Single‐cell RNA sequencing reveals heterogeneity and differential expression of decidual tissues during the peripartum period

The decidua is a tissue that contacts both maternal and foetal components and is pivotal to labour onset due to its location. Due to the heterogeneity of decidual tissue, it is challenging to study its role in the peripartum period. Herein, we analysed the transcriptomes of peripartum decidua at single‐cell resolution.


| INTRODUC TI ON
Abnormal onset of labour is a direct cause of several adverse pregnancy events, especially preterm birth. [1][2][3][4] As a pregnancy-specific tissue, the maternal-foetal interface is essential for delivery. [5][6][7][8][9] Decidua is one of the central tissues found within the interface that contacts both maternal and foetal components; it is a tissue with high heterogeneity. Decidua is the target of decidualization during implantation 10 and plays an important role in labour onset. 11 Various cell types in decidual tissue, such as decidual stromal cells and T cells, have been uncovered by classical cell morphology classification, flow cytometry and other methods. The recent development of single-cell RNA sequencing has resulted in a more comprehensive understanding of tissue components. Published studies on the placenta in the first and second trimesters have revealed their role in preeclampsia and other pathological pregnancies. [12][13][14] Studies on the third trimester have revealed insights on the differentiation of placental trophoblasts, placental cellular dynamics and the interactions between placental trophoblast cells and decidual stromal cells. 15,16 Importantly, these studies also found that the decidua was at the centre of intercellular signal transduction. These studies demonstrate the essential role of the decidua throughout pregnancy, as well as the importance of studying the cell types, cell ratios and cell heterogeneity of the decidua for understanding labour onset.
Here, we performed single-cell RNA sequencing of the decidua before and after delivery, and identified the major cell populations and subpopulations. The functions of different cell types and subtypes during labour onset were evaluated by bioinformatics. The results revealed decidual cell changes during labour onset. Overall, the results from this study provide the basis for further research.  Table S1.

| Decidual tissue dissociation
According to the general steps of making single-cell suspension, [12][13][14] for term pregnancy (labour onset), three decidual samples were obtained after delivery and washed in phosphate-buffered saline (37°C). Similarly, three samples before delivery were obtained during cesarean section (without labour). Subsequently, the samples were dissociated with tweezers and scissors, and a singlecell suspension was made after dissociation, digestion (collagenase Sigma-Aldrich, and phosphate-buffered saline, 37°C, 45 minutes) and filtration (Falcon 40-μm cell strainer; Corning). Then, the cell viability assessment was performed before the next step.

| Single-cell RNA-sequencing data analysis
Cell Ranger 2.0.1 was used to analyse the data. The data from different samples were merged through a Cell Ranger aggr pipeline and normalized by equalizing the read depth. Principal component analysis and t-distributed stochastic neighbour embedding were performed using the R t-distributed stochastic neighbour embedding package of r software. According to the guide, output and presentation files were displayed and analysed by the Cell Browser 2.0.0 (10 × Genomics).

| Gene ontology and pathway analysis
For the output gene list of each cluster, we analysed it through Metascape. 18 KEGG pathway, gene ontology biological processes, reactome gene sets, canonical pathways and CORUM were used for the pathway and process enrichment analysis. Based on the accumulative hypergeometric distribution, the P-values were calculated.

| Pseudotemporal ordering analysis
We used the Monocle package (r) for pseudotemporal ordering analysis. Genes that met the following standards were selected: gene expressed in no less than 10 cells, average expression value greater than 0.5, q value of the differential expression analysis less than 0.01 and dispersion value greater than or equal to the expected dispersion value. After selection, the multidimensional space was reduced to two-dimensional space, and then, the single cell was sorted (Monocle).

| Construction of cell spectrum of decidual tissue during the peripartum period
All of the decidual tissue samples were collected for single-cell sequencing ( Figure 1A). A total of 29 231 cells were detected and divided into 16 cell clusters, with different relative unique molecular identifier value per cell ( Figure 1B We analysed the expression of genes found specifically in the decidual stromal cell population and found that the main functions of decidual stromal cells at this stage included extracellular matrix organization, protein processing and cell-substrate adhesion. Through single-cell RNA sequencing, we also found that although decidual stromal cells may perform the above functions, there still exists intra-group heterogeneity ( Figure 2M,N). In addition to extracellular matrix organization and cell-substrate adhesion, the characteristic genes expressed by cluster 2 were found to be mainly involved in regulation of insulin-like growth factor (IGF) transport and muscle structure development ( Figure 2N). The difference was that cluster 6 genes focused on processes including protein processing in the endoplasmic reticulum and regulation of hormone levels. We also noted that this cluster expressed a higher level of classical decidual suggested extravillous trophoblasts had a foetal source. Using our data, we found five extravillous trophoblast subgroups with different markers: cluster 5 highly expressed ALPP, cluster 7 highly expressed FABP7, and cluster 12 highly expressed DSG1 ( Figure 3D-H).
Trophoblasts regulate invasive functions such as neutrophil activation and cell adhesion. The high heterogeneity of extravillous trophoblasts may be one of the reasons for its multiple functions.
Specifically, the top gene expressed by cluster 5 indicated that this cluster might be involved in reproductive structure development and epidermal development ( Figure 3I). The main function of cluster 11 (placenta development, vasculature development) indicated that this cluster was primarily involved in the establishment and development of placenta. Considering the gestational week (more than 37 weeks) of the samples, this suggested that this group may be involved in placenta formation from the first to third trimesters. In particular, as the smallest extravillous trophoblast group, cluster 14 has mainly two functions, DNA replication and cell division, indicating that this was a subgroup of cell proliferation ( Figure 3I).
We can identify the T-cell group based on the expression of CD3 (CD3E) and CD3D ( Figure 3J,K). We found that T cells had two sets of clusters, which were not strictly classified in the classical way (eg, CD4+ TC and CD8+ TC), but had unique transcriptome differences.
The two subgroups (clusters 10 and 15) of T cells can be distinguished by RRM2 (cluster 15+ and cluster 10−; Figure 3L).

As immune cells, the main functions of T cells are lymphocyte
activation and the adaptive immune response. Cluster 10 focused on T-cell activation, CD8+ T-cell receptor response pathways, and may be a cell population that primarily functions in the immune response ( Figure 3M). We observed that cluster 15 cells were significantly different from those of cluster 10, which consisted of a significant proportion of cells at various stages of the cell cycle, including mitosis, presumably related to T-cell proliferation ( Figure 3M).

| Change in cell number and proportion in the decidua during the peripartum period
Analysis of 29 231 cells (eight types, 16 clusters) in our study showed that the top three cell types were endothelial cell (10 004, 34.22%), decidual stromal cell (6422, 21.97%) and smooth muscle cell (3720, 12.73%; Figure 1B, Table S2). In the first trimester, decidual leucocytes accounted for more than 40% of all cells; endothelial cells were few in number. 14 During the peripartum period, the proportion of decidual endothelial cells increased greatly and that of fibroblasts decreased. The proportional change in decidual stromal cell numbers was relatively small. Compared to the first-trimester data, 14 the proportions of certain cell types varied greatly during the peripartum period ( Figures 1B and 4, Table S2).

| Heterogeneity of expression profiles of different cell types during the peripartum period
Next, we analysed transcriptome changes during labour onset. We found that not only did the expression profiles of different cell types vary unequally, but also one of the subtypes was also inconsistent. The Our results showed that the functions of T cells were lymphocyte activation and the adaptive immune response during this period.
Unexpectedly, the results of cluster 15 were significantly different from that of cluster 10, which consisted primarily of cells in mitosis and the cell cycle ( Figure 6 and Figure S2). We speculated that this phenomenon might be related to T-cell proliferation. Although the role of decidual immune cells in the maternal-foetal interface has not been fully elucidated, our results suggest that heterogeneity exists in the T-cell population.

| Differences in cell cycle and proliferation in perinatal decidual tissue
To evaluate cell-cycle differences in perinatal decidual tissue, we analysed the expression of cell-cycle genes. 26 Cells in S phase were mainly concentrated in clusters 4 (a small part of smooth muscle cell), 14 and 15 ( Figure 6A). These clusters (cluster 4 smooth muscle cell, cluster 14 extravillous trophoblast and cluster 15 T cell) also highly expressed G2/M phase cell-cycle genes ( Figure 6B). These results suggest that cells in these clusters may have shorter cell cycles or more active proliferation. Compared to other clusters, the cells in these three clusters were more proliferative and may not be affected by labour. During pregnancy, the immune system is functional and highly active. 27 Among active decidual cells, T cells make up the largest proportion, suggesting a high degree of immune cell proliferation during the perinatal period.

| Pseudotemporal ordering reveals a differentiation relationship between major cell types
Next, we performed pseudotemporal ordering of decidual stromal cells ( Figure S3A-C). Our results showed an evolutionary trajectory of decidual stromal cell subtypes and a differentiation relationship with fibroblasts ( Figure 7A,B). We found a differentiation relationship from cluster 2 to cluster 6 by decidual stromal cell pseudotemporal ordering and believe cluster 6 may be at the end of differentiation ( Figure 7A,B and Figure S3A,B). When FBs were included in our analysis, the new trajectory was the direction from fibroblast to decidual stromal cell ( Figure 7A,B). Along this trajectory, the expression of CFH and TFPI gradually increased ( Figure 7C). We divided the genes into 6 'clusters' to classify genes that followed similar dynamic trends in the trajectory, and found that the expression of most genes gradually decreased ( Figure 8A). In 'cluster' 3 of Figure 8A, the decreasing genes were involved in processes including extracellular structure organization, inflammatory response and cell junction organization. This change reflected the process of functional evolution or the change from fibroblast to decidual stromal cell. An analysis of branch events revealed differential genes associated with branches ( Figure 8B). We found that IGFBP1 ( Figure 8B) was one of the key genes for cell fate 2, which was related to the decidual stromal cell fate.
Among extravillous trophoblasts, cluster 11 to cluster 12 followed a similar pseudotemporal ordering ( Figure 7D,E and Figure S3D,E). These results suggest the associations among these subgroups. Along the trajectory, the expression of TIMP2 showed a gentle downward trend ( Figure 7F). The high expression of TIMP2 in decidual extravillous trophoblasts is important for matrix remodelling and controlled trophoblast invasion during placentation. 28 Our results suggest that the changes in extravillous trophoblast subgroups occur during invasion during this period.

| D ISCUSS I ON
The maternal-foetal interface is complex; the various cells involved can be identified via single-cell sequencing. Prior single-cell sequencing papers studied placental and decidual heterogeneity during early pregnancy. [12][13][14] However, pregnancy is a period with changes, and

F I G U R E 5
Comparison of the functions of upregulated genes among some cell populations after delivery. A, Heat map of enriched terms across upregulated genes of decidual stromal cell subgroups after delivery. Log10(P) is the P-value in log base 10. B, Network of enriched terms in decidual stromal cell subgroups, coloured by cluster ID, where nodes that share the same cluster ID are typically close to each other. C, Overlap between genes of decidual stromal cell subgroups, including the shared term level, where blue curves link genes that belong to the same enriched ontology term. The inner circle represents gene lists, where hits are arranged along the arc. Genes that hit multiple lists are coloured in dark orange, and genes unique to a list are shown in light orange. D, Heat map of enriched terms across upregulated genes of extravillous trophoblast subgroups after delivery. After delivery, the other two clusters had no significantly upregulated genes or the number of upregulated genes was not sufficient for functional analysis. Log10(P) is the P-value in log base 10. E, Overlap between genes of extravillous trophoblast subgroups, including the shared term level, where blue curves link genes that belong to the same enriched ontology term. The inner circle represents gene lists, where hits are arranged along the arc. Genes that hit multiple lists are coloured in dark orange, and genes unique to a list are shown in light orange. Cluster 12 of extravillous trophoblast had no significant upregulated gene after delivery. DSC, decidual stromal cell; EVT, extravillous trophoblast  protein 1 and interleukin-6 signalling pathways. It is worth noting that the role of activator protein 1 and interleukin-6 signalling in labour has been confirmed. [38][39][40][41] As one of the extravillous trophoblast subgroups (cluster 12) was not active during labour, its function in delivery is unclear. These results suggest that extravillous trophoblasts may participate in labour by regulating immune tolerance and F I G U R E 8 Differential gene expression (DSC clusters 2, 6 and 13; and FB cluster 9) heat map of pseudotemporal ordering in decidua. A, Differential gene cluster heat map (for DSC clusters 2, 6 and 13; and FB cluster 9). Each column represents the average expression in the current cell state. The 'cluster' in this figure represents a cluster of genes with similar dynamic trends, which is different from the 'cluster' of cells in this article. B, Differential gene pseudotemporal expression trajectory map (for DSC clusters 2, 6, and 13; and FB cluster 9). The arrow indicates the expression of IGFBP1 gradually increased along the trajectory of cell fate 2. DSC, decidual stromal cell; FB, fibroblasts the activation of multiple signalling pathways. There is no doubt that the participation of various extravillous trophoblast subtypes in the initiation of labour onset is a complex process that warrants needs further investigation.
Among decidual stromal cells, which are characteristic decidual cells, we found that clusters 2 and 6 were mainly involved in labour onset. In general, these subgroups were involved in the activator protein 1 pathway, suggesting that decidual stromal cells in delivery may be regulated through this pathway. This pathway is also related to the activation of smooth muscle cells. 42 Although smooth muscle cells are currently recognized as effectors during delivery, the exact mechanism is still unclear. 43  Extravillous trophoblasts with high expression of HLA-G are related to immune tolerance, and in the perinatal period, the expression profile of decidual T cells that regulate immunity is heterogeneous.
Our results showed that most T cells perform immune responses, while a small number retain their ability to proliferate. In normal pregnancy, early decidual CD8+ T cells have both activation and dysfunction. 44 As pregnancy progresses, the reactivity of maternal-foetal antigen-antibody gradually changes and adapts to each other. We discovered differences in the proportions of different functional subgroups of decidual T cells. This difference may be closely related to the foetal-maternal immune tolerance response during labour onset.
During pregnancy, T cells maintain the maternal-foetal tolerance, thus a normal pregnancy. 45 Decidual T-cell numbers change during delivery, perhaps in an effort to maintain homeostasis. 46 Dendritic cells regulate T-cell responses; T cells interact via a complex web during pregnancy. 47 Decidual immune cells may thus play important roles in pregnancy maintenance and labour onset.
Data on early pregnancy have shown the existence of pseudotemporal ordering from fibroblasts to decidual stromal cells, 14 which is consistent with our results. Cluster 6 may be the end of the ordering (cell fate 2 in Figure 8B) in decidual stromal cells. According to the ordering, the expression change in complement factor H (which can inhibit CD47-mediated resolution of inflammation) in the pseudotemporal ordering suggests the functional heterogeneity of decidual stromal cells. 48 TFPI has been implicated in angiogenesis-related processes 49 ; this was similar to the function of cluster 6 after labour onset ( Figure 5A). The data indicate that the pseudotemporal ordering might be preparation for labour onset. In the extravillous trophoblast ordering, TIMP2 expression showed a relatively slow decreasing trend ( Figure 7F), suggesting heterogeneity of extravillous trophoblast subgroups. Previous studies have shown that TIMP2 is associated with homeostatic regulation of extracellular matrix remodelling in labour onset. 50 We used single-cell sequencing to study peripartum decidua. Further work is required; this is a limitation of our research. Highlevel expression of cell cycle-relevant genes in certain cell populations may indicate that these cells can proliferate. However, cell-cycle signals may mask non-cell cycle differences, which may need further study. We sampled only six pregnant women; sampling error may be in play in terms of cell type/subtype identification and RNA expression patterns. Therefore, more comprehensive researches may help to solve this issue.
In conclusion, our study used single-cell RNA-sequencing technology and revealed new information on the peripartum decidual atlas and cell proportions. Furthermore, we revealed heterogeneity in the subgroups and found differential gene expression and functional changes. The various cell types participate in labour to varying degrees, demonstrating the complexity of cell networks that form the decidual tissue system associated with delivery. In summary, our results provide a new perspective for the study of delivery, and provide important single cell-based evidence for detailed studies on the decidua during the peripartum period.

CO N FLI C T O F I NTE R E S T
The authors declare no conflicts of interest.

DATA AVA I L A B I L I T Y S TAT E M E N T
The data that support the findings of this study are available from the corresponding author on reasonable request.