related to inflammation, oxidative stress, and aging in MCs was partly restored postpartum. These results indicated that aging-like immune changes induced by pregnancy could be partly reversed postpartum and that antioxidant agents administration may promote this restoration after delivery.

There were some limitations to this study. First, there were fewer samples from the HC group than from the pregnancy groups in our single-cell analysis. Moreover, the number of samples used for each pregnancy stage was not sufficient to evaluate statistical significance in our single-cell analysis. Finally, the current study was a cross-sectional study, and blood samples from a single woman during pre-pregnancy and different pregnancy stages were not available.

5. Conclusion

In summary, we evaluated the dynamic immune landscape using PBMCs from pregnant women during three pregnancy stages. During pregnancy, a proportional and functional reduction in lymphocytes, particularly a decrease in the cytotoxic phenotype, was observed. The aberrant cell-cell communication among APCs and lymphocytes may underlie these changes. Conversely, MCs exhibited upregulation of the inflammation-related phenotype and may compensate for the dampened TC response. We also demonstrated the dynamic immune adaptations specific to each of the three trimesters. Based on multiple similarities in the immune changes induced by pregnancy and aging, we proposed that pregnancy may promote aging and that these effects may be reversed postpartum. Our findings may establish a basis for further exploration of the complex immune mechanisms underlying the physiological and pathological changes during pregnancy.

CRediT authorship contribution statement

Xiuxing Liu: Formal analysis, Methodology, Writing – original draft. Lei Zhu: Formal analysis, Writing – original draft. Zhaohao Huang: Formal analysis, Writing – original draft. Zhaohuai Li: Methodology, Writing – original draft. Runping Duan: Methodology, Writing – original draft. He Li: Resources, Supervision. Lihui Xie: Resources, Supervision. Xiaozhen Chen: Resources, Supervision. Wen Ding: Resources, Writing – original draft. Binyao Chen: Resources, Writing – original draft. Yuehan Gao: Resources, Writing – original draft. Juan Su: Visualization. Xianggui Wang: Visualization. Wenru Su: Visualization.

Biographies

Xiuxing Liu earned his M.B. at the Medical School of Nanchang University, and M.D. at Zhongshan Ophthalmic Center, Sun Yat-sen University. Now he is a Ph.D. student at Zhongshan Ophthalmic Center, Sun Yat-sen University. His research focuses mainly on immunology and autoimmune diseases.

Wenru Su earned his M.B. at Bengbu Medical School, M.D. and Ph.D. at Zhongshan Ophthalmic Center, Sun Yat-sen University, and he did his postdoctoral studies at University of Southern California (USC). Now he is a professor at State Key Laboratory of Ophthalmology, Zhongshan Ophthalmic Center, Sun Yat-sen University. His research focuses on the pathogenesis of autoimmune and allergic diseases, especially for ocular diseases.