Human placental biology at single-cell resolution: a contemporaneous review

Abstract

Single-cell technologies capture cellular heterogeneity to focus on previously poorly described subpopulations of cells. While work by our lab and many others have metagenomically characterized a low biomass intrauterine microbial community, alongside microbial transcripts, antigens, and metabolites, the functional importance of low biomass microbial communities in placental immuno-microenvironments are still being elucidated. Given their hypothesized role in modulating inflammation and immune ontogeny to enable tolerance to beneficial microbes while warding off pathogens, there is a need for single-cell resolution. Herein, we summarize the potential for mechanistic understandings of these and other key fundamental early developmental processes by applying single-cell approaches.

Introduction: why science resolved at a single-cell level matters.

Mammalian tissues are heterogenous and complex. Being composed of numerous distinct and variant cell types with diverse functions, microarchitecture, and differentiation states, biochemical and molecular assays performed on homogenates or tissue en masse are limited to describing the population average of many cells. As such, they fail to provide the level of discriminate details and may miss rare but biologically important signals. However, recent advances in special and single cell sequencing approaches have launched a high-resolution era of tissue assessment that is now upon us. Single-cell DNA and RNA sequencing was named Nature Method of the Year in 2013, multi-modal single-cell analyses in 2019, and spatial transcriptomics in 2020.(1–3) Single-cell technologies have great merit in the field of maternal-fetal medicine with the potential for transformative discoveries, personalized therapies, and ultimately novel therapeutics.

In reproductive biology and medicine, observations derived from heterogeneous cell populations have resulted in residual unanswered questions which necessitate single-cell resolution to answer (Fig. 1). For example, standard clinical approaches for in vitro fertilization include pre-implantation single cell genomics. In parallel, developmental biologists have resolved whole-transcriptome profiles on preimplantation human embryos from the zygote, to the 2-cell stage, 4-cell, 8-cell, and further stages with single-cell technologies.(4–8) Additionally, it has been shown that it is now possible to integrate single cell technologies into a multi-omics approach of single-cell RNA-seq (scRNA-seq), combining single-cell whole genome sequencing with single-cell bisulfite sequencing to define multi-layered differentiation pathways.(9) Copy number variants (CNV) and single nucleotide polymorphisms (SNP) analyses captured both euploid and a rare subpopulation of aneuploid cells with unsynchronized X chromosome silencing. Single-cell DNA methylation revealed dynamic re-methylation patterns partitioning distinct developmental stages. Several other groups have developed human trophoblast stem cell systems as an alternative to utilizing human embryo blastocysts. These are engineered by (i) differentiating naïve pluripotent stem cells or (ii) by reprogramming human fibroblasts into induced-pluripotent stem cells.(10, 11) Additionally, 3D-blastocyst organoid systems have recapitulated cellular heterogeneity and complex cellular architectures found in vivo.(12–14) Collectively, these studies have demonstrated the utility of scRNA-seq and multi-omics approaches in rendering molecular snapshots of crucial intervals of embryonic development.

Figure 1.

Increasing resolution of single-cell technologies. A) Classical microscopy techniques: DNA and RNA in situ hybridization, immunofluorescence. B) Bulk assays: NGS DNA-seq, RNA-seq, Western blots. C) Single-cell: flow cytometry, single-cell assay for transposase-accessible chromatin (scATAQ-seq), single-cell RNA-seq (scRNA-seq), spatial transcriptomics. D) Multi-omics: scRNA-seq + scATAQ-seq, proteogenomics, Visium. Created with BioRender.com.

Only recently have single-cell technologies sought to similarly resolve the far more complex and heterogenous fetal and trophoblast-derived tissues characteristic of the fuller spectrum of eutherian mammalian development. For example, we and others have used cultivation independent approaches to provide evidence for a low biomass intrauterine and fetal microbial community.(15,16,62–64) More recently, several groups have tested functional hypotheses related to the role of these low biomass communities of microbiota in fetal immune development and tolerance. (20–22) Single-cell approaches will be key to elucidating host-microbiome interactions in these communities, and resolving these functional hypothesis. From human fetal intestine aged 14–22 weeks, subpopulations of functional memory CD4⁺ T cells were found and characterized by scRNA-seq, CyTOF, and TCR linage traced.(20) A recent study rigorously characterized bacterial species in human fetal organs during the 2nd-trimester by metagenomics, cultivated the live bacteria, and visualized bacteria in the gut by electron microscopy.(21) Further, these authors demonstrated activation of memory T cells from the fetal lymph node by specific bacteria, providing further evidence that transient or persistent bacteria have a functional role in fetal immune development. Additionally, the existence of germ-free animal models is often used as an argument against the presence of a functional low biomass intrauterine or fetal microbial community.(15, 16) In fact, gnotobiotic animals are decidedly abnormal, and are riddled with immune, metabolic, and behavioral deficiencies which cannot be corrected with postnatal colonization but can be partially restored with in utero, early gestational colonization.(17, 18) Similarly, cross-species microbiota transplants have demonstrated that both the composition of the transplanted community of microbes and immune pathways were consistently shaped by the donor, not the recipient.(19) In addition, multi-omic analyses of fetal tissues have supported consistent or transient exposure to microbes. In the near future, single-cell resolution will be important in identifying cell-to-cell interactions and potential bystander activation (or tolerance) in fetal and intrauterine immune microenvironments.(22)

Herein, we provide an overview of these technologies and summarize the findings of single-cell studies thus far analyzing over 414,237 cells from human placental tissues and identify areas of unmet need. As one key example, we will highlight how single-cell science can illuminate the “when, where, and how” pertaining to the development of the fetal repertoire of hematopoietic lineages and immune cells. This rich and complex environment, which includes mesenchymal stem cells, decidual fibroblasts, maternal and fetal red blood cells (RBC), Natural Killer (NK) cells, T cells, and placental-specific Hofbauer macrophages, remains poorly understood with respect to both the constituents and their functions.(23) With a lens on how single-cell technologies may be used to elucidate any functional importance of low biomass microbial communities or microbial antigen exposure in the intrauterine environment, we will provide a framework for understanding how the potential for mechanistic understandings of these and other key fundamental early developmental processes can be enabled through the application of 21^st century single-cell approaches.

Single-cell sequencing: the science has arrived.

To date, there are at least 46 developing single-cell platforms and nearly 200 specific applications.(24) There are over 940 analysis packages in over 30 categories currently being tracked on scRNA-tools.org.(25) Current single-cell ‘omics approaches include targeting DNA, RNA, protein, and epigenetic context. In single-cell DNA sequencing (scDNA-seq), applications include detecting copy number variants (CNV), single-nucleotide polymorphisms (SNPs), inserts and deletions, as well as aneuploidies (e.g. Fluidigm C1, Takara Bio PicoPLEX Gold).(26–28) For epigenetics, Assay for Transposase-Accessible Chromatin using sequencing (ATAQ-seq) identifies sequences accessible to a transposase that are likely being transcribed. Conversely, this can infer epigenetically repressed regions. Isolation of nuclei is necessary for scATAQ-seq, and nuclei are incubated in a transposase mixture where adapter sequences are added to DNA fragments as they are transposed. The combination of scATAQ-seq with scRNA-seq in a multi-omic approach can directly correlate changes in chromatin accessibility with mRNA levels.(29) At the protein level, strides have been made and ideal single-cell proteomics systems are still in development.(30) Leading single-cell proteomics approaches combine flow cytometry with liquid chromatography and tandem mass spectrometry (LC-MS/MS) (e.g. Fluidigm CyTOF Helios). Thus far, this technology has been optimized to capture <50 proteins in single cells. The need for labeling targets with heavy metal isotopes is a current limiting factor.

Single-cell transcriptomics.

scRNA-seq is often used and readily available on several platforms (e.g. Fluidigm C1, Drop-seq, 10x Genomics Chromium Next GEM Single Cell 5’/3′, Takara Bio SMART-seq)(31). Cells in culture or from tissue are enzymatically or mechanically disrupted and put through cell strainers that catch clumps of cells and only allow single cells to pass. These cells are then pelleted and washed to remove ambient RNA from lysed cells, with an optional secondary enrichment by gradient centrifugation or flow cytometry into single wells. Alternatively, single-cell suspensions are loaded into microfluidics systems that only allow a single cell to pass and each cell is then captured in a gel emulsion matrix (GEM). Each GEM (or well) has a unique cell barcode and all of the reagents necessary for reverse transcription and NGS library preparation.(32, 33) Sequencing yields millions of reads, each with a unique cell barcode, cDNA insert, and unique molecule identifiers (UMIs) that help determine mRNA copy numbers.

scRNA-seq analysis workflow.

Bioinformatic processing and analysis of these scRNA-seq reads includes pre-processing of the adapters, quality filtering, alignment to a reference transcriptome, followed by binning and counting of cell transcripts according to each barcode(34). This results in a feature:barcode counts matrix, composed of single-cell transcriptomes for thousands to millions of cells (e.g., x=10,000 cell-barcodes and y=21,000 transcripts). Prior to clustering, each count matrix undergoes secondary quality control steps to remove low-quality cells (e.g. apoptotic, mitochondria, and doublets), where GEMs with low transcript counts and transcript complexity are removed. More stringent filtering excludes apoptotic cells, where a GEM captures mitochondria with intact membranes from lysed cells bioinformatically identifiable by a high proportion of mitochondrial transcripts (>5–10%). Next, the data are log-transformed, scaled, and dimensionally reduced via principal component analysis (PCA). Differential expression analysis identifies transcripts that are most responsible for variance and cellular stratification, which is then used for clustering into sub-populations defined by unique and shared transcriptomic features. Cell clusters are visualized in 2-dimensional space via Unique Manifold Approximation and Projection (UMAP)(35) and annotated either manually by known marker transcripts or automatically based on comparisons to reference datasets. Beyond differential expression analysis, numerous downstream analyses further define rare subtypes of cells, predict fundamental cell-cell interactions, and identify perturbed molecular pathways. For instance, pseudotime trajectory analysis has been shown to be able to infer longitudinal relationships between cell states of differentiation based on a defined starting point.(36, 37) Additionally, pseudotime trajectories can identify transition states and characterize unique cell fates. Pearson’s correlation coefficient analyses can identify novel transcription programs that define particular cells.(38) RNA velocity can be calculated based on spliced or unspliced transcripts.(13) Ligand-receptor interactions can be inferred to describe potential cell-cell interactions.(39, 40) To summarize, scRNA-seq is a discovery tool that describes heterogeneity and can identify unique cell types with canonical or novel biomarkers, which must be validated with complementary approaches such as immunofluorescence, flow cytometry, or RNAscope. The biological significance of these distinct sub-populations should be examined by genetic perturbations including knocking out, knocking down, or overexpressing marker transcripts of interest.

Spatial transcriptomics.

The field of spatial transcriptomics pairs scRNA-seq with histology, making it possible to overlay near single-cell transcriptome data with an H&E stain or immunofluorescence images that probe for up to 6 antigens (e.g. 10x Genomics Visium or Seq-Scope).(41, 42) Each spot on a capture area of a Visium slide contains approximately 5,000 spots that each have coordinates and a DNA barcode. Recently, a version of Visium has been optimized for formalin-fixed paraffin-embedded tissue. This approach utilizes pools of RNA-hybrid pools for 20,000–30,000 transcripts in human and mouse species. The current resolution of 10x Genomics Visium (v2) is 5–10 cells per spot, an improvement from 100 cells/spot (v1), and ultra-resolution is being developed to have single-cell or sub-cellular resolution.

Multi-omic immune profiling.

Classically, immunophenotyping quantifies the abundance of cell surface proteins on individual cells and can be utilized to infer function or associate functional ratios with disease. Examples include quantifying cell surface proteins to characterize M1/M2 macrophage polarization or T_H1/T_H2 ratios of T cells. Multi-omics immune profiling by single cell technology provides greater resolution of the transcriptome, and can simultaneously quantitate transcript expression of cell-surface proteins, VDJ recombination lineage(33), and antigen-specific epitope information from a single cell (utilizing technologies and platforms such as dCODE Dextramer, 10x Genomics Chromium Next GEM Single Cell 5’, and BioLegend TotalSeq). Since these multi-omic approaches provide clarity to the dynamic and multi-dimensional nature of complex human pathologies, they are ripe for application aimed at unveiling the intricacies of the maternal-fetal interface across gestation.

Placental biology at single-cell resolution: opportunities abound.

Several studies have examined human placental cells by scRNA-seq at progressive stages of gestation (recently reviewed in (43–45), see Table 1). However, unlike most tissue types, scRNA-seq of the placenta has unique technical challenges, including delineation of anatomical site being sampled (e.g. decidual basalis or spongiosa), varying methodologies of generating single-cell suspensions, and the inherent dynamic nature of the transcriptomic profile throughout development, from implantation to parturition (Fig. 2). In addition, capturing and analyzing large and multinucleated syncytiotrophoblasts with current platforms is inherently challenging as unbiased analyses would require nucleus-isolation or spatial single-cell analytics. Currently, syncytiotrophoblasts may be captured and successively purged at quality control steps where doublets, outliers with more than twice the RNA content of all other cells, are filtered out. An intriguing feature of placental sampling is the ability to differentiate maternal and fetal cells by comparing SNPs, or when analyzing male-derived fetuses by expression of transcripts from the X or Y chromosomes, therein aiding in the annotations of maternal and fetal cell clusters downstream.

Table 1. Known human placental scRNA-seq datasets.

Summary of 424,238 scRNA-seq transcriptomes related to placental biology.

First Author	Year	Samples	Cells	Data Repository	Platform
Pavličev	2017	term, cesarean-section placentas from two patients	87	GEO: GSE87726	Fluidigm C1 and laser microdissection
Tsang	2017	term placentas (two female and two male babies)	20,518	EBI: EGAS00001002449	10X Genomics 3' (v1)
Suryawanshi	2018	First trimester (6–11 weeks) >14,300 placental villi from eight patients and 6.7k decidual cells from five patients	14,341 (placental villi), 6,754 (decidua)	BioProject ID: PRJNA492324	10X Genomics and Drop-seq
Liu	2018	first- and second-trimester (8- and 24-week) placental samples	1,471	GEO: GSE89497	SMART-seq2
Vento-Tormo	2018	first-trimester (6–11 weeks) placenta (5 patients, 18,547 cells), decidua (11 patients, 40,512), and matched-PBMCs (6 patients, 10,001 cells)	70,325	ArrayExpress: E-MTAB-6701 (10x Genomics) and E-MTAB-6678 (SMART-seq2)	SMART-seq2 or 10X Genomics 3' (v2)
Pique-Regi	2019	term chorioamniotic membranes, placental villi, and basal plate samples (with or without labor at 38–40 weeks) were compared with samples from pre-term labor (33–34 weeks) patients.	79,906	NIH dbGAP: phs001886.v1.p1	10X Genomics 3' (v2)
Pique-Regi	2020	snRNA-seq from 32 placental villi and decidual samples collected from in the second or third-trimester	26,501	NIH dbGAP: phs001886.v2.p1	10X Genomics 3' (v2 3rd trim. or v3 2nd trim.)
Sun	2020	first-trimester (6–11 weeks) placental villi were analyzed from 6 patients (3 male and 3 female offspring)	7,245	GEO: GSE131696	10X Genomics 3' (v2)
Han	2020	chorionic villus cells from a 13-week male and additional placenta cells from a 10 week female	19,493	GEO: GSE134355 **700k cells- select (a) placenta and (b) chorionic villus samples.	Microwell-seq
Cao	2020	5 males and 6 females estimated at 12–17-weeks	29,876	GEO: GSE156793 **4.9M cells- select 11 placenta samples	Sci-RNA-seq3
Saha	2020	first-trimester placentas (6 and 7.6 weeks)	14,603 (independent analysis)	GEO: GSE145036	10X Genomics 3' (v, unk.)
Guo	2021	24 human 1st- trimester decidual samples (9 RPL and 15 healthy) that were presorted for CD45+ cells	10,142 (controls) and 8,504 (RPL)	GSA: CRA002181	10X Genomics 3' (v2)
Yang	2021	Full-term placentas collected during caesarean section (2 healthy and 2 GDM) were subject to scRNA-seq	14,591 (GDM) and 12,629 (control)	GEO: GSE173193	10X Genomics 3' (v3)
Shannon	2021	healthy placental villi from 7 patients during the first-trimester and additional trophoblast organoid cultures	12,794 (placenta) and 37,996 (organoid)	ArrayExpress: E-MTAB-6701	10X Genomics (v, unk)

Figure 2.

Multi-modal single-cell approaches. A) Example samples may include placental or decidual biopsies, preimplantation embryos from culture, or PBMCs. B) Cells are enzymatically digested for single-cell suspensions or slides are made for spatial transcriptomics. C) H&E or immunofluorescence occurs prior to cDNA synthesis (5k spots/capture) D) scRNA-seq transcriptomes are analyzed to identify known and differential expression identifies novel cell types. Markers of novel cell types are confirmed with complementary approaches. Created with BioRender.com.

To this end, annotation of cell clusters relies on canonical biomarkers that have been used for decades to differentiate cell types in microscopy. For trophoblasts, broad markers include HLA-G, KRTY, and hCG paired with the lack HLA-A, HLA-B, or MHC-II. Trophoblasts can be further delineated as proliferative villous cytotrophoblasts (VCTs) by PARP1, or differentiated extravillous trophoblasts (EVTs) by HLA-G, and differentiated syncytiotrophoblasts by ERVFRD-1.(reviewed in (46)) Characteristics indicative of an ideal cellular marker transcript or protein include a high degree of sensitivity and accuracy. Additionally, the perfect marker would be specific and not observed in other cell types. Manual annotations of canonical marker transcripts based on extensive literature searches can identify broad cell types. Additionally, top differentially expressed transcripts among clusters may lead to the discovery of novel markers, pending validation. As an alternative, clusters can be annotated automatically using bulk RNA-seq from known cell populations as reference datasets (e.g. SingleR(47)). To aid in manual annotation of placental scRNA-seq clusters, in this review outline expected placental cell types (Fig. 3) and we have curated a list of single-cell marker transcripts for >90 placental cell types and subtypes (Table S1).

Figure 3.

Cells annotated in placental scRNA-seq studies. Known and recently described cell-types broadly identified either in placental villi, decidual, or amniotic membranes. Created with BioRender.com.

Human placental scRNA-seq: key initial studies of healthy term placentas.

The first scRNA-seq from human placental samples was published in 2017.(48) From term, cesarean-delivered placentas collected from 2 subjects, villous tissue was dissected and minced before enzymatic disruption for single-cell isolation followed by gradient-centrifugation. scRNA-seq utilizing the Fluidigm C1 system resulted in 87 single-cell transcriptomes. In addition, since syncytiotrophoblasts are multinucleated, these authors prudently isolated two single-cell transcriptomes by laser microdissection. Clustering resulted in 5 clusters annotated as 3 distinct VCTs clusters, maternal dendritic cells (DCs), and EVTs (see Table 1). Among top differentially expressed transcripts were ribosomal RNAs, which are often regressed out of the analysis in more recent scRNA-seq analysis workflows.(49–51) Receptor-ligand analyses inferred cell-communication networks and suggested changes resulting from uterine decidualization caused by the WNT signaling pathway. Subsequently, the importance of WNT signaling was verified by immunofluorescence experiments showing decidua-condition media increased β-catenin nuclearization.

Another scRNA-seq study of term placenta (2 female and 2 male offspring) yielded 20,518 transcriptomes from the 10x Genomics platform.(50) Twelve clusters were identified, and fetal or maternal origin was determined based on SNP-analysis or Y chromosome transcription. Trophoblasts are known to lack HLA-A expression and subtypes including EVTs were defined by HLA-G and CYP19A1. In contrast, VCTs were identified by PARP1, and syncytiotrophoblasts were marked by GH2 and CGA. Hofbauer cells were found by CD209 and CD163 markers. Maternal macrophages were marked by AIF1, CD53, CSF1R, and CD14. T cells were identified by GZMA and CD3G. Maternal DCs were marked by HLA-DR, HLA-DQ, and CD83. Vascular endothelial cells were identified by CDH5, CD34, and ICAM1 while vascular smooth muscle cells had expression of CNN1 and MYH11. Villous stromal cells expressed COL3A1, COL1A1, and VIM. Contaminant RBCs were identified by HBB, HBA1, and HBG1. Pseudotime analyses predicted differentiation paths starting at VCTs branching towards either EVTs or syncytiotrophoblasts driven by SLC1A2, ADHFE1, and DEPDC1B transcription. Cell-type-specific transcription programs were used to identify differences in pathology from plasma from preeclamptic gravidae and healthy controls. Complementary bulk RNA-seq data from cell-free RNA from these gravidae showed more EVTs transcripts in the plasma, suggesting these cells are disrupted. Pathway analyses revealed preeclampsia upregulated transcripts in cell proliferation, cell death, and apoptosis, while DNA damage repair and antigen presentation were downregulated compared to plasma from term placentas.

Placenta scRNA-seq from the 1^st- and 2^nd-trimesters.

Suryawanshi et al. (2018), analyzed 1^st-trimester (6–11 weeks gestation) placental villi from eight gravidae yielded 14,341 cells and 6,754 decidual cells from five subjects by both 10x Genomics and Drop-seq scRNA-seq platforms.(39) Data analysis was done using Seurat(49) and included quality-control filtering of highly expressed ribosomal and nuclear noncoding-RNA MALAT1 transcripts. Clustering resulted in 9 villous and 11 decidual clusters. Specifically, canonical and potentially novel markers delineated Hofbauer cells, multiple immune lineages, EVTs, syncytiotrophoblasts, and VCTs (see Table 1 for a more complete delineation). Distinctions in transcription and interactions between these villi and decidual cellular subtypes were further characterized by inferred ligand-receptor interactions. In a separate study, Liu et al. (2018) assessed 1^st- and 2^nd-trimester (8- and 24-weeks) placental samples yielding 1,471 cells with the SMART-seq2 platform.(52) Utilizing a magnetic activated cell sorting (MACS) workflow, EVTs (HLA-G), cytotrophoblasts (CDH1), syncytiotrophoblasts (mouth pipetted based on size), and villous stromal cells (HLA-G and CDH1⁻) were enriched. These 5 cell types were broadly annotated based on canonical markers and an additional 14 total subtypes were characterized by differential expression and functional pathways. Immunofluorescence and immunohistochemistry analyses confirmed the presence of proteins derived from top marker transcripts. There were two distinct Hofbauer macrophages subtypes distinguished by expression of CD68 and MRC1, representing active and resting subpopulations respectively. Pseudotime trajectories of trophoblasts showed in distinctions between the 1^st- and 2^nd-trimester EVTs, where LAIR2 and PAGE4 were higher in the 1^st-trimester compared to later upregulation of PRG2, TAC3, and SERPINE2.

To generate an encyclopedia of cells at the maternal-fetal interface, Vento-Tormo et al. (2018) performed scRNA-seq on 70,325 cells from 1^st-trimester (6–11 weeks) placenta (5 patients, 18,547 cells), decidua (11 patients, 40,512), and matched-PBMCs (6 patients, 11,266 cells).(40) Their approach utilized flow cytometry enrichment prior to both 10x Genomics and SMART-seq2 platforms. This resulted in 21 clusters of immune cells and 17 non-immune. In addition, these authors developed a comprehensive ligand-receptor database called CellPhoneDB, which predicted cell-cell interactions and defined distinct ligand-receptor pairs in trophoblast endpoints. Pseudotime analysis of trophoblast cells recapitulated differentiation paths of VCTs into EVTs or syncytiotrophoblasts(52, 53) and CellPhoneDB extended these insights by inferring syncytiotrophoblasts interactions with EGFR, NRP2, and MET receptors and EVTs interactions with CXCL6, TGFB1, and PAPPA. Furthermore, the compacta and spongiosa layers of the decidua were characterized using differential expression to find top markers, which were confirmed by immunohistochemistry of ACTA2 and RNAscope probing for PRL, ACTA2, and IGFBP1. Notably, three distinct decidual NK cell subtypes were discovered and characterized. Using genomic references, KIR-related genes were haplotyped and determined as HLA-C in dNK1, HLA-E in dNK2, and a distinct lack of HLA-E in dNK3. The dNK1 subtype was defined by CYP26A1, ID3, and KIR2DS1 transcripts, increased expression of glycolysis enzymes, and confirmed by FACS sorting of additional samples gating for CD39, CYP26A1, and B4GALNT1. dNK1 cells were functionally distinguished by Giemsa staining, which showed increased granule components, a function primed after the first pregnancy. The dNK2 subtype was most abundant and differentiated by CD27, GZMH, and NKG2E transcripts and confirmed by FACS sorting for ANXA1. The dNK3 subtype was defined by CD160, CXCR4, and RGS2 and could be enriched by gating for CD160, KLRB1, and CD103. Overall, Vento-Tormo and colleagues (2018) thoroughly exemplified rigorous characterization of previously unreported cell types, developed a widely used ligand-receptor database, and brought attention to decidual microenvironments.

Recently, Shannon et al. (2021) analyzed 7,798 cells from healthy placental villi from 7 gravidae during the 1^st-trimester and an additional 6,309 cells from trophoblast organoid cultures by scRNA-seq.(54) By comparing results with the published Vento-Tormo et al. (2018) dataset, they found BCAM was a marker of progenitor cytotrophoblasts. Subset analyses of 7,798 trophoblasts identified “column trophoblasts”, defined by expression of NOTCH1, ITGA2, and SOX9 or low expression of EGFR, TFAPC, and YAP1. They also analyzed three different trophoblast organoid lines, each treated with WNT⁺ or WNT⁻ media for 7 days to promote proliferation or extravillous trophoblast differentiation, respectively. The scRNA-seq analyses revealed the trophoblast organoids closely resembled those freshly isolated from placenta, providing a promising model for experimentation. Subsequent knockdown of BCAM by siRNAs in the organoid model resulted in impaired trophoblast organoid growth, confirming its importance in early trophoblast proliferation.

Placental sexual dimorphism.

Biological sex has many known influences on immune regulatory networks. Without sorting or lysing RBCs, 1^st-trimester (6–11 weeks) placental villi were analyzed from 6 gravidae (3 male and 3 female offspring) yielding 7,245 high-quality scRNA-seq transcriptomes.(55) Cluster analysis defined trophoblasts, stromal fibroblasts, Hofbauer cells, antigen-presenting cells, and endothelial cells by canonical markers. Cells then assigned male or female to determine cell-type-specific transcription changes based on sex. Across all cell types, significant differences were observed including increased levels of RPS4Y1, EIF1AY and DDX37 in males and increased MAGEA4, TMSB4X, XIST in females. Sub-setting trophoblast cells, identified 7 subtypes distinguished by pseudotime trajectories branching towards either HLA-G or ERVFRD-1 cell fates, reflecting extravillous or syncytiotrophoblasts, respectively. Immunohistochemistry localized MUC15 and NOTUM to the syncytiotrophoblasts and EVTs, respectively, which were found to be among the most sexually dimorphic transcripts and transcribed from autosomal chromosomes. Analyses to predict upstream regulators identified TGFB1 and estradiol to influence each of the cell types and were both found to be significantly upregulated in males. These results were supported by bulk RNA-seq data from matched decidua and placental villi samples, 22 females and 17 males, at 11–13 weeks, which were utilized to infer sex-based differences in receptor-ligand pairings.

Comprehensive single cell tissue atlases, including human placenta.

Valiant efforts have generated comprehensive single-cell atlases comparing transcriptomes and chromatin accessibility across human tissues including adult, placental, and other fetal tissues (56–58). Using microwell-seq, Han et al. (2020) profiled 702,968 cells including 9,898 chorionic villus cells from a 13-week male and 9,595 placenta cells from a 10-week female (57). These cells were included in their analysis of all human tissues compared to murine cells, but specific analyses focused on the placental cells were not included. Cao et al. (2020), analyzed 4,979,593 from 28 fetal samples from 10–18-weeks (58). Placental cells were isolated from 5 male and 6 female samples estimated at 12–17-weeks for 29,876 cells. Placental cells were broken down into 12 main groups (e.g. stromal, decidual, labyrinthine trophoblast) based on canonical markers(39, 40, 52, 59–61) and further dissected into 29 different subtypes (Syncytiotrophoblasts types 1 through 8). In a complementary study, Domcke et al. (2020) assessed the single-cell chromatin accessibility of 334,718 placental cells (3 female, 2 male) at 15–17-weeks (56). This information aided in annotating cells from the RNA dataset and vice versa. Placental-specific observations included the mis-annotation of sex from some of the cells in the RNA dataset due to surprisingly low levels of expression from the Y chromosome, which were maternal decidual or endometrial cells. Accessibility of transcription factors and motif enrichment analysis confirmed the FOS::JUN motif known to be active in EVTs and GATA::TAL1 motifs of erythroblasts in unannotated placental clusters. Shared amongst these comprehensive atlas studies were comparisons to mouse single-cell atlases(37, 60), which confirmed known similarities and identified many divergences to be further explored.

Extending placental single-cell studies to pregnancy complications.

Given the presumptive role of pathobionts, or aberrant microbiome community structures of functions, in several complications of pregnancy(62–65), it is worthwhile reviewing the available, albeit limited literature on single-cell technologies describing variant host immune profiling with these same disorders. Thus far, placental scRNA-seq approaches have primarily focused on host transcription but future studies will likely focus on host-microbe interactions to examine potential functional roles for microbes in host response, tolerance, and fetal immune programming. In this section we will limit our discussion to focus on recurrent pregnancy loss, preterm birth, gestational diabetes, and the contemporaneous threat of COVID-19.

Recurrent pregnancy loss (RPL).

Two placental scRNA-seq studies have focused on RPL. First, Saha and colleagues (2020) examined the potential role of TEAD4 in idiopathic RPL in mice, in human 1^st-trimester placentas (6 and 7.6 weeks), and in trophoblast stem cell cultures derived from RPL patients.(66) By scRNA-seq, they identified 22 clusters in the two human placenta samples from our estimated 14,603 high quality cells. Their annotated clusters were based on canonical markers, and they consistently observed an association between expression of TEAD4 and undifferentiated cytotrophoblasts with RPL. This was validated in mouse embryos by immunofluorescence, and the importance of TEAD4 in implantation was further confirmed in crosses of TEAD4-knockout mice, and by siRNA knockdown in trophoblast-like stem cell cultures.

An additional study addressed RPL by scRNA-seq analysis from 24 human 1^st-trimester decidual samples (9 RPL and 15 healthy) that were presorted for CD45⁺ cells.(67) From RPL specimens, 8,504 cells were compared to 10,142 control cells. Clustering identified progenitors, and immune cells including macrophages, NK cells, and T cells. Comparisons of the healthy and RPL samples showed an increase in T cells, decreased macrophages, and no change in NK cell proportions, which were validated by flow cytometry of additional samples (24 healthy and 23 RPL). Analysis of known decidual NK cell subsets (originally described in (40)) and a subset of progenitor NK cells (distinguishing markers not apparent from the figures or text) revealed RPL-associated changes in their proportions, where dNK3 cells increased drastically. To complement this single-cell approach, the subsets of dNK cells were sorted and subject to bulk ATAQ-seq to identify changes in chromatin accessibility, revealing multi-modal shifts from angiogenic LILRB1-expressing NK cells (dNK1) to cytokine secreting subtypes associated with RPL (dNK2 and dNK3). Pseudotime trajectory analyses of dNK cells identified a subtype of dNK2 cells that diverge into dNK1 cells (termed “Path T cells”), and RPL-specific developmental fate characterized by increased cytokine signaling associated with dNK3 cells. Path T cells were defined by low levels of CD39 and CD18, verified by additional flow cytometry from RPL and healthy patients. Next, they examined macrophages and identified two subtypes, where “mac1” cells were slightly increased in RPL cells and “mac2” cells were drastically decreased in RPL cells. Pathway analyses and subsequent validation studies revealed mac1 cells were associated with NK cell chemotaxis and healthy controls, while mac2 cells were related to positive regulation of T cell chemotaxis (in CD4⁺, CD8⁺, and FOXP3⁺ T reg cells) and RPL.

Preterm birth.

To date, a single study has performed scRNA-seq of placenta in relation to preterm birth. Pique-Regi and colleagues dissected term chorioamniotic membranes, placental villi, and basal plate samples (with or without labor at 38–40 weeks) and compared with preterm subjects samples (33–34 weeks).(68) Resultant scRNA-seq data from 79,906 cells were analyzed and compared with previously reported markers to identify 19 clusters which differentially associated with term (labor and unlabored) or preterm gestations, including trophoblasts, lymphoid, myeloid, stromal, and endothelial cells. Interestingly, they also identified non-proliferative interstitial cytotrophoblasts (“npiCTBs”) specifically in placental villi, marked by reduced expression of XIST, DDX3X, and EIF1AX with increased levels of PAGE4. There were additional unique cellular phenotypes in the chorioamnion, which were characterized as maternal lymphatic endothelial decidual cells marked by EDNRB, CD34, and TIE1. Subsequent visualization by immunofluorescence of LYVE-1 and CD31 confirmed that these cells were not found in placental villi or the basal plate, suggesting they had been infiltrated by maternal lymphocytes. When comparing cell-type specific transcription in the term labor, term no labor, and preterm subjects samples, cytotrophoblasts and EVTs were consistently observed to bear the most differentially expressed transcripts, with NFKB1 expression in maternal macrophages being associated with labor.(68) As a complementary approach, these authors utilized bulk RNA-seq data from whole blood at different gestational periods to determine if they could identify single-cell transcriptome signatures, which could be used to monitor preterm labor risks non-invasively. They observed subtypes of macrophages in 5-week intervals from week 15–40, “macrophage 2”, monocyte, NK cells, and activated T cell signatures.

Gestational diabetes mellitus (GDM).

Recently, Yang et al. (2021) examined GDM at single-cell resolution.(69) Full-term placentas collected during Cesarean delivery (2 healthy and 2 GDM) were subjected to scRNA-seq, yielding 14,591 GDM and 12,629 control cells for comparison. Nine cell types were identified including VCTs, syncytiotrophoblasts, EVTs, granulocytes, myelocytes, T/NK cells, B cells, monocytes, and macrophages. Subset analyses of trophoblasts and differential expression identified GDM enrichment of estrogen and antigen presentation pathways and decreases in IL-17 signaling. Subpopulations of immune cells were identified and associations with GDM were tested then validated by flow cytometry. They observed significant increases in NK cells and cytotoxic T cells, enhancement of M2 (CD206⁺) macrophages, and a generally decreased inflammatory response (ligand-receptor predictions with lack of RPS19-C5AR1, SPP1-PTGER1, and SPP1-CD44 complexes) in association with GDM.

COVID-19.

In a series of timely studies seeking to determine whether or SARS-CoV-2 retains the capacity for vertical transmission in utero, we observe the potential limitations to utilizing scRNA-seq data as a surrogate for functional viral replication experiments in vitro. Initial reports observed low-level transcript expression of ACE2 or TMPRSS2 in fetal tissues (55), low co-expression in placental cells (56), and almost no transcripts in iPSC-derived embryonic lines.(70) Similarly, single-nuclei RNA-seq (snRNA-seq, to enable sequencing from large multinucleated syncytiotrophoblasts) generated from 26,501 cells derived from n=32 placental villi and decidual 2^nd- and 3^rd-trimester samples failed to generate robust levels of ACE2 nor TMPRSS2 transcripts.(68) Nonetheless, in some cases vertical transmission of SARS-CoV-2 has been shown to occur(71–79) and SARS-CoV-2 was recently shown to replicate very inefficiently in cultured placental cells(80). While more in vivo work is necessary to determine the parameters by which SARS-CoV-2 does or does not transmit to the fetus during development, these publications are a poignant reminder of the limitations to inferring biology based on single transcript expression in large, static datasets.

Host-microbe interactions illuminated by scRNA-seq.

Multi-omic approaches to characterize host-microbe interactions are being applied to single-cell technologies (recently reviewed in (81)). For example, scRNA-seq has been utilized to examine unique host and viral transcription programs during infection of a wide range of viruses including influenza, Dengue, SARS-CoV-2, HSV-1, HSV-2, and CMV.(82–94) Viral transcriptomics provide unique challenges including antisense transcripts, increased mitochondrial transcription as an apoptosis host defense, and known changes to host splicing and transcript elongation.(95–99) Nonetheless, optimizing existing approaches for examining host-immune responses during viral infection of the placenta by scRNA-seq may provide key insights into mechanisms of vertical transmission and the role transient exposures to microbes play in fetal immune development. As an example, placenta scRNA-seq, combined with proteogenomics, TCR/BCR immune-profiling, and spatial transcriptomics in an immunocompetent vertical transmission model (e.g. congenital Zika virus or CMV) would be of great interest.

Bacterial single-cell transcriptomics remains challenging. Individual microbes have low quantities of RNA (about 1 femtogram) compared to host cells (1 picogram).(81) Despite these limitations, single cell technologies have been applied to intracellular microbes (Salmonella spp), and intercellular microbial species (Escherichia coli, Pseudomonas, Staphylococcus aureus, and Bacillus subtilis).(100–104) In addition, a recent study utilized a triple bulk RNA-seq strategy to analyze host-pathogen transcriptomics in DCs co-infected with CMV and Aspergillus fumigatus.(105) Utilizing similar approaches to investigate host-microbe transcripts in gnotobiotic germ-free mice, antibiotic-treated mice, recently microbial colonized gnotobiotic mice, and specific-pathogen free mice would test the hypothesis that microbes, microbial components or products (including metabolites) are key to normal immune tolerance and host metabolic and immune development. Similarly, expansion of ours(64) and others high-resolution spatial 16S analysis combined with single-cell transcriptomics would enable better appreciation of how low biomass communities are potentially maintained as such, and how foci with microbes vary from those free of microbes. Soon, multi-omics approaches optimized for microbes combined with reporter constructs to trace intracellular transmission (106, 107) will allow for biologic directionality and the capacity to delineate longitudinal transcription in the host transcription because of different microbes being present or absent. To our knowledge, Visium spatial resolution of host transcription in the placenta with or without microbes has not yet been reported.

Conclusions and future perspectives.

Applications of single-cell ‘omics technologies in normal and diseased tissues, at various stages in development, have illuminated the degree of cellular heterogeneity. The next step for single-cell applications will include functional biology, with a primary aim to resolve how highly specified populations of immune cells in a single space vary in their transcriptional activity. To date, while human placenta scRNA-seq studies have analyzed over 414,237 cells and identified canonical lineages of immune sub-populations, they have yet to be applied to functional biology. Application of single-cell approaches aimed at understanding immune development in situations known to result in disease in the offspring (namely congenital infections or gnotobiotic, germ free animals) are of paramount importance. In the short-term interval, studies with multi-omic single-cell approaches in the placenta will be vital in determining functional roles of low-biomass and sparse communities in fetal immune development and describing heterogeneous microenvironments where pathobionts or beneficial microbes exist or persist.