Colony‐stimulating factor 1 positive (CSF1+) secretory epithelial cells induce excessive trophoblast invasion in tubal pregnancy rupture

Abstract Tubal ectopic pregnancy (TEP) occurs when an embryo aberrantly implants in the fallopian tube, leading to abortive or ruptured tubal ectopic pregnancy (AEP or REP). Poor outcomes of REP include maternal infertility or mortality. Current studies on the prevention and treatment of ruptured tubal ectopic pregnancy (REP) are unfortunately hampered by a lack of the cell spectrum and cell–cell communications in the maternal–foetal interface. Here, we investigate the mechanisms of tubal rupture through single‐cell transcriptome profiling of the fallopian tube‐trophoblast interface in REP, AEP and intrauterine pregnancy patients. In REP, extravillous trophoblast (EVTs) cells form a dominant cell population, displaying aggressive invasion and proliferation, with robust differentiation into three subsets. Cell communication analysis identified colony‐stimulating factor 1 (CSF1), overexpressed by fallopian tube secretory epithelial cells in REP, with CSF1R on EVTs and macrophages, as a ligand/receptor pair that stimulates EVT invasion and macrophage accumulation. CSF1+ secretory epithelial cells stimulate EVTs migration and invasion, leading to a tubal rupture in REP. These results provide a mechanistic context and cellular milieu leading to tubal rupture, facilitating further study and development of therapeutics for REP in early pregnancy.


| INTRODUCTION
Ectopic pregnancies (EPs) result from embryo implantation at sites other than the uterine endometria (intrauterine pregnancy, IP), 98% of which occur within fallopian tubes. 1 This can lead to infertility and/or even death in affected patients, and accounting for 75% of early pregnancy-related deaths. [2][3][4] In clinical practice, the actual morbidity rate may be even higher. 5,6 Upon implantation in the fallopian tube, tubal EPs (TEPs) can evolve into either abortive or ruptured EPs (hereafter termed AEP and REP, respectively). Both are diagnosed and confirmed by a combination of beta-subunit human chorionic gonadotropin (β-hCG) dynamics, ultrasonography and laparoscopy. 7 AEPs are typically characterized by low and gradual declining β-hCG dynamics and haemodynamic stability with intact fallopian tubes and are therefore monitored at home or treated with methotrexate as necessary, a widely used and cost-effective intervention that preserves patient fertility. 7 By contrast, REP, the most common cause of maternal mortality among all tubal EPs, is commonly characterized by elevated β-hCG levels, an active foetal heart, and haemodynamic instability. 8,9 REPs thus require more invasive management, such as emergency laparoscopic resection of the fallopian tube and foetus. While the factors associated with increased risk of tubal rupture have been intensely investigated, such as maternal age (≥35), implantation in the isthmus and >5000 IU/L of β-hCG, 9,10 it remains uncertain whether these risk factors in REP are causal, due a lack of mechanistic understanding of the distinct events leading to AEP and REP. [11][12][13] The human placenta is a highly specialized and multifunctional organ that is essential for foetal growth and survival. Three placentaspecific cell types are responsible for its primary functions, including villous cytotrophoblasts (vCTBs, the stem cell population in trophoblasts), syncytiotrophoblasts (STBs), and extravillous trophoblasts (EVTs). 14 VCTBs further differentiate into two lineages that either fuse to an external layer of multinucleated STB or undergo an epithelialmesenchymal transition to give rise to EVT cells. 14 After implantation, EVTs display a migratory and invasive phenotype, ensuring the success of the pregnancy by anchoring the placenta to the uterus.
The aberrant regulation of EVT invasion can lead to a series of obstetric syndromes. Incomplete invasion is associated with preeclampsia, foetal growth restriction (FGR) and stillbirth, 15,16 while excessive invasion results in placental accreta and choriocarcinoma. 17 Histopathology studies have demonstrated that tubal rupture is positively correlated with the depth of EVT invasion. 18,19 Invasion of the fallopian tube wall by EVT can potentially compromise its structure and disrupt its functions in pregnancy by inducing an inflammatory response. 18 In IP, several factors secreted from decidua, uterine smooth muscle cells and decidual natural killer (dNK) cells act as the key regulators of EVT function. 17,20 In addition, macrophages are the predominant immune cells at the maternal-foetal interface in EP, and the classically activated phenotype (Macro1) promotes trophoblast apoptosis and inhibits trophoblast invasion, resulting in tubal abortion. 21,22 Thus, an imbalance in trophoblast invasion provides obvious evidence of the pivotal regulatory contribution of the maternal-foetal interface microenvironment. 23 Under certain physiological conditions, the motion of tubal cilia, smooth muscle contraction and tubal secretory fluids play a central role in tubal transport. Altered cilia function, a perturbed chemotactic tubal microenvironment and dysregulation of migration by the fertilized egg can all result in stranding the embryo in the tube, and consequently, tubal pregnancy. 24,25 However, after implantation, the cellular interactions within the fallopian tube that regulate EVT functions and the mechanisms leading to different pregnancy outcomes are still unclear.
Here, we conducted single-cell sequencing to comprehensively determine the cell states and subtypes involved in the development of tubal rupture and maternal-foetal communication in the narrow space of fallopian tubal walls. Bioinformatic analyses were used to evaluate the functions of different cell types and subtypes, and generate detailed molecular and cellular maps of the human fallopian tubalplacental interface. These analyses provide an in-depth understanding of tubal rupture at the single-cell level and can help guide the development of novel therapeutic strategies to combat the aetiology of REP.

| scRNA-seq data analysis
Mapping to GRCh38 human genome, quality control and read count-

| Differential expression and enrichment analysis
The Seurat FindAllMarkers function performed marker genes expression analysis for each cluster using the Wilcoxon rank sum test (adjust p-value <0.05 and fold-change threshold >1.2). The list of DEGs (differential expressed genes) per cluster then were subjected to functional enrichment analyses using gene ontology

| Cell-cell communication analysis
Number of significant ligand-receptor pairs between any pair of two cell clusters was analysed by the method of CellChat. 30

| Explant culture
Explants were cultured according to a previously published protocol. 31 The placenta tissue was dissected out pieces of villus tips (2-3 mm)

| Macrophage subsets differ between AEP and REP patients
The implantation immune microenvironment is a complex, multilevel interaction network that provides major contributions to regulating EVT function. Identification of immune-related cells using known marker genes revealed that macrophages/monocytes were the dominant cell type in both AEP and REP, while NK cells accounted for a relatively small population (Figures 4A and S3A). [32][33][34] Moreover, the proportion of macrophages/monocytes was higher in REP patients than in AEP ( Figures 4B and S3B-3C). Volcano plots of differentially expressed genes (DEGs) in macrophages ( Figure S3D) and GO analysis of up-regulated DEGs genes between the REP and AEP groups ( Figure 4C) demonstrated that macrophages in REP patients were  20 an M1 subset with high S100A8/S100A9 expression, similar to the Macro1 subset reported in IP, 35 and an M2/3 subset with high CD206 expression similar to the alternatively activated phenotype (Macro2) subset in the IP 34 (Figures 4D S3G, H).
Based on scRNA data obtained from normal fallopian tubes (ampullar), this subcluster of macrophages was only found in the fallopian tube of a tubal pregnancy (including both REP and AEP) ( Figure S3H, I), 36 suggesting that the presence of macrophage subsets in TEP was related to pregnancy ( Figure S3H, I). These overall observations suggested that the M1 subgroup, which was dominant in AEP, could most likely promote an inflammatory response and inhibit trophoblast invasion ( Figure 4E, Figure S3E, F, Table S2). In contrast, the M2/3 subgroups, dominant in REP patients, potentially functioned in maintaining the microenvironment balance and conceptus ( Figure 4E, Figure S3E, F, Table S2). At the maternal-foetal interface in EP patients, high

| Accumulation of M2 macrophages and EVT invasion induced by a CSF1/CSF1R regulatory axis in FTSECs
To explore the association between maternal epithelia and EVT, we ( Figure 5A). The latter of these subtypes, EPPK1+ undefined epithelia, was specific to AEP patients and enriched with feature genes related to phenotypes, such as epithelial proliferation, migration and development ( Figure 5B, Table S3). Ciliated cells were enriched with genes related to normal tubal cilia function and the HPlEpC subset was comprised of foetal epithelial cells ( Figure 5B, Table S3). Among these clusters, FTSECs showed enrichment for feature genes involved in immune response and lymphocyte activation, as well as cell adhesion, suggesting that these cells participated in EVT invasion ( Figure 5B, Table S3). FTSECs accounted for more than 30% of all epithelial cells in REP, but only 6.7% in AEP samples ( Figure 5C, Figure S4A). In light of the remarkably high proportion of these cells, we performed immunostaining for OVGP1, a secretory epithelial marker and HLA-G, an F I G U R E 6 Legend on next page.
EVT marker, revealed that FTSEC populations were smaller in AEP patients compared with REP patients and EVT was located under the intact FTSECs in REP ( Figure 5D).
Aside from epithelial cells, the fibroblast/stroma/endothelia cells were also identified in nine cell types ( Figure S4B-4C). The F3/F4 cells were dominant in the AEP group, and DEGs enriched in this population were largely related to neutrophil activation ( Figure S4D), suggesting that the F3/F4 subsets were likely responsible for actively promoting granulocyte accumulation in AEP patients ( Figure S4D).
Collectively, these results implied that cytokines and chemokines secreted from FTSECs could interact with macrophages and EVTs at the maternal-foetal interface, stimulating macrophage proliferation and potentially modulating EVT function via transcriptomic reprogramming.
CellChat analysis was conducted to explore this possibility, which identified FTSECs as the predominant intercellular communication "hub," with secreted signals from these cells predicted to interact with both macrophages and EVTs ( Figure S5A). To further investigate the mechanism of cell-to-cell communication, we applied CellPhoneDB to identify interaction pairs that were enriched in EP ( Figure 6A). In IP, the colony-stimulating factor-1 (CSF1) is synthesized in high concentrations by the endometrium during pregnancy and targeted to CSF1R of trophoblast. 38,39 Many studies have revealed that CSF1 regulates the proliferation, differentiation and survival of macrophages, 40  and CSF1R, implying that they could interact with FTSECs ( Figure 6C).
In addition, FTSECs in REP samples displayed higher CSF1 expression than in AEP samples ( Figure S5B, C), while its receptor, CSF1R, was expressed at similar levels by EVTs in both pregnancy types ( Figure S5D). Moreover, the FTSECs in AEP were enriched with genes (BID, EIF5A and HIF1A) related to apoptotic signalling ( Figure S5B). [47][48][49] FTSECs were almost completely absent at the fallopian tube-trophoblast interface in AEP patients, while those that were detected expressed CSF1 at low levels ( Figure 5D, Figure S5B, C). The overexpression of CSF1 in the FTSECs of REP suggested a possible role in regulating EVT proliferation and invasion.
To investigate whether the CSF1/CSF1R ligand-receptor pair indeed played a role in promoting EVT proliferation and invasion, we conducted functional assays using HTR8-SVneo cells and villous explant cultures. Previous studies showed that upregulation of CSF1 expression can induce a dose-dependent increase in trophoblast proliferation. 42 To test this possibility, we applied GW2580, a CSF1R antagonist that blocks CSF1-induced proliferation, to cells in in vitro wound closure assays ( Figure 6D), transwell assay ( Figure 6E) and villous explant cultures ( Figure 6F). The results indicated that exogenous CSF1 could obviously stimulate the migration and invasion of HTR8-SVneo, while exposure to GW2580 inhibited their migration and invasion ( Figure 6D,F, Figure S5E). QPCR-based relative expression assays indicated that stimulation with exogenous CSF1 (100 ng/mL) significantly increased MMP2 expression in the HTR8-SVneo ( Figure 6G). These cumulative results indicated that high CSF1 levels could likely increase MMP2 expression, thereby promoting EVT invasion. In addition, after inducing THP-1 monocytes to differentiate into macrophages, culturing these macrophages in a medium containing purified CSF1 resulted in their polarization towards a CD206 + CD80-Macro2-like phenotype ( Figure S5F) The decidua spongiosa contain hyper-secretory glands which provide histotrophic nutrition to support early embryonic growth. 20 54 These results suggest that further mechanistic studies are needed to determine how VEGFs function in modulating EVT3 invasion. Moreover, the initiation of tubal rupture is a complex process that requires contributions from several EVT subpopulations, thus warranting further experimental examination.
In the normal fallopian tube, in the nonpregnant state, the immune landscape is dominated by T cells (especially CD8+ T cells), 36,55 whereas in EP, macrophages are the most abundant leukocyte population within the human fallopian tube, 32