Preeclampsia-specific immune cell network in placenta revealed by Cytometry by time of flight and single-cell RNA-seq

Reviewer #1 (Public review):

Summary:

In this study, the authors utilized human placental samples together with multiple mouse models to explore the mechanisms whereby inflammatory macrophages and T cells are linked to preeclampsia (PE). The authors first undertook CyTOF of placental samples from women with normal pregnancies, PE, gestational diabetes mellitus (GDM), and GDM with superimposed PE (GDM+PE). The authors report an increase of memory-like Th17 cells, memory-like CD8+ T cells, and pro-inflammatory macrophages in PE cases, but not GDM or GDM+PE, together with diminished γδT cells, anti-inflammatory macrophages, and granulocyte myeloid-derived suppressor cells (gMDSC). The authors then undertook several experiments using scRNA-seq, bulk RNA-seq, and flow cytometry in a RUPP model to first show that the transfer of pro-inflammatory macrophages from RUPP mice into normal pregnant mice with depleted macrophages resulted in increased embryo resorption and diminished fetal weight and size. Moreover, pro-inflammatory macrophages induced memory-like Th17 cells in mice. Similarly, injection of T-cells from RUPP mice resulted in increased embryo resorption and diminished fetal weight and size. Such mice that received RUPP-derived T cells displayed similarly worsened outcomes in their second pregnancy in the absence of any additional T cell transfer. The authors identified the IGF1-IGF1R ligand-receptor pair as a factor involved in the macrophage-mediated induction of memory-like Th17 cells, as confirmed by experiments using an IGF1R inhibitor. Finally, the authors transferred IGF1R inhibitor-treated T cells to a pregnant mouse that was administered LPS and depleted of T cells and observed improved outcomes compared to mice that received non-treated T cells. The authors conclude that their study identifies a PE-specific immune cell network regulated by pro-inflammatory macrophages and T cells.

Strengths:

Utilization of both human placental samples and multiple mouse models to explore the mechanisms linking inflammatory macrophages and T cells to preeclampsia (PE).
Incorporation of advanced techniques such as CyTOF, scRNA-seq, bulk RNA-seq, and flow cytometry.

Identification of specific immune cell populations and their roles in PE, including the IGF1-IGF1R ligand-receptor pair in macrophage-mediated Th17 cell differentiation.
Demonstration of the adverse effects of pro-inflammatory macrophages and T cells on pregnancy outcomes through transfer experiments.

Weaknesses:

Inconsistent use of uterine and placental cells, which are distinct tissues with different macrophage populations, potentially confounding results.

Missing observational data for the initial experiment transferring RUPP-derived macrophages to normal pregnant mice.

Unclear mechanisms of anti-macrophage compounds and their effects on placental/fetal macrophages.

Difficulty in distinguishing donor cells from recipient cells in murine single-cell data complicates interpretation.

Limitation of using the LPS model in the final experiments, as it more closely resembles systemic inflammation seen in endotoxemia rather than the specific pathology of PE.

Reviewer #2 (Public review):

Summary:

Fei, Lu, Shi, et al. present a thorough evaluation of the immune cell landscape in pre-eclamptic human placentas by single-cell multi-omics methodologies compared to normal control placentas. Based on their findings of elevated frequencies of inflammatory macrophages and memory-like Th17 cells, they employ adoptive cell transfer mouse models to interrogate the coordination and function of these cell types in pre-eclampsia immunopathology. They demonstrate the putative role of the IGF1-IGF1R axis as the key pathway by which inflammatory macrophages in the placenta skew CD4+ T cells towards an inflammatory IL-17A-secreting phenotype that may drive tissue damage, vascular dysfunction, and elevated blood pressure in pre-eclampsia, leaving researchers with potential translational opportunities to pursue this pathway in this indication.

They present a major advance to the field in their profiling of human placental immune cells from pre-eclampsia patients where most extant single-cell atlases focus on term versus preterm placenta, or largely examine trophoblast biology with a much rarer subset of immune cells. While the authors present vast amounts of data at both the protein and RNA transcript level, we, the reviewers, feel this manuscript is still in need of much more clarity in its main messaging, and more discretion in including only key data that supports this main message most effectively.

Strengths:

(1) This study combines human and mouse analyses and allows for some amount of mechanistic insight into the role of pro-inflammatory and anti-inflammatory macrophages in the pathogenesis of pre-eclampsia (PE), and their interaction with Th17 cells.

(2) Importantly, they do this using matched cohorts across normal pregnancy and common PE comorbidities like gestation diabetes (GDM).

(3) The authors have developed clear translational opportunities from these "big data" studies by moving to pursue potential IGF1-based interventions.

Weaknesses:

(1) Clearly the authors generated vast amounts of multi-omic data using CyTOF and single-cell RNA-seq (scRNA-seq), but their central message becomes muddled very quickly. The reader has to do a lot of work to follow the authors' multiple lines of inquiry rather than smoothly following along with their unified rationale. The title description tells fairly little about the substance of the study. The manuscript is very challenging to follow. The paper would benefit from substantial reorganizations and editing for grammatical and spelling errors. For example, RUPP is introduced in Figure 4 but in the text not defined or even talked about what it is until Figure 6. (The figure comparing pro- and anti-inflammatory macrophages does not add much to the manuscript as this is an expected finding).

(2) The methods lack critical detail about how human placenta samples were processed. The maternal-fetal interface is a highly heterogeneous tissue environment and care must be taken to ensure proper focus on maternal or fetal cells of origin. Lacking this detail in the present manuscript, there are many unanswered questions about the nature of the immune cells analyzed. It is impossible to figure out which part of the placental unit is analyzed for the human or mouse data. Is this the decidua, the placental villi, or the fetal membranes? This is of key importance to the central findings of the manuscript as the immune makeup of these compartments is very different. Or is this analyzed as the entirety of the placenta, which would be a mix of these compartments and significantly less exciting?

(3) Similarly, methods lack any detail about the analysis of the CyTOF and scRNAseq data, much more detail needs to be added here. How were these clustered, what was the QC for scRNAseq data, etc? The two small paragraphs lack any detail.

(4) There is also insufficient detail presented about the quantities or proportions of various cell populations. For example, gdT cells represent very small proportions of the CyTOF plots shown in Figures 1B, 1C, & 1E, yet in Figures 2I, 2K, & 2K there are many gdT cells shown in subcluster analysis without a description of how many cells are actually represented, and where they came from. How were biological replicates normalized for fair statistical comparison between groups?

(5) The figures themselves are very tricky to follow. The clusters are numbered rather than identified by what the authors think they are, the numbers are so small, that they are challenging to read. The paper would be significantly improved if the clusters were clearly labeled and identified. All the heatmaps and the abundance of clusters should be in separate supplementary figures.

(6) The authors should take additional care when constructing figures that their biological replicates (and all replicates) are accurately represented. Figure 2H-2K shows N=10 data points for the normal pregnant (NP) samples when clearly their Table 1 and test denote they only studied N=9 normal subjects.

(7) There is little to no evaluation of regulatory T cells (Tregs) which are well known to undergird maternal tolerance of the fetus, and which are well known to have overlapping developmental trajectory with RORgt+ Th17 cells. We recommend the authors evaluate whether the loss of Treg function, quantity, or quality leaves CD4+ effector T cells more unrestrained in their effect on PE phenotypes. References should include, accordingly: PMCID: PMC6448013 / DOI: 10.3389/fimmu.2019.00478; PMC4700932 / DOI: 10.1126/science.aaa9420.

(8) In discussing gMDSCs in Figure 3, the authors have missed key opportunities to evaluate bona fide Neutrophils. We recommend they conduct FACS or CyTOF staining including CD66b if they have additional tissues or cells available. Please refer to this helpful review article that highlights key points of distinguishing human MDSC from neutrophils: https://doi.org/10.1038/s41577-024-01062-0. This will both help the evaluation of potentially regulatory myeloid cells that may suppress effector T cells as well as aid in understanding at the end of the study if IL-17 produced by CD4+ Th17 cells might recruit neutrophils to the placenta and cause ROS immunopathology and fetal resorption.

(9) Depletion of macrophages using several different methodologies (PLX3397, or clodronate liposomes) should be accompanied by supplementary data showing the efficiency of depletion, especially within tissue compartments of interest (uterine horns, placenta). The clodronate piece is not at all discussed in the main text. Both should be addressed in much more detail.

(10) There are many heatmaps and tSNE / UMAP plots with unhelpful labels and no statistical tests applied. Many of these plots (e.g. Figure 7) could be moved to supplemental figures or pared down and combined with existing main figures to help the authors streamline and unify their message.

(11) There are claims that this study fills a gap that "only one report has provided an overall analysis of immune cells in the human placental villi in the presence and absence of spontaneous labor at term by scRNA-seq (Miller 2022)" (lines 362-364), yet this study itself does not exhaustively study all immune cell subsets...that's a monumental task, even with the two multi-omic methods used in this paper. There are several other datasets that have performed similar analyses and should be referenced.

(12) Inappropriate statistical tests are used in many of the analyses. Figures 1-2 use the Shapiro-Wilk test, which is a test of "goodness of fit", to compare unpaired groups. A Kruskal-Wallis or other nonparametric t-test is much more appropriate. In other instances, there is no mention of statistical tests (Figures 6-7) at all. Appropriate tests should be added throughout.

Author response:

Reviewer #1:

Strengths:

Utilization of both human placental samples and multiple mouse models to explore the mechanisms linking inflammatory macrophages and T cells to preeclampsia (PE).
Incorporation of advanced techniques such as CyTOF, scRNA-seq, bulk RNA-seq, and flow cytometry.

Identification of specific immune cell populations and their roles in PE, including the IGF1-IGF1R ligand-receptor pair in macrophage-mediated Th17 cell differentiation.
Demonstration of the adverse effects of pro-inflammatory macrophages and T cells on pregnancy outcomes through transfer experiments.

Weaknesses:

Comment 1. Inconsistent use of uterine and placental cells, which are distinct tissues with different macrophage populations, potentially confounding results.

Response1: We thank the reviewers' comments. We have done the green fluorescent protein (GFP) pregnant mice-related animal experiment, which was not shown in this manuscript. The wild-type (WT) female mice were mated with either transgenic male mice, genetically modified to express GFP, or with WT male mice, in order to generate either GFP-expressing pups (GFP-pups) or their genetically unmodified counterparts (WT-pups), respectively. Mice were euthanized on day 18.5 of gestation, and the uteri of the pregnant females and the placentas of the offspring were analyzed using flow cytometry. The majority of macrophages in the uterus and placenta are of maternal origin, which was defined by GFP negative. In contrast, fetal-derived macrophages, distinguished by their expression of GFP, represent a mere fraction of the total macrophage population, signifying their inconsequential or restricted presence amidst the broader cellular landscape. We will added the GPF pregnant mice-related data in Figure 4-figure supplement 1 to explain the different macrophage populations in the uterine and placental cells.

Comment 2. Missing observational data for the initial experiment transferring RUPP-derived macrophages to normal pregnant mice.

Response 2: We thank the reviewers' comments. In our experiments, PLX3397 or Clodronate Liposomes was used to deplete the macrophages of pregnant mice, and then we injected RUPP-derived pro-inflammatory macrophages and anti-inflammatory macrophages back into PLX3397 or Clodronate Liposomes-treated pregnant mice. And We found that RUPP-derived F480+CD206- pro-inflammatory macrophages induced immune imbalance at the maternal-fetal interface and PE-like symptoms (Figure 4E-4H and Figure 4-figure supplement 1 A-C).

Comment 3. Unclear mechanisms of anti-macrophage compounds and their effects on placental/fetal macrophages.

Response 3: We thank the reviewers' comments. PLX3397, the inhibitor of CSF1R, which is needed for macrophage development (Nature. 2023, PMID: 36890231; Cell Mol Immunol. 2022, PMID: 36220994), we have stated that on line 189-191. However, PLX3397 is a small molecule compound that possesses the potential to cross the placental barrier and affect fetal macrophages. We will discuss the impact of this factor on the experiment in the discussion section.

Comment 4. Difficulty in distinguishing donor cells from recipient cells in murine single-cell data complicates interpretation.

Response 4: We thank the reviewers' comments. Upon analysis, we observed a notable elevation in the frequency of total macrophages within the CD45+ cell population. Then we subsequently performed macrophage clustering and uncovered a marked increase in the frequency of Cluster 0, implying a potential correlation between Cluster 0 and donor-derived cells. RNA sequencing revealed that the F480+CD206- pro-inflammatory donor macrophages exhibited a Folr2+Ccl7+Ccl8+C1qa+C1qb+C1qc+ phenotype, which is consistent with the phenotype of cluster 0 in macrophages observed in single-cell RNA sequencing (Figure 4D and Figure 5E). Therefore, we believe that the donor cells is cluster 0 in macrophages.

Comment 5. Limitation of using the LPS model in the final experiments, as it more closely resembles systemic inflammation seen in endotoxemia rather than the specific pathology of PE.

Response 5: We thank the reviewers' comments. Firstly, our other animal experiments in this manuscript used the Reduction in Uterine Perfusion Pressure (RUPP) mouse model to simulate the pathology of PE. However, the RUPP model requires ligation of the uterine arteries in pregnant mice on day 12.5 of gestation, which hinders T cells returning from the tail vein from reaching the maternal-fetal interface. In addition, this experiment aims to prove that CD4+ T cells are differentiated into memory-like Th17 cells through IGF-1R receptor signalling to affect pregnancy by clearing CD4+ T cells in vivo with an anti-CD4 antibody followed by injecting IGF-1R inhibitor-treated CD4+ T cells. And we proved that injection of RUPP-derived memory-like CD4+ T cells into pregnant rats induces PE-like symptoms (Figure 6). In summary, the application of the LPS model in Figure 8 does not affect the conclusions.

Reviewer #2:

Strengths:

(1) This study combines human and mouse analyses and allows for some amount of mechanistic insight into the role of pro-inflammatory and anti-inflammatory macrophages in the pathogenesis of pre-eclampsia (PE), and their interaction with Th17 cells.

(2) Importantly, they do this using matched cohorts across normal pregnancy and common PE comorbidities like gestation diabetes (GDM).

(3) The authors have developed clear translational opportunities from these "big data" studies by moving to pursue potential IGF1-based interventions.

Weaknesses:

Comment 1. Clearly the authors generated vast amounts of multi-omic data using CyTOF and single-cell RNA-seq (scRNA-seq), but their central message becomes muddled very quickly. The reader has to do a lot of work to follow the authors' multiple lines of inquiry rather than smoothly following along with their unified rationale. The title description tells fairly little about the substance of the study. The manuscript is very challenging to follow. The paper would benefit from substantial reorganizations and editing for grammatical and spelling errors. For example, RUPP is introduced in Figure 4 but in the text not defined or even talked about what it is until Figure 6. (The figure comparing pro- and anti-inflammatory macrophages does not add much to the manuscript as this is an expected finding).

Response 1: We thank the reviewers' comments. According to the reviewer's suggestion, we will proceed with making the necessary revisions. Firstly, We will modify the title of the article to be more specific. Then, we will introduce the RUPP mouse model when interpreted Figure 4. Thirdly, we plan to simplify or consolidate the images from Figure5 to Figure7 to make them easier to follow. Finally, We will diligently correct the grammatical and spelling errors in the article. As for the figure comparing pro- and anti-inflammatory macrophages, The Editor requested a more comprehensive description of the macrophage phenotype during the initial submission. As a result, we conducted the transcriptomes of both uterine-derived pro-inflammatory and anti-inflammatory macrophages and conducted a detailed analysis of macrophages in single-cell data.

Comment 2. The methods lack critical detail about how human placenta samples were processed. The maternal-fetal interface is a highly heterogeneous tissue environment and care must be taken to ensure proper focus on maternal or fetal cells of origin. Lacking this detail in the present manuscript, there are many unanswered questions about the nature of the immune cells analyzed. It is impossible to figure out which part of the placental unit is analyzed for the human or mouse data. Is this the decidua, the placental villi, or the fetal membranes? This is of key importance to the central findings of the manuscript as the immune makeup of these compartments is very different. Or is this analyzed as the entirety of the placenta, which would be a mix of these compartments and significantly less exciting?

Response 2: We thank the reviewers' comments. Placental villi rather than fetal membranes and decidua were used for CyToF in this study. This detail about how human placenta samples were processed will be added to the Materials and Methods section.

Comment 3. Similarly, methods lack any detail about the analysis of the CyTOF and scRNAseq data, much more detail needs to be added here. How were these clustered, what was the QC for scRNAseq data, etc? The two small paragraphs lack any detail.

Response 3: We thank the reviewers' comments. The detail about the analysis of the CyTOF and scRNAseq data will be added in the Materials and Methods section.

Comment 4. There is also insufficient detail presented about the quantities or proportions of various cell populations. For example, gdT cells represent very small proportions of the CyTOF plots shown in Figures 1B, 1C, & 1E, yet in Figures 2I, 2K, & 2K there are many gdT cells shown in subcluster analysis without a description of how many cells are actually represented, and where they came from. How were biological replicates normalized for fair statistical comparison between groups?

Response 4: We thank the reviewers' comments. In Figure 1, CD45+ immune cells were clustered into 10 subpopulations, which included gdT cells. While Figure 2 displays the further clustering analysis of CD4+T, CD8+T, and gdT cells, with gdT cells being further subdivided into 22 clusters (Figure 2-figure supplement 1C). The number of biological replicates (samples) is consistent with Figure 1.

Comment 5. The figures themselves are very tricky to follow. The clusters are numbered rather than identified by what the authors think they are, the numbers are so small, that they are challenging to read. The paper would be significantly improved if the clusters were clearly labeled and identified. All the heatmaps and the abundance of clusters should be in separate supplementary figures.

Response 5: We thank the reviewers' comments. The t-SNE distributions of the 15 clusters of CD4+ T cells, 18 clusters of CD8+ T cells, and 22 clusters of gdT cells are shown separately in Figure 2A, F, and I. The heatmaps displaying the expression levels of markers in these clusters of CD4+ T cells, CD8+ T cells, and gdT cells are presented in Figure 2-figure supplement 1A, B, and C, respectively. The t-SNE distributions of the 29 clusters of CD11b+ cells are shown in Figure 3A, and the heatmap displaying the expression levels of markers in these clusters is presented in Figure 3B. As for sc-RNA sequencing, the heatmap and UMAP distributions of the 15 clusters of macrophages are shown separately in Figure 5C and 5D. The UMAP distributions and heatmap of the 12 clusters of T/NK cells are shown in Figure 6A and 6B. The UMAP distributions and heatmap of the 9 clusters of T/NK cells are shown in Figure 7A and 7B.

Comment 6. The authors should take additional care when constructing figures that their biological replicates (and all replicates) are accurately represented. Figure 2H-2K shows N=10 data points for the normal pregnant (NP) samples when clearly their Table 1 and test denote they only studied N=9 normal subjects.

Response 6: We thank the reviewers' careful checking. During our verification, we found that one sample in the NP group had pregnancy complications other than PE and GMD. The data in Figure 2H-2K was not updated in a timely manner. We will promptly update this data and reanalyze it.

Comment 7. There is little to no evaluation of regulatory T cells (Tregs) which are well known to undergird maternal tolerance of the fetus, and which are well known to have overlapping developmental trajectory with RORgt+ Th17 cells. We recommend the authors evaluate whether the loss of Treg function, quantity, or quality leaves CD4+ effector T cells more unrestrained in their effect on PE phenotypes. References should include, accordingly: PMCID: PMC6448013 / DOI: 10.3389/fimmu.2019.00478; PMC4700932 / DOI: 10.1126/science.aaa9420.

Response 7: We thank the reviewers' comments. We have done the Treg-related animal experiment, which was not shown in this manuscript. We will add the Treg-related data in Figure 6. The injection of CD4+ T cells derived from RUPP mouse, characterized by a reduced frequency of Tregs, could induce PE-like symptoms in pregnant mice. Additionally, we will add a necessary discussion about Tregs.

Comment 8. In discussing gMDSCs in Figure 3, the authors have missed key opportunities to evaluate bona fide Neutrophils. We recommend they conduct FACS or CyTOF staining including CD66b if they have additional tissues or cells available. Please refer to this helpful review article that highlights key points of distinguishing human MDSC from neutrophils: https://doi.org/10.1038/s41577-024-01062-0. This will both help the evaluation of potentially regulatory myeloid cells that may suppress effector T cells as well as aid in understanding at the end of the study if IL-17 produced by CD4+ Th17 cells might recruit neutrophils to the placenta and cause ROS immunopathology and fetal resorption.

Response 8: We thank the reviewers' comments. Although we do not have additional tissues or cells available to conduct FACS or CyTOF staining, including for CD66b, we plan to utilize CD15 and CD66b antibodies for immunofluorescence staining of placental tissue. Suppressing effector T cells is a signature feature of MDSCs, and T cells may also influence the functions of MDSCs, we will refer to this review and discuss it in the Discussion section of the article.

Comment 9. Depletion of macrophages using several different methodologies (PLX3397, or clodronate liposomes) should be accompanied by supplementary data showing the efficiency of depletion, especially within tissue compartments of interest (uterine horns, placenta). The clodronate piece is not at all discussed in the main text. Both should be addressed in much more detail.

Response 9: We thank the reviewers' comments. We already have the additional data on the efficiency ofmacrophage depletion involving PLX3397 and clodronate liposomes, which were not present in this manuscript, and we'll add it to the manuscript. The clodronate piece is mentioned in the main text (Line 197-201), but only briefly described, because the results using clodronate we obtained were similar to those using PLX3397.

Comment 10. There are many heatmaps and tSNE / UMAP plots with unhelpful labels and no statistical tests applied. Many of these plots (e.g. Figure 7) could be moved to supplemental figures or pared down and combined with existing main figures to help the authors streamline and unify their message.

Response 10: We thank the reviewers' comments. We plan to simplify or consolidate the images from Figure5 to Figure7 to make them easier to follow.

Comment 11. There are claims that this study fills a gap that "only one report has provided an overall analysis of immune cells in the human placental villi in the presence and absence of spontaneous labor at term by scRNA-seq (Miller 2022)" (lines 362-364), yet this study itself does not exhaustively study all immune cell subsets...that's a monumental task, even with the two multi-omic methods used in this paper. There are several other datasets that have performed similar analyses and should be referenced.

Response 11: We thank the reviewers' comments. We will search for more literature and reference additional studies that have conducted similar analyses.

Comment 12. Inappropriate statistical tests are used in many of the analyses. Figures 1-2 use the Shapiro-Wilk test, which is a test of "goodness of fit", to compare unpaired groups. A Kruskal-Wallis or other nonparametric t-test is much more appropriate. In other instances, there is no mention of statistical tests (Figures 6-7) at all. Appropriate tests should be added throughout.

We thank the reviewers' comments. As stated in the Statistical Analysis section (lines 601-604), the Kruskal-Wallis test was used to compare the results of experiments with multiple groups. Comparisons between the two groups in Figures 6-7 were conducted using Student's t-test. The aforementioned statistical methods will be included in the figure legends.