Retinoic acid regulates fetoplacental vascularization via notch signaling and a SEMA3E/F-PLEXIND1 axis

Summary

The placenta is vital for fetal development, and altered placental vascularization, the most common placental pathology, underlies prevalent disorders, including fetal growth restriction, prematurity, and pregnancy complications. Impaired placental vascularization is associated with Vitamin A deficiency, but the mechanisms are undefined. To investigate this, we used retinoic acid (RA)-deficient Raldh2^−/− embryos, and found they exhibit allantoic and placental endothelial hyperproliferation and impaired arterial-venous remodeling, which were rescued by providing all-trans-RA (ATRA) via maternal diet. Single-cell RNA sequencing of E9.5 Raldh2^+/+, Raldh2^−/−, and Raldh2^−/− + ATRA placental cells, and functional assays, revealed that RA regulates endothelial growth and vascular remodeling via Notch signaling. We also uncovered a PLEXIND1–SEMA3E/F signaling axis between fetal endothelial cells and chorionic trophoblast precursors that is impaired with RA deficiency and rescued with ATRA. Our data suggest that RA-mediated signaling regulates allantois and placental endothelial cell growth, specification, and guidance required for chorioallantoic fusion and fetoplacental vascularization.

Graphical abstract

Highlights

• •Retinoic acid (RA) deficiency disrupts fetoplacental vascular remodeling
• •RA limits endothelial proliferation via Notch-mediated cell-cycle control
• •RA regulates endothelial guidance via SEMA3–PLEXIND1 signaling
• •Maternal ATRA supplementation rescues placental vascular defects

Molecular biology; Cell biology; Developmental biology

Introduction

The placenta is a unique organ with two separate blood circulatory systems: the maternal-placental and fetal-placental (fetoplacental) systems.¹ Both human and mouse placentation are hemochorial, with maternal blood flowing in direct contact with fetal trophoblasts. Additionally, in both human and mouse, the fetoplacental vascular system extends into syncytiotrophoblast-covered spaces bathed in maternal blood.

Formation of the fetoplacental vascular system begins with the de novo differentiation of endothelial cells (ECs) from mesodermal progenitors, and elaboration of a vascular plexus in the allantois, the precursor of the umbilical cord.^2,3 The allantois then attaches to the chorionic mesoderm and, once fused, the nascent umbilical vessels interact with chorionic trophoblasts.^4,5 Interactions between umbilical ECs and chorionic trophoblasts regulate branching morphogenesis to form a fetoplacental circulatory network⁶; however, underlying mechanisms are undefined.

In the placenta, the establishment of complementary vascular circulatory systems and their tight regulation is critical to achieve efficient exchange of gas, nutrients, and waste between mother and fetus. Altered placental vascularization, one of the most common placental pathologies, affecting 10–50% of pregnancies in the US,⁷ has detrimental consequences, causing early pregnancy loss, intrauterine fetal growth restriction (IUGR), and premature birth, as well as pregnancy complications such as preeclampsia,^8,9 a leading cause of US maternal mortality.^10,11 Major risk factors for placental insufficiency and poor pregnancy outcomes are nutritional deficiencies, including Vitamin A deficiency (VAD)¹²; yet the role of retinoic acid (RA), the bioactive form of Vitamin A, in placental vascularization remains poorly defined and is the focus of our study.

Our previous studies showed that RA plays a critical role in embryonic vascular development via upregulation of cell cycle inhibitors p21 and p27 in developing ECs, causing G1 arrest,^13,14 which enables their arterial-venous specification.¹⁵ In addition, we found that RA signaling promotes hemogenic EC specification in a Notch- and cell cycle-dependent manner.¹⁶ However, no studies have yet defined the role of RA and downstream effectors in fetoplacental vascularization.

We investigated the role of RA in chorioallantoic fusion and fetoplacental vascularization using RA-deficient (Raldh2^−/−, lacking retinaldehyde dehydrogenase 2 and the ability to generate RA from Vitamin A¹⁷; and Raldh2^+/+ wild type embryonic day (E)8.0-E9.5 mouse embryos. Deletion of the Raldh2 gene leads to embryonic lethality at ∼ E11 due to defects in hindbrain and heart development. In this study, we investigated fetoplacental vascularization at early stages of embryogenesis and utilized a short-term maternal all-trans-RA (ATRA) supplementation to rescue the lethality of Raldh2^−/− embryos and assess the role of RA in placental development. Although RA-rescued mutants still develop some organ abnormalities, such as persistent truncus arteriosus, that prevent postnatal survival, they are largely comparable to wild-type embryos up to E14.5.¹⁷

Using immunofluorescence imaging and quantitative analyses, we found that RA deficiency causes allantoic and placental EC hyperproliferation and impaired arterial-venous remodeling, and these defects could be rescued by feeding pregnant females ATRA. To gain a mechanistic understanding of how RA regulates these processes, we performed single-cell RNA sequencing (scRNA-seq) of placental cells isolated from E9.5 Raldh2^+/+ and Raldh2^−/− embryos, as well as Raldh2^−/− + ATRA. Data analyses and functional assays revealed that RA regulates EC growth and vascular remodeling via Notch signaling. We also uncovered a PLEXIND1-SEMA3E/F signaling axis between fetal ECs and chorionic trophoblasts that is impaired with RA deficiency and rescued with ATRA. Collectively, our data suggest that RA-mediated signaling and downstream Notch and Semaphorin/Plexin effectors regulate allantois and placental EC growth, specification, and guidance that are required for chorioallantoic fusion and fetoplacental vascularization.

Results

RA deficiency impairs allantois and fetoplacental vascular development

Vitamin A is converted to RA by retinaldehyde dehydrogenase (RALDH) enzymes. Raldh2 is predominantly expressed in early embryonic development, and its deletion (Raldh2^−/−; further referred to as Raldh2 KO) causes embryonic RA deficiency.¹⁷ In our studies, the pregnant mother is heterozygous for Raldh2 deletion (Raldh2^+/−). Therefore, these embryos provide a model to investigate RA function in allantois and fetoplacental vascularization without indirect maternal effects.

We collected Raldh2^+/+ (further referred to as Raldh2 WT) and Raldh2 KO placentas at E8.5 and E9.5, as shown in Figure 1A, and performed whole-mount immunohistochemical studies for EC-enriched proteins, namely ERG (nuclei) and CD31 (membrane). At E8.5, primitive vascular plexi formed in both Raldh2 WT and KO placentas (Figure S1A); however, at E9.5, Raldh2 KO placentas failed to develop large-caliber, lumenized vessels, resulting in decreased vessel diameter compared to Raldh2 WT placentas (Figures 1B and 1C). Moreover, this defect was associated with a higher number of ERG⁺ ECs (Figure 1D) and proliferative phospho-HISTONE H3 (pHH3)⁺ ECs (Figures 1E and S1B) in Raldh2 KO placentas. We did not observe changes in apoptosis levels in E9.5 Raldh2 WT and KO placentas, quantified via immunostaining for cleaved CASPASE-3 (Figure S1C). Interestingly, when pregnant Raldh2^+/− females (gestational stage E7.5) were fed chow supplemented with 1 μM ATRA, the vascular development (Figures 1B–1D) and EC hyperproliferation (Figures 1E and S1B) defects in the KO placentas were prevented. Importantly, this RA treatment did not impair the vascularization of Raldh2 WT placentas (Figures 1C–1E and S1B), consistent with previous studies showing this RA dose does not cause vascular development defects.¹⁴

Figure 1.

RA is required for allantois and fetoplacental vascular development

(A) Schematic illustrating mouse placenta collection the whole-mount immunostaining at E8.5–9.5 (left panel), and schematic illustrating arterial and venous vessels in the developing placenta, indicating that placental arteries transport deoxygenated fetal blood to the placenta for exchange while veins that return oxygenated blood to the fetus; key endothelial markers used in this study are annotated (arterial: DLL4; venous: ENDOMUCIN) (right panel; adapted from.¹⁸

(B–E) (B) Confocal images showing E9.5 Raldh2 WT and KO (+/−ATRA, 100 mg/g) whole-mount placentas immunostained for ECs (anti- CD31 and ERG). Scale bars, 200 μm. Bar graphs show quantification for CD31⁺ vessel diameter (C), ERG⁺ ECs (D), and pHH3⁺/ERG⁺ ECs (E).

(F) Confocal images show E9.5 Raldh2 WT and KO (+/−ATRA, 100 mg/g) whole-mount placentas immunostained for arteries (anti-DLL4), and veins (anti-ENDOMUCIN).

(G and H) Scale bars, 200 μm. Bar graphs show DLL4-fluorescence intensity (G) and ENDOMUCIN-fluorescence intensity (H) in E9.5 Raldh2 WT and KO (+/−ATRA) placentas.

(I) Schematic illustrating mouse allantois collection and explant culture.

(J) The timeline of allantois explant culture.

(K) Confocal images show allantois explants from E8.0 Raldh2 WT and KO placentas (+/−RA, post-48h culture) immunostained for ECs (anti-ERG), arteries (anti-DLL4), and veins (anti-ENDOMUCIN).

(L) Scale bars, 100 μm.

(M and N) Bar graphs show the quantification for ERG⁺ EC (L), EdU⁺ EC (M), and vessel branching points (N). N ≥ 4 different animals/group, 5 regions of interest analyzed/group. ns: not significant; ∗p < 0.05, ∗∗p < 0.01, ∗∗∗p < 0.001, and ∗∗∗p < 0.0001 (Data are mean ± SEM, one-way ANOVA with Holm-Šidàk’s multiple comparison test). A and I Created with BioRender.com.

Since past studies from our group have shown a link between EC proliferation and arterial-venous specification,^15,19,20 we next assessed arterial and venous EC identity in E9.5 Raldh2 WT and KO placentas immunostained with antibodies against arterial- (DLL4) and venous-enriched (ENDOMUCIN) proteins. Raldh2 KO placentas exhibited fewer DLL4⁺ vessels compared to WT (Figures 1F, 1G, and S1D), and this defect was prevented by RA supplementation (Figures 1F, 1G, and S1D). Despite Raldh2 KO placentas exhibiting architectural venous network defects, we did not observe quantitative differences in ENDOMUCIN expression in Raldh2 KO vs. WT (Figures 1F, 1H, and S1D). Together, these data demonstrate the important role of RA in regulating placental EC proliferation and arterio-venous vessel development.

To determine whether RA affects earlier stages of fetoplacental vascularization (e.g., in the allantois and umbilical vessels), we used the allantois explant culture assay, as depicted in Figure 1I. We collected allantoides from E8.0 Raldh2 WT and KO embryos, and explants were cultured for 48 h +/−1 μM RA (Figure 1J), a dose that did not affect WT allantois vascularization (Figures 1L–1N and S1E). We observed an increased number of ERG⁺ ECs (Figures 1K and 1L) and proliferative EdU⁺ ECs (Figures 1M and S1E) in Raldh2 KO allantoides compared to WT. These defects were associated with the impaired formation of DLL4⁺ arteries and vascular remodeling, characterized by increased vessel branching in Raldh2 KO allantoides (Figures 1K and 1N). Exogenous ATRA rescued the EC proliferation and vascular remodeling defects observed in Raldh2 KO (Figures 1M, 1N, and S1E). These collective data support that RA is an important regulator of allantois and placental EC proliferation and vascular development.

scRNAseq of E9.5 placental cells reveals dysregulated EC specification in the Raldh2 KO

To gain insight into mechanisms by which RA regulates fetoplacental vascularization, we performed scRNAseq analysis of E9.5 Raldh2 WT and KO placental cells (+/−ATRA). For each sample, the collected tissues were pooled to prepare a single-cell suspension, in which red blood cells were lysed, and ECs were enriched using anti-CD31 microbeads. To all samples, 5% of all live placental cells were added back, and all cells underwent scRNAseq using the 10X Genomics Chromium system (Figure 2A), followed by downstream analysis with Cell Ranger and Seurat.^21,22 We obtained ∼28,000 mean reads/cell with a median of ∼2,000 genes/cell across all samples. Fourteen distinct cell clusters were identified using unsupervised graph-based clustering (Figures S2A and S2B). Each cluster was annotated based on the expression of known and previously reported marker genes.^{23,24,25,26,27,28} Within the clusters, we identified fetal ECs as exhibiting high relative mRNA expression of Pecam (CD31) and Kdr (Figure S2A). Further analysis of the fetal EC cluster, based on the expression of known subtype-enriched genes, identified five subtypes, including arterial, capillary (cap)-arterial, venous, cap-venous, and proliferating ECs (Figures 2B and S2C).

Figure 2.

scRNAseq analysis of Raldh2 WT, KO, and KO + ATRA placental cells reveals EC hyperproliferation and impaired specification

(A) Schematic illustration of the scRNAseq experimental setup. Created with BioRender.com.

(B) Dot plot of fetal EC subtype marker expression.

(C) PHATE dimensionality reduction plot of fetal ECs shows annotated EC subtypes in the WT, KO, WT + ATRA, and KO + ATRA samples.

(D) Stack plot and table show proportions of fetal EC subclusters in the WT, KO, WT + ATRA, and KO + ATRA samples.

To investigate the effects of RA on EC identity, we used potential of heat-diffusion for affinity-based trajectory embedding (PHATE)²⁹; dimensionality reduction, to visualize and evaluate the EC subtype distribution across experiments. PHATE organizes cells into a tree-like hierarchy and captures continuous trajectories based on the cell-to-cell gene expression continuums, enabling us to adequately map the transitional states of EC specification in WT and KO placental tissues. PHATE analysis and quantification of EC subtypes revealed that the proportions of cap-arterial ECs were reduced in the Raldh2 KO placentas, compared to WT, and this distribution was restored to WT levels with ATRA supplementation (Figures 2C and 2D). Additional analysis, using arterial and venous gene expression module scores (Table S1), further supports that arterial and venous relative gene expression were reduced in Raldh2 KO placentas and higher with ATRA supplementation (Figures S2D and S2E).

Consistent with results shown in Figure 1, scRNAseq analysis revealed a higher proportion of proliferating ECs in Raldh2 KO placentas, compared to WT, which was decreased in the KO + ATRA group (Figure 2D). Further analysis of the fetal ECs for the expression of known cyclin dependent kinase (CDK) inhibitors, Cdkn1c (p57), Cdkn1b (p27), Cdkn1a (p21), Cdkn2a (p16), and Cdkn2b (p15^INK4b), performed on an equal number of ECs across conditions, showed lower expression of CIP/KIP family members Cdkn1c (p57), Cdkn1b (p27), and Cdkn1a (p21) in KO ECs, compared to WT, and these levels were elevated in the KO + ATRA group (Figure S2F). Expression of INK4 family members, Cdkn2a (p16^INK4a) and Cdkn2b (p15^INK4b), was minimal across conditions. These collective transcriptomics analyses are consistent with our immunofluorescence studies (Figures 1 and S1) and further support that RA deficiency leads to placental EC hyperproliferation and impaired specification, and these defects are rescued with ATRA supplementation.

Notch signaling is downregulated in Raldh2 KO placental ECs

To identify potential regulators of placental EC proliferation and vascular development that function downstream of RA, we performed pathway enrichment analysis of our scRNAseq data from WT vs. KO fetal ECs. We found Notch signaling to be among the significantly enriched pathways in WT vs. KO ECs (Figure 3A). We then further explored the differences in Notch signaling-related gene expression in fetal ECs using a Notch pathway gene expression module score (Table S2), and found the expression of key pathway components to be enriched in our dataset, namely the receptors Notch1 and Notch4, ligand jagged 2 (Jag2), and Notch target gene, Hes1. On a heatmap representation, we show that those genes are downregulated in Raldh2 KO placental ECs, compared to WT, and increased in the KO ECs with ATRA supplementation (Figure 3B).

Figure 3.

Notch signaling functions downstream of RA to regulate EC proliferation and vascular remodeling

(A) Pathway enrichment analysis of differentially expressed genes in ECs of WT vs. KO scRNAseq data.

(B) Heatmap shows the differential expression of genes related to the Notch pathway in Raldh2 WT, KO, and KO + ATRA ECs.

(C) Schematic illustrating mouse allantois collection, and the timeline of allantois explant culture. Created with BioRender.com.

(D) Confocal images of allantois culture immunostained for proliferation (EdU) and EC nuclei (ERG) in WT and KO allantois explants cultured +/−1 μM ATRA, with or without 10 μM Notch activation inhibitor DAPT (gamma-secretase inhibitor). Scale bars, 20 μm. Yellow arrowheads show pHH3⁺/ERG⁺ ECs. (N = 4–7 animals/group, 5 regions of interest analyzed/group).

(E) Assessment of proliferation in allantois culture by EdU uptake in the KO allantois explant cultures ± RA and ± DAPT.

(F) Bar graph shows the total number of vascular branching points in KO allantois explant culture ± RA and ± DAPT. (N = 4–7 animals/group, 5 regions of interest analyzed/group). ∗p < 0.05 and ∗∗∗p < 0.0001 (Data are mean ± SEM, one-way ANOVA with Holm-Šidàk’s multiple comparison test).

To test whether Notch signaling functions downstream of RA to regulate fetoplacental EC proliferation and vascular remodeling, we used the allantois explant culture assay. Raldh2 KO allantois explants were cultured with 1 μM ATRA, with or without 10 μM Notch activation inhibitor DAPT (gamma-secretase inhibitor), as depicted in Figure 3C. In Raldh2 KO, we found that Notch inhibition abrogated the beneficial effects of RA (shown in Figure 1M) in inhibiting EC proliferation (as measured via EdU uptake) (Figures 3D and 3E), and in promoting allantois vascular remodeling (Figure 3F). These collective results suggest that Notch signaling functions downstream of RA to regulate fetoplacental EC proliferation and vascular remodeling.

RA deficiency dysregulates placental vascular branching morphogenesis and SEMA3E/F- PLXND1-mediated interactions between fetal EC-chorionic trophoblasts

To investigate the impact of RA on vascular branching morphogenesis in the placenta, we performed confocal imaging of E9.5 WT, KO and KO + ATRA placental cross-sections immunostained for LAMININ (marks fetal side of placenta), ISOLECTIN-BS1 (chorionic trophoblasts), and ERG (EC nuclei), and measured the distance from the edge of the fetal placenta to the end of the vasculature (ERG⁺ cells), branching toward the maternal side (Figures 4A and 4B). We found that, in the absence of RA in Raldh2 KO, branching morphogenesis is reduced compared to WT, as assessed by the distance of penetration of ERG⁺ ECs/LAMININ⁺ cells toward ISOLECTIN-BS1⁺ trophoblasts. Supplementation with ATRA restored branching distance in Raldh2 KO placentas to WT levels (Figures 4B and 4C).

Figure 4.

RA regulates fetal EC-trophoblast interactions during chorioallantoic fusion and placental vascular branching morphogenesis

(A) Schematic illustrating a placental cross-sectional view.

(B) Confocal images of E9.5 placenta cryo-sections (10 μm) immunostained for LAMININ to show the fetal side of the placenta, ISOLECTIN BS1 to identify chorionic trophoblasts, and ERG to label ECs. Scale bars, 50 μm.

(C) Bar graph shows the quantification of average vessel branching distance (d) within the placenta. Data are represented as mean ± SEM.

(D) Heatmap of the aggregated scaled expression of semaphorin and plexin pathway components in all the cells from scRNAseq data from Raldh2 WT and KO, and KO + ATRA.

(E) Violin plots of Plxnd1, Sema3e, Sema3f, and Vegfa genes in WT placental cell types.

(F) Cell chat analysis of Plxnd1-Sema3e/f signaling strength between Raldh2 WT and KO (+/−ATRA) placental cells.

(G) Ex vivo culture of pre-placental tissues for 24 h with or without Sema3-Plxnd1 pathway inhibition using furin inhibitors (FI). The top panel illustrates the experimental setup of collecting chorion and allantois from Raldh2 WT and KO embryos and culturing them ex vivo. The lower panel illustrates the representative confocal images of vibratome sections (250 μm) of the chorioallantoises stained for ISOLECTIN BS1 to identify chorionic trophoblasts, and ERG to label ECs. Scale bars, 50 μm.

(H) Bar graph shows the quantification of the number of ERG⁺ ECs over ISOLECTIN BS1⁺ trophoblast within the chorioallantois. A and G Created with BioRender.com. ∗p < 0.05 and ∗∗∗∗p < 0.00001 (Data are mean ± SEM, one-way ANOVA with Holm-Šidàk’s multiple comparison test).

To gain insights into RA-regulated molecular cues that may be guiding fetal EC-trophoblast interactions that enable chorioallantoic fusion and placental vascular branching morphogenesis, we analyzed our scRNAseq datasets from E9.5 Raldh2 WT, KO, and KO + ATRA placental cells, with a particular focus on reciprocal signaling pathways that may regulate these processes. On a heatmap representation, we show that semaphorins (Sema3e and Sema4d), the chemorepellents known to play a role in EC guidance and vascular branching morphogenesis in the retina and kidney,^30,31 are downregulated in Raldh2 KO, and their expression is elevated in KO treated with ATRA. The expression of plexins (Plxnd1, Plxna3, and Plxna4), the known receptors for semaphorins, is upregulated in the KO cells, compared to WT, and reduced in the KO + ATRA, similar to multiple VEGF signaling components, including Vegfa, which is known to upregulate Plxnd1 (Figure 4D). Placental cell type-specific expression analysis revealed that Sema3e, Sema3f, and Vegfa are enriched in Syncytio-Trophoblast II precursor (SynTII-P) and sinusoid trophoblast giant cell precursor (S-TGC-P) chorionic trophoblasts, while Plnxd1 is enriched in fetal ECs (Figure 4E).

We then performed Cell Chat analysis of our scRNAseq data³² to interrogate Plxnd1-Sema3e/f interaction strength among placental cell types in the E9.5 Raldh2 WT, KO, and KO + ATRA groups. CellChat analysis predicts potential contact- and non-contact-dependent paracrine signaling pathways among cell types represented within a dataset. In WT cells, interactions between fetal ECs and S-TGC-P and Spongiotrophoblast cell precursor/spiral artery trophoblast giant cell/glycogen cells (SpT/SpA-TGC/GC) were high, as indicated by thick red and brown lines, respectively, linking those cell populations. In contrast, these interaction strengths were reduced in the KO cells, as depicted by thinner red and brown lines between these cell types. Interestingly, the fetal ECs and S-TGC-P interaction was restored in the KO + ATRA condition, where branching morphogenesis is restored to WT levels (Figure 4F). To functionally test the importance of the predicted SEMA3E/F-PLEXIND1 signaling axis in mediating fetal EC and chorionic trophoblast interactions during branching morphogenesis, we performed ex vivo culture of pre-placental tissues³³ for 24 h with or without SEMA3-PLEXIND1 pathway inhibition using furin inhibitors,³⁴ which block the proteolytic activation of SEMA3 ligands required for PLEXIND1 engagement. Inhibition of this pathway resulted in impaired vascular branching, phenocopying the defects observed in Raldh2 KO placentas (Figures 4G and 4H), thereby validating the functional relevance of predicted SEMA3–PLEXIND1 interactions in this process. These results suggest that SEMA3E/F-PLXND1 signaling may function downstream of RA to regulate fetal EC-chorionic trophoblast interactions to guide chorioallantois fusion and vascular branching morphogenesis.

Collectively, our data and functional tests suggest a working model (Graphical Abstract) in which RA functions to upregulate Notch signaling in fetal ECs to promote their growth inhibition, which is required for arterial-venous specification.¹⁵ Additionally, as suggested by our transcriptomic data, RA may function to directly inhibit VEGF-A and increase SEMA3E/F production, as it does in other cell types.^35,36 Thus, in the absence of RA, Notch signaling is downregulated in fetal ECs, leading to their hyperproliferation and impaired specification that contributes to dysregulated vascular remodeling and branching morphogenesis. In addition, SEMA3E/F-PLXND1 signaling is dysregulated and may contribute to impaired fetal EC-chorionic trophoblast interactions that guide chorioallantoic fusion and fetoplacental vascular branching morphogenesis.

Discussion

Impaired vascularization, the most common placental pathology, is associated with VAD and leads to pregnancy loss, IUGR, preterm birth, and pregnancy complications. We sought to better understand mechanisms by which RA, the bioactive form of Vitamin A, regulates placental vascularization, especially in the first trimester of pregnancy when VAD leads to congenital malformations.^11,12,37

Using the Raldh2^−/− mouse model, combined with immunohistochemical and transcriptomic analyses and functional assays, we found that RA deficiency leads to allantois and placental EC hyperproliferation and impaired vascular remodeling, and these defects could be prevented/rescued by providing the embryos bioactive ATRA via maternal diet. While interventions, such as maternal Vitamin A supplementation, are standard clinical practice and have been shown to improve maternal and fetal outcomes,^8,38 the exact mechanisms of these treatments were not known, limiting the ability to fine-tune their application for specific pregnancy complications. Thus, further studies with this and other animal models in which RA dosing during pregnancy can be manipulated may be useful for further dissecting Vitamin A needs and effects on placental vascularization and fetal development.

Our analyses further revealed that RA mediates effects on fetoplacental EC proliferation and vascular remodeling via Notch signaling. This is consistent with our previous studies of hemogenic EC specification, wherein we showed that RA-activated Notch signaling upregulates p21 and p27 to promote EC cycle arrest, which is required for hemogenic specification.¹⁶ In addition, during retinal vascularization, we showed that shear-induced Notch-p27 signaling is required for EC growth control and arterial-venous specification.^15,20 In hyperproliferative RA-deficient placental ECs, p21 and p27, as well as p57, are downregulated, and their expression is “normalized” with ATRA supplementation. Thus, this RA-Notch-p27 signaling axis may also be driving the observed EC growth inhibition in ATRA-treated KO placentas, to enable arterial-venous development therein.

Other studies have shown that RA and Notch pathways interact during embryonic development in a context-dependent manner. For instance, Liu and colleagues³⁹ described a Nkx2-5/Notch/RA axis regulating cardiac hematopoiesis, while Ying and coworkers⁴⁰ demonstrated that RA induces the early transcriptional activation of developmental regulators, including Notch and Wnt. These findings suggest that RA can modulate Notch activity either directly through RAR/RXR binding to promoter regions or indirectly via upstream transcription factors. In a preliminary analysis of the Notch1 promoter, we found non-canonical RARE motifs (DR9 and IR0) within the proximal region, which could enable direct RAR/RXR binding and RA-mediated transcriptional activation. This, together with the upregulation of Notch1 and Hes1 following ATRA treatment and the loss of this effect with DAPT, supports a model in which RA acts upstream of Notch signaling to control EC proliferation and promote branching morphogenesis, which is similar to its role in the specification of hemogenic EC.^13,14

Our studies also showed that interactions between fetal ECs, which arise in the allantois to form the umbilical vessels, and chorionic trophoblasts, are dysregulated by RA-deficiency and normalized with ATRA supplementation. Although these cell-cell interactions are known to direct chorioallantoic fusion and branching vascular morphogenesis to form the fetoplacental circulatory network, their regulatory mechanisms were undefined. Based on our immunohistochemical studies and transcriptomic analysis, we propose that “normal” levels of VEGF-A, produced by chorionic trophoblasts, promote angiogenic sprouting of fetal ECs,³⁰ inducing tip cell formation and branching toward the chorion. We further propose that SEMA3 ligands, via PLEXIND1 receptors, guide the tip cells toward the trophoblasts, to enable chorioallantoic fusion and patterning of the labyrinthine network (Graphical Abstract). In the absence of RA, Vegfa and Plxnd1 are upregulated, and Sema3e is downregulated, disrupting this tightly controlled process.

The importance of Vegfa signaling for regulating EC differentiation in chorionic villi of human placentas has been reported⁴¹; however, roles for SEMA3E/F-PLXND1 in fetoplacental vascularization were not previously identified. In support of our findings, sema3f KO mice die at E10 with impaired labyrinth development,⁴² and the expression of Sema3f is reduced in cases of pregnancy complications.⁴³ Furthermore, consistent with our computational predictions, ex vivo inhibition of SEMA3–PLEXIND1 signaling recapitulated the branching defects observed in Raldh2 KO embryos, supporting a functional role for this signaling axis in regulating fetal EC guidance and chorioallantoic vascular morphogenesis. Given that these pathways regulate diverse EC functions (proliferation, migration, guidance, and identity)⁴⁴; in other tissues, their role in the unique fetoplacental vascularization process warrants further investigation.

In summary, our studies identify RA-mediated signaling as a central regulator of allantois and placental EC growth, specification, and guidance required for chorioallantoic fusion and fetoplacental vascularization. This work suggests that modulating RA signaling, and/or its downstream Notch and Semaphorin/Plexin effectors, could have therapeutic potential in restoring or enhancing placental vascular integrity in compromised pregnancies to improve fetal health and survival.

Limitations of the study

This study relies primarily on the Raldh2^−/− mouse model, which represents a severe and global loss of RA synthesis and may not fully reflect the spatial, temporal, or partial RA deficiencies observed in human pregnancy. Although maternal ATRA supplementation rescued key vascular defects, this approach does not distinguish direct endothelial effects from indirect effects mediated by trophoblasts or other placental cell types, nor does it define precise ATRA dose or timing requirements. In addition, while our data support RA acting upstream of Notch and Semaphorin/Plexin signaling, evidence for direct transcriptional regulation remains preliminary. Finally, ex vivo and computational analyses cannot fully capture the dynamic in vivo maternal-fetal environment. Future studies using cell type-specific and dose-controlled models will be important to refine these mechanisms and their translational relevance.

Resource availability

Lead contact

Further information and requests for resources and reagents should be directed to and will be fulfilled by the lead contact, Karen K. Hirschi, kkh4yy@virginia.edu.

Materials availability

This study did not generate new unique reagents. Any further information about reagents and requests for resources should be directed to and will be fulfilled by the lead contact.

Data and code availability

• •Data: Single-cell RNA sequencing (scRNAseq) data have been deposited at NCBI Gene Expression Omnibus (GEO) and are publicly available as of the date of publication. Accession numbers are listed in the key resources table.
• •Code: All original code has been deposited at GitHub and is publicly available as of the date of publication. The repository is listed in the key resources table.
• •Any additional information required to reanalyze the data reported in this paper is available from the lead contact upon request.

Acknowledgments

This work was supported by NIH grants R01DK118728 (KKH), R01HL146056 (KKH), R01HL171284 (KKH), AHA Predoctoral Fellowship 23PRE1026720, The Robert R. Wagner Fellowship Fund (AC), The Robert R. Wagner Fellowship Fund (ZM), AHA Career Development Award CDA1051175 (NG), AHA Career Development Award CDA938744 (GG), F31 CA288057 (SC), F31 HL172647 (JA). University of Virginia Genome Analysis and Technology Core (RRID:SCR_018883, 5P30CA044579) performed the single-cell RNA sequencing. University of Virginia School of Medicine Research Histology Core Facility (RRID:SCR_025470) performed frozen tissue sectioning.

Author contributions

K.K.H. and A.C. planned and carried out the experiments. U.P. performed the scRNAseq data analysis, and all authors contributed to the interpretation of the results. K.K.H. and A.C. wrote the first draft of the article. All authors provided critical feedback and helped shape the research and analysis. All authors reviewed and approved the final article.

Declaration of interests

Authors declare no competing interests.

Declaration of generative AI and AI-assisted technologies in the writing process

Authors declare no usage of AI or AI-assisted technologies.

STAR★Methods

Key resources table

REAGENT or RESOURCE	SOURCE	IDENTIFIER
Antibodies
Rabbit Anti-Erg	Abcam	RRID: AB_2630401
Donkey Anti-Rabbit Alexa Fluor 405	Abcam	RRID: AB_2923541
Rat Anti-CD31	BD	RRID: AB_393571
Donkey Anti-Rat Alexa Fluor 594	Invitrogen	RRID: AB_2535795
Goat Anti-DLL4	R&D Systems	RRID: AB_354770
Donkey Anti-Goat Alexa Fluor 488	Abcam	RRID: AB_2687506
Rat Anti-ENDOMUCIN	Fisher Scientific	RRID: AB_891527
Rat Anti-LAMININ	Sigma	RRID: AB_1587233
FITC-lectin from Bandeiraea simplicifolia (Griffonia simplicifolia)	Sigma	RRID: AB_2314051
Chemicals, peptides, and recombinant proteins
PBS	Sigma	Cat# T5648
Sucrose	Sigma	Cat# C8267
OCT	Gibco	Cat# 10010023
Super Frost Microscope Slides	Fisher Scientific	Cat# S5-3
Triton X-100	Sakura	Cat# 4583
BSA	Fisher Scientific	Cat# 22-037-246
Fluoromount-G	Sigma	Cat# 9002-93-1
HBSS	SouthernBiotech	Cat# 0100-01
Collagenase I from Clostridium histolyticum	Sigma	Cat# C0130
Retinoic acid (RA)	Sigma	Cat# 302-79-4
DAPT	Fisher Scientific	Cat #AAJ65864MA
Normal Donkey Serum	Southern Biotech	Cat# 0030-01
All trans retinoic acid (ATRA)	Sigma	Cat #302-79-4
Furin Inhibitor	EMD Millipore	Cat #344930-1 MG
Critical commercial assays
Click-iT® EdU Imaging Kit	Fisher Scientific	Cat# C10337
Emeraldamp® GT PCR Master Mix	Takara Bio USA	Cat# RR310A
CD 45 microbeads	Miltenyi	Cat# 130-097-418
CD 31 microbeads	Miltenyi	Cat#130-052-301
Deposited data
Single cell RNA sequencing of E9.5 placental cells.	This Study	GEO: GSE304388
Software and algorithms
Custom analysis code	This paper; GitHub	https://github.com/udp3f/ScRNAseq_E9.5_placenta
Experimental models: Organisms/strains
Raldh2	(Niederreither 1999)12	N/A
Software and algorithms
CellRanger	Illumina	v8.0.1
R	R project	v4.3.3
RStudio	Posit	v2023.12.1 + 402
Seurat	(Zheng 2017; Stuart 2019)21,22	v4.4
Graphpad Prism	Graphpad	V10
Adobe Illustrator	Adobe	N/A
Angio Tool	National Cancer Institute; Zudaire et al.	N/A
ImageJ	NIH	N/A
Other
Leica DMi8 Lightning Confocal (SP8)	Leica	RRID:SCR_018169
Next Seq 2000	Illumina	RRID:SCR_023614
MiSeq	Illumina	RRID:SCR_016379

Experimental model and study participant details

Mice

All animal procedures were approved by the University of Virginia Animal Care and Use Committee (protocol No. 4277) and adhered to all ethical guidelines. Animal housing was controlled by University of Virginia Vivarium Staff, maintaining a 14 h light to 10 h dark cycle, an ambient temperature of 21.5 °C (range 20°C–26 °C), and ambient humidity of 40% (range 30–70%).

In this study, endogenous expression of the murine Raldh2 (retinaldehyde dehydrogenase 2) gene was disrupted in CD1 mice, as previously described.¹⁷ Maintenance of Raldh2^+/− heterozygote breeding pairs was achieved through sibling mating. Timed pregnancies were established by paired matings, with the morning of vaginal plug detection designated as embryonic day (E)0.5. Tissues were isolated from embryos at E8.0 (allantoides), E8.5, and E9.5 (placentas).

Sex as a biological variable

Sex was not determined for E8.0–E9.5 embryos due to the lack of phenotypical sexual dimorphism at these early developmental stages. This is acknowledged as a limitation of the study.

Method details

Genotyping

Genotyping of embryos was performed by PCR using gene-specific primers flanking the murine Raldh2 gene locus. Genomic DNA was extracted from whole tissue lysates by digesting tissue in 50 mM NaOH at 95°C for 15 min, followed by neutralization with 100 mM Tris-Cl (pH 8.0). PCR was conducted using Emeraldamp® GT PCR Master Mix (Takara Bio USA).

Allantois explant culture

Allantoides were isolated from E8.0 Raldh2^+/+ (WT) and Raldh2^−/− (KO) embryos using an established method.⁴⁵ Allantoides were cultured on a 12-well plate on the coverslip covered with 0.1% gelatin, with the DMEM media. Explants were cultured with DMSO for the control, or with retinoic acid (RA) (1 μM) (CAS #302-79-4; Sigma-Aldrich, St. Louis, MO), or DAPT (10 μM) (Cat #AAJ65864MA; Thermo Fisher Scientific, Waltham, MA) for 48 h, then treated with Click-iT® EdU Imaging Kit (Cat# C10337; Thermo Fisher Scientific, Waltham, MA) to assess proliferation and fixed by 4% PFA in PBS for 20 min. Allantoides were permeabilized and blocked for 30 min using 0.1% Triton™ X-100 (Cat# T8787; Sigma-Aldrich, St. Louis, MO) with 10% Normal Donkey Serum (NDS) (Cat# 0030-01; Southern Biotech, Birmingham, AL) in PBS. Primary and Secondary antibodies were dissolved in 0.1% NDS in PBS. We assessed the extent of vascular remodeling and arterial-venous development in Raldh2 KO vs. WT allantoides via immunohistochemistry (IHC) using antibodies against EC nuclei (ERG- Rabbit Monoclonal ERG antibody conc. 1:100) (Cat# ab92513; Abcam, Waltham, MA), EC membrane (CD31- Rat Monoclonal CD31 antibody conc. 1:100) (Cat# 550274; BD, Franklin Lakes, NJ), arterial ECs (DLL4- Mouse DLL4 Antibody conc. 1:200) (Cat# AF1389; R & D Systems, Minneapolis, MN), and venous ECs (Endomucin-Endomucin Rat anti Mouse Monoclonal Antibody conc. 1:200) (Cat# 14-5851-82; Thermo Fisher Scientific, Waltham, MA) enriched proteins, with fluorescent secondary antibodies. The following secondary antibodies were used: D&Rabbit 405 (Cat# ab175651; Abcam, Waltham, MA; conc. 1:500); D&Goat 488 (Cat# ab150129; Abcam, Waltham, MA; conc. 1:500); D&Rat 594 (Cat# A21209; Invitrogen, Waltham, MA; conc. 1:500). Stained explants were imaged using confocal microscopy (Leica SP8) for EdU quantification and further analysis. Using Angio Tool software and established methods,⁴⁶ the number of vessel branch points, as an indicator of arterial-venous remodeling was quantified.

Ex vivo chorioallantoic culture, tissue processing, and analysis

Mesometrial halves of decidua containing pre-attachment chorions from Raldh2 WT and KO embryos were cultured together with allantoic explants, as previously described,³³ and analyzed by confocal microscopy. For decidua/EPC/chorion and allantois explants, samples were positioned with the distal side containing the chorionic plate facing upward. A single allantois was placed on top of the chorion at an approximately 45° angle in 500 μL of culture medium in 24-well tissue culture plates (Figure 4G). The association between the allantois and chorion was confirmed under a dissecting microscope prior to culture. The culture medium consisted of RPMI 1640 supplemented with 50% FBS, 4 μM L-glutamine, and 50 μg/mL penicillin/streptomycin. To inhibit the SEMA3–PLEXIND1 signaling pathway,³⁴ 25 μM Furin Inhibitor (EMD Millipore 344930-1 MG) or DMSO (vehicle control) was added to the medium. Explants were cultured for 24 h at 37°C in a humidified incubator with 5% CO₂. Following culture, explants were fixed in 4% paraformaldehyde (PFA) in 1× PBS overnight at 4 °C, embedded in 2% agarose, and sectioned at 150 μm thickness using a vibratome (Leica). The central section of each explant was stained for endothelial cells (ERG) and trophoblasts (Isolectin BSI), using the same staining conditions as those applied for placental cryosections. Vibratome sections were imaged using a Leica SP8 confocal microscope at 20× magnification. Z-stacks were collected to capture the chorioallantoic interface. Image analysis was performed in ImageJ. Vascular branching was quantified by calculating the ratio of ERG⁺ EC number over Isolectin BSI⁺ trophoblast cells, providing a measure of branching extent across experimental conditions.

Placenta collection and immunostaining

Whole mount

Placentas were carefully dissected from E9.5 embryos, maternal decidua was separated from the fetal labyrinthine side and discarded. Fetal placentas were fixed in 4% PFA for 2 h at 4°C and washed in 1X PBS (Cat# 10010023; Thermo Fisher Scientific, Waltham, MA). For immunostaining, whole placentas were permeabilized in 0.05% Triton X-100 (Cat #9002-93-1; Sigma-Aldrich, St. Louis, MO) in 1X PBS, then blocked for 1 h at room temperature in 1X PBS containing 0.01% Triton X-100 and 10% BSA (Cat# 126575-100 GM; Sigma-Aldrich, St. Louis, MO). Primary antibodies were incubated overnight in the blocking solution. Placentas were washed three times in 1X PBS +0.01% Triton X-100, then incubated with secondary antibodies in blocking solution overnight. After three additional washes in 1X PBS+0.01% Triton X-100, placentas were mounted with glass coverslips using Fluoromount-G (Cat# 0100-01; SouthernBiotech, Birmingham, AL).

Cross sections

Placentas were fixed in 4% PFA for 2 h at 4°C, washed in 1X PBS (Cat# 10010023; Thermo Fisher Scientific, Waltham, MA), and cryopreserved in a 30% sucrose/OCT (Cat# S53; Thermo Fisher Scientific, Waltham, MA) solution before being embedded in OCT blocks. Cryosections were prepared at 10 μm thickness and mounted onto slides (Cat# 22-037-246; Thermo Fisher Scientific, Waltham, MA). For immunostaining, sections were permeabilized in 0.05% Triton X-100 (Cat #9002-93-1; Sigma-Aldrich, St. Louis, MO) in 1X PBS, then blocked for 1 h at room temperature in 1X PBS containing 0.01% Triton X-100 and 10% BSA (Cat# 126575-100 GM; Sigma-Aldrich, St. Louis, MO). Primary antibodies were incubated overnight in the blocking solution. Antibodies used: Rat monoclonal anti-Laminin (Cat# MAB1914P; Sigma-Aldrich, St. Louis, MO; conc. 1:100), FITC-lectin from Bandeiraea simplicifolia (Griffonia simplicifolia) (Cat# L9381; Sigma-Aldrich, St. Louis, MO; conc. 1:50 from a stock solution 1 mg/ml), ERG (same as above). Sections were washed three times in 1X PBS +0.01% Triton X-100, then incubated with secondary antibodies in blocking solution for 1 h. After three additional washes in 1X PBS +0.01% Triton X-100, slides were mounted with glass coverslips using Fluoromount-G (Cat# 0100-01; SouthernBiotech, Birmingham, AL).

Quantifications

All quantifications of tissue immunostaining were performed on at least three independent replicate experiments including at least three separate litters. Samples were imaged using fluorescence confocal microscopy. Quantification of vessel diameters, cell number, fluorescence intensity and branching distance in the placental and allantois vasculature were performed with Image J software.

Sample preparation for single cell RNA sequencing (scRNAseq)

Single cell RNA sequencing was performed on E9.5 Raldh2^+/+ (WT), and Raldh2^−/− (KO) mouse placentas (n = 2). The rescue experiment included feeding the pregnant dam with all trans retinoic acid (ATRA) (1 μM) (CAS #302-79-4; Sigma-Aldrich, St. Louis, MO) in normal chow food starting at day E7.5, and collecting placentas at E9.5 (WT + ATRA, and KO + ATRA; n = 1). Placentas were collected and enzymatically digested into a single cell suspension with 1 mL of 10% Collagenase type I from Clostridium histolyticum (Sigma Cat# C0130-500 MG) with agitation at 37°C for 1 h. Cells were passed through a 70 μm cell strainer and rinsed with 1 mL ice-cold HBSS. Placentas were passed through a 70 μm cell strainer into 1X red cell lysis buffer (1.5 mM ammonium chloride, 100 μM sodium bicarbonate, and 11 μM EDTA disodium) and incubated for 10 min at room temperature. The sample was depleted of CD45⁺ cells and enriched for CD31⁺ cells to increase the number of ECs using microbeads (Cat# 130-097-418, #130-052-301; Miltenyi Biotec, Germany). 5% of all other cells were spiked back to the final sample to explore cell-cell interactions.

Quantification and statistical analysis

scRNAseq data collection: Library preparation and sequencing

Samples were processed for scRNAseq using standard 10X Genomics methods.⁴⁷ UVA Genome Analysis and Technology Core (RRID:SCR_018883, 5P30CA044579) at University of Virginia, School of Medicine University of Virginia performed the library construction and RNA sequencing. Single cell suspensions were loaded into a multi-channel microfluidic chip within the Chromium Controller (10x Genomics Inc., Pleasanton, CA, USA) to generate Gel Beads in Emulsion (GEMs), targeting a recovery of 5,000–10,000 cells. The 10x Genomics Chromium Single Cell System was used to process the suspension. Gene expression libraries were prepared with the Chromium™ Next GEM Single Cell GEM-X 3′ kit (1000691 Chromium™ GEM-X Single Cell 3′ Kit v4, 16 rxns), sequencing on an Illumina P2-100 kit (WT and KO samples) and P4-100 kit (WT + ATRA and KO + ATRA samples). A small portion of the library was initially sequenced on the MiSeq for quality control, including cell count verification, library assessment, and rapid cluster analysis. Final sequencing was performed on the Illumina NextSeq 2000 instrument. Bcl2fastq (Illumina) conversion software was used to generate sample-specific FASTQ files from the raw sequencing data. The resulting FASTQ files were processed with Cell Ranger (v8.0.1, 10x Genomics) tool for cell calling and to generate gene expression counts per cell. For this, reads were mapped to the M33 mouse reference transcriptome using the Cell Ranger count function, yielding a total of 101126 cells across all samples (48302 in WT, 14710 in KO, 19711 in KO + ATRA, and 18,403 in WT + ATRA). We had ∼28,000 mean reads/cell and detected ∼25,000 total genes, with a median of ∼2,000 genes/cell across all samples. The sequencing data supporting these findings are available at GEO.

scRNAseq data processing

The downstream analysis of the scRNAseq data was performed in R, using Seurat v4.4, that included standard preprocessing steps to visualize and filter cells based on quality control metrics. Cells with high mitochondrial gene expression and low or very high feature counts were filtered out. SoupX was used to correct for any contamination due to ambient RNA (lysed RBCs in this case) and doublets were filtered out following their identification with the scDblFinder package in R. To mitigate the effects of cell cycle heterogeneity in the single cell sequencing data, the difference between the G2M and S phase scores was regressed out from the data. The data was normalized using ‘sctransform’ and data across all experiments was integrated to anchor shared cell populations. Principal Components were estimated based on a diagnostic plot (elbow plot) to capture the major structure in the data, which were then used to determine UMAP and PHATE dimensionality reductions. Following cluster identification, markers were identified for each cluster using the “FindAllMarkers” function in Seurat. An additional filtering step was performed at this point to eliminate cells (possibly dying cells) that showed high expression of Mt-Co3 gene (>5 log normalized expression) as Mt genes showed up as markers in two identified clusters. Re-clustering was performed for the final dataset of 75,796 cells and cell types were determined based on known markers from published literature and genes identified with marker analysis.^27,28 The arterial and venous module score graphs were generated by sampling an equal number of cells across experimental conditions and then scoring the cells for the expression of the venous/arterial gene set (module scores). The means of the venous/arterial module scores for each experiment were compared.

Pathway enrichment analysis

Enrichment analysis was performed using the Enrichr⁴⁸ functionality in R. The input data used for this analysis were the differentially expressed genes in the WT vs. KO Fetal ECs which was obtained by marker analysis using Seurat’s “FindMarkers” functionality. The enrichment results shown here are for the signaling pathways from the Reactome pathways database for an adjusted P-value cut-off of less than or equal to 0.05 (adjusted P-value <0.05).

Heatmaps generation

Heatmaps were generated with the “pheatmap” function in R. The heatmap input gene data was either from differential expression results for the WT & KO Fetal ECs (e.g., Notch pathway genes) or the aggregated expression of genes for each experimental condition across all cell types for the Sema-Plexin-Vegf genes.

Intercellular communication analyses using cell chat

The Cell Chat tool, quantitatively infers intercellular communication networks from single cell data and uses a comprehensive curated database of ligand-receptor interactions to achieve this. We used the Cell Chat (v2.1.2) functionality in R to analyze the cell-cell communication among the identified placenta cell types using default parameters. Each experimental condition was analyzed individually. We computed the communication probabilities, excluding interactions when specific cell groups contained fewer than 10 cells. We extracted cellular communication network at a signaling pathway level, for the annotated Seurat subset objects of WT, KO & KO ATRA. All three cellchat objects were subsequently merged for a comparative analysis. The Cell Chat DB mouse was used and differential interactions and their strengths among various cell types, including comparisons between WT, KO, KO + ATRA groups, were systematically analyzed. Circle plots were generated to show differential interaction strength of signaling pathways among cell populations across all three datasets. This analysis identified dysregulation of signaling pathways, along with associated upregulated and downregulated ligand-receptor pairs.

Statistical analysis

All statistical details of experiments, including the exact value of “n” and what “n” represents (e.g., number of embryos, number of litters, or independent replicates), are reported in the figure legends. For all embryo experiments, a minimum of n = 3 to n = 8 biological replicates from at least three separate litters were used. Statistical significance was determined using GraphPad Prism (v10.0). Continuous data are presented as mean ± standard error of the mean (SEM). Comparisons between two groups were performed using Student’s t test (two-tailed), and multiple group comparisons were performed using one-way ANOVA followed by Tukey’s multiple comparisons test. A P-value of <0.05 was considered statistically significant.

Published: March 3, 2026