Feto-placental blood vessel development Open Access

Formation of the placental circulation is crucial for fetal survival and a successful pregnancy. In both humans and mice, the placental circulation is made up of two separate blood supplies, one maternal and one fetal. Maternal nutrients and gases must cross the placental barrier to be delivered to the fetal circulation. While the maternal side has large spiral arteries carrying maternal blood to the sinusoids of the placenta, on the fetal side, the feto–placental blood vessels transport nutrients and gases from the maternal circulation to the fetus. Despite the important role these placental vessels play during development, our understanding of their developmental origin, including the progenitor cells that give rise to them and the genetic pathways that control their formation, is decades behind that of the embryonic vasculature in both mice and humans. Poor adoption of placental vasculature research by embryonic vascular biologists and clinicians is likely due to several reasons. Firstly, the placental circulation is counter-intuitive, in that the placental arteries carry deoxygenated blood whereas the embryonic arteries carry oxygenated blood. Secondly, until recently, we have lacked genetic markers and genetic tools to compare placental endothelial cells (ECs) to embryonic ECs in mouse models. Finally, in humans, we are similarly limited by access to early post-implantation stage tissue, technology and resolution of imaging modalities to track their formation during pregnancy. Given these caveats in diagnostic imaging, the prediction and treatment of vascular-associated disorders are lacking. Furthermore, formation of these placental vessels in the mouse occurs in a series of stages for embryonic viability; however, such checkpoints in human development are yet to be investigated.

The placenta contains cells from three different cell lineages: maternal uterine cells making up the decidua (uterine lining), trophoblast subtypes derived from the trophectoderm (outermost cellular layer of the early embryo) and extra-embryonic mesoderm (originating from the epiblast; see Glossary, Box 1), which contributes to the feto–placental vasculature (see Glossary, Box 1). The vessels of the placenta are located within structures known as the ‘labyrinth’ in the mouse and ‘villi’ in humans (see Glossary, Box 1). Materno–fetal exchange occurs by embryonic day (E)10.5 in mice and by ∼10 post-conception weeks (pcw; see Glossary, Box 1) in humans, based on experiments examining maternal spiral arterial blood flow into the intervillous space and increased oxygen tension (Jauniaux et al., 2000) (Fig. 1).

In the mouse, the allantois (see Glossary, Box 1) gives rise to the umbilical artery and vein, in addition to the placental vessels that connect to them, to create the ‘chorio–allantoic’ vasculature (see Glossary Box 1) (Fig. 2). Before this vasculature becomes functional at E10.5, the ‘chorio–vitelline’ vasculature of the yolk sac provides initial nutrients and gas exchange. The yolk-sac vasculature envelops the embryo in the mouse, whereas in the human it is a sac adjacent to the connecting stalk (see Glossary, Box 1). In humans, the allantois (or ‘allantoic diverticulum’; see Glossary, Box 1) is a separate structure from the umbilical cord (see Glossary, Box 1) vessels and the ‘feto–placental’ vessels, which both derive from the connecting stalk (Fig. 3). Here, the human allantois forms a singular extension from the hindgut, which exists temporarily between the two umbilical arteries. It regresses from the umbilical cord to connect the hindgut to the cloaca and eventually becomes the ‘urachus’ or ‘umbilical ligament’ (for more information see O'Rahilly and Müller, 1987, Kruepunga et al., 2018, Kalisch-Smith, 2025). In this Review, I use both ‘chorio–allantoic’ and ‘feto–placental’ vasculature terms to describe the equivalent structures between mouse and humans.

Following the onset of oxygenated blood circulation in the embryo after materno–fetal exchange at E10.5 (mouse) and 10 weeks (humans), deoxygenated blood flows from the dorsal aorta to the umbilical artery. As the umbilical artery moves into the placenta, it is continuous with the large placental stem arteries radiating outwards, creating an arterial tree-like structure (Fig. 2, E11.5; Fig. 3, 4+ pcw). These arteries branch further into an expansive capillary plexus for nutrient/gas exchange over the placental barrier from the maternal blood supply (supplied by spiral arteries and returned by venous channels). Fetal capillaries then pass oxygenated blood into large venous vessels, which connect to the umbilical vein, flowing on to the embryonic heart for distribution (Figs 2 and 3). ECs line all the vasculature of the embryo, yolk sac, umbilical and chorio–allantoic placenta, providing vascular structure.

This Review highlights current understanding of feto–placental vascular development, focusing on arterio–venous formation in mice and humans, and illustrates known models exhibiting perturbed formation. I detail major unanswered questions in the placental vascular field, of EC developmental lineage, when arterio–venous formation occurs, and their relevance in human pathologies of pregnancy, including miscarriage, preeclampsia, congenital heart disease and stillbirth. With improvements in genomic technologies, next generation sequencing, as well as organoids and stem cell-based embryo models, solutions to some of these age-old problems are tantalisingly close.

The mouse has been used as a robust model to test genetic and environmental perturbations of pregnancy. While the mouse placental vasculature is smaller and less branched in structure compared to humans, there are many analogous cell types that perform similar functions. Setting up the materno–fetal interface of the placental barrier, separating maternal and fetal blood supplies, is crucial (Simmons et al., 2008). In mice, this barrier is composed of allantois-derived ECs, two syncytiotrophoblast layers (called SynT-I and SynT-II) and a perforated layer of sinusoidal trophoblast giant cells that line the maternal blood sinusoids of the placenta (Simmons et al., 2008).

The chorio–allantoic blood vessels develop primarily from extra-embryonic mesoderm (Tremblay et al., 2001; Inman and Downs, 2007; Nehme et al., 2025), the differentiation of which occurs over a number of stages before distal attachment to the chorion by E8.5 (Figs 1 and 2) (reviewed by Inman and Downs, 2007; Arora et al., 2012). While incremental steps have been made to understand the formation of the extra-embryonic vasculature, a clear junction between embryonic and allantoic-derived vessels is yet to be defined. For genetic mutations that affect allantois development and differentiation, from allantois budding to chorio–allantoic branching, see Table 1. For gene markers of vascular and vascular-associated cells in mice and humans see Table 2.

The extra-embryonic mesoderm, which expresses T (brachyury) and Fgf8, extrudes a bud-like structure (now referred to as allantoic mesoderm) into the exocoelomic cavity towards the chorion (see Glossary, Box 1) (Ellington, 1985; Inman and Downs, 2007; Nahaboo et al., 2022). The allantoic mesoderm extrudes via proliferation and ongoing deposition of mesoderm into the allantois from the primitive streak (Downs and Bertler, 2000; Tam and Beddington, 1987). Mouse genetic knockouts of genes that impact mesodermal proliferation (e.g. T, Bmp4, etc. see Table 1) prevent allantois budding and/or growth (extension). These knockouts are embryonic lethal due to chorio–allantoic attachment defects, with either a complete absence of or poor placental vascular formation (Lawson et al., 1999; Inman and Downs, 2007). Before allantoic budding, the extra-embryonic mesoderm also gives rise to a thin mesothelial layer that lines the exocoelomic cavity (including the chorionic region), sometimes referred to as ‘chorionic mesothelium’ or ‘chorionic mesenchyme’ (see Glossary, Box 1; Pereira et al. 2011; Nahaboo et al., 2022; Paulo et al., 2011), which has hematopoietic potential (Zeigler et al., 2006).

In order for the allantois to expand and extend toward the chorion, the distal portion of the allantois increases dramatically in extracellular space by cavitation, giving the allantois a foamy-like appearance (Downs and Bertler, 2000). During this time, the allantois accumulates extracellular matrix proteins and proteoglycans including hyaluronic acid. Both groups have protein members that mediate EC tube formation, which may impact on formation of the allantois vascular plexus (see Elongation) (Arora et al., 2012; Crosby et al., 2005). Defects in cavitation have also been associated with aberrant cellular adhesion, another important process for EC tube formation (Naiche and Papaioannou, 2003).

By four-somite pairs, the allantois has elongated and is now covered by a mesothelial layer of squamous epithelial cells (Downs et al.,1998). The mesothelium (Car4⁺, Cdh11⁺ and Stard8⁺), also known as the allantois cortex, is located at the distal region of the allantois. The mesothelium begins to express adhesive properties (e.g. VCAM1) that permits future chorio–allantoic attachment (Kwee et al., 1995; Nahaboo et al., 2022). By E8.25, a subpopulation of allantoic mesoderm cells differentiates into Etv2⁺ angioblasts (Lee et al., 2008), which signal and induce a signalling cascade within Kdr⁺ (Flk1⁺) endothelial progenitors (Downs et al., 1998; Huber et al., 2004) (Table 1). The endothelial progenitors in the allantois form a primitive Pecam1⁺ endothelial network at ∼E8.25 by vasculogenesis, a process by which new blood vessels are formed (Downs et al., 1998). At this time, in the chorion, the labyrinth trophoblast progenitors (precursors to SynT cells) become patterned and express markers for all trophoblast layers (e.g. Gcm1, Syna, Hand1, Rhox4b) (Simmons et al., 2008, reviewed by Simmons, 2013).

At this stage, the allantoic mesothelium attaches to the chorion (Downs and Gardner, 1995). What happens to the mesothelial layers (covering the allantois and early chorion) after this stage is unknown. Following chorio–allantoic attachment, Gcm1⁺ chorionic trophoblast cells (early SynT-II clusters) mark future branch points for the allantois to extend into the chorion (Anson-Cartwright, et al., 2000; Simmons et al., 2008). Kdr⁺ EC progenitors and definitive ECs are already differentiated before chorio–allantoic attachment, which further expand into the chorion. By E8.5, part of the primary vascular plexus in the allantois also remodels into the umbilical artery (Downs et al., 1998).

In order for allantois-derived ECs to be incorporated into the trilaminar labyrinth, these ECs must first invade into the chorion. This process coincides with embryo turning, where the body axis rotates or ‘inverts’ from a concave to convex ‘C’ shaped position. It is hypothesised that cell shape changes and the movement of Gcm1⁺ chorion cells pass through the chorion, causing it to fold, creating simple branches in which the ECs reside (Simmons et al., 2008). Continued EC branching morphogenesis allows creation of the labyrinth vasculature to support the nutritional needs of the embryos for rapid growth.

The umbilical vein is starting to form in the embryo and is connected to the allantois by the 16-somite stage (Walls et al., 2008). How the umbilical vein forms and the progenitors that contribute to its development is currently unknown. Allantoic-derived ECs continue to invade and branch to create the labyrinth vasculature, making a connected arterial and a venous vascular tree (Fig. 2). Many mouse genetic mutations that affect labyrinth EC formation have been investigated and are embryonic lethal at different stages of gestation from E9.5 (Table 1).

In addition to the allantois cell types mentioned above, transcriptomic analysis of the allantois has revealed further cell subpopulations at E7.75 (Nahaboo et al., 2022), including Cdx4⁺ cells in the allantois base (closest to primitive streak), and Cdh5⁺/Sox7⁺/Runx1⁺ erythroid progenitors (Nahaboo et al., 2022). A small population of Dll4⁺ cells have also been found at this age, indicating the presence of early arterial/sprouting ECs (Nahaboo et al., 2022). Compared with other EC populations at E8.25, allantoic EC progenitors express a unique signature of posterior-associated HOXA genes: Hoxa10, Hoxa11 and Hoxa13 (Ibarra-Soria et al., 2018; Scotti and Kmita, 2012). Hoxa13 has since been used to engineer mouse Cre/CreERT2 tools to target allantois and placental ECs (discussed below). Further allantois heterogeneity has been shown in a mouse embryonic gastrulation dataset at E8.5, where Tbx4⁺ allantois cells are distinct from allantoic EC progenitors (Pijuan-Sala et al., 2019). Lineage tracing of Tbx4:Cre:lacZ allantoic mesenchymal cells shows that they do not give rise to umbilical ECs but do give rise to support cells surrounding umbilical and placental labyrinth vessels, likely pericytes and smooth muscle (Naiche et al., 2011). Only sporadic placental ECs show labelling at E10.5, although this has not been rigorously quantified.

In humans, the feto–placental vasculature originates from the connecting stalk and attaches to the chorion from 2 weeks gestation (Figs 1 and 3). Connecting stalk cells extend across the chorionic cavity to make up the feto–placental vasculature and the supporting chorionic plate (see Glossary, Box 1), which contains the major collecting vessels connecting to the umbilical artery and vein (Kalisch-Smith, 2025). In humans, these feto–placental vessels are arranged into a tree-like structure to permit nutrient exchange, and are contained within ‘villi’ – protrusions of trophoblast from the chorionic plate (Fig. 3). The barrier between fetal and maternal blood is composed of two layers of trophoblast cells (syncytiotrophoblast cells in contact with maternal blood and villous cytotrophoblast cells), as well as a layer of ECs. In this way, the human differs from the mouse with the types of cellular layers between maternal and fetal blood supplies (the mouse has two syncytiotrophoblast layers), but the individual cell types (or layers) are functionally similar (Figs 2 and 3). Villous cytotrophoblast cells extend into columns, differentiate and form invasive extravillous trophoblasts, which anchor the villi to the decidua.

The developmental lineage of the connecting stalk and feto–placental vasculature in humans has been unclear, with hypoblast, primitive endoderm, yolk sac and extra-embryonic mesoderm all suggested as potential progenitors. However, recent production and integration of single-cell RNA sequencing (scRNA-seq) of human embryos (Tyser, et al., 2021; Zhao et al., 2025), together with stem cell-based models of post-implantation embryos (Oldak et al., 2023), has revealed new insights. In stem cell-based embryo models created from human embryonic stem cells, cells with an extra-embryonic mesoderm signature (e.g. VIM⁺, POSTN⁺, HAND2⁺, TBX4⁺, HGF⁺), form a connecting stalk towards trophoblast cells across a chorionic cavity at day 8 of culture, equivalent to day 13/14 pcw (Oldak et al., 2023). This extra-embryonic mesoderm signature is similar to that from a human Carnegie stage (CS) 7 embryo (2.5 pcw), expressing FOXF1, POSTN, VIM and PITX2 (Tyser et al., 2021; Zhao et al., 2021 preprint; Oldak et al., 2023) and is the most promising data we currently have in the field as to developmental lineage. While models are yet to show umbilical formation, EC differentiation and chorionic attachment, these will be required in the next iteration of embryo models to ensure developmental competence for later embryonic stages. Further, the genetic signature of extra-embryonic mesoderm in humans mimics non-human primate (cynomolgus macaque) connecting stalk cells (Zhao et al., 2025), in which extra-embryonic mesoderm contributes to placental villous ECs (Jiang et al., 2023).

Endothelial progenitors have been predicted within the connecting stalk from 15-21 days post-conception (from CS5c-6) (Boyd and Hamilton, 1970; Kaufmann et al., 2004; https://www.ehd.org/virtual-human-embryo/), but are yet to be recapitulated in embryo models to determine expression profiles, 3D orientation or exact timing. Culture conditions are likely to influence how these early embryos grow and are yet to fully recapitulate the uterine environment. In the early human placenta, endothelial progenitors have been assessed at 5-9 pcw (CS16-20) and show primitive ‘cords’ located in the chorionic plate and central villus, but not the peripheral villi in which the later feto–placental capillaries develop (Aplin et al., 2015); here, the large stem vessels express markers for EC progenitors (e.g. KDR⁺) (Aplin et al., 2015). Vasculogenesis, likely of angioblasts (see Glossary, Box 1), gives rise to these endothelial progenitors of the chorionic plate, while the capillaries arise by angiogenesis of pre-existing blood vessels from 4 pcw (see Kaufmann et al., 2004). Vasculogenesis has also been hypothesised to occur at terminal capillaries from surrounding stromal mesenchyme (Aplin et al., 2015). However, this is yet to be established in human organoid models, which contain stromal mesenchyme but typically lack endothelial cells due to low abundance (compared with trophoblast and stromal cells) and culture techniques that were developed primarily for isolation of trophoblast cells (Turco et al., 2018). Similar experiments could perhaps give insight into ‘vascular pruning’ whereby apoptosis is seen at higher levels in distal capillaries at 7-8 pcw (Tertemiz, et al., 2005). Human organoid and embryo models are likely to yield many insights into early human placental villus formation in the near future.

Even by 4.5 pcw, the human placental vasculature is remarkably complex. Adjacent cytotrophoblast cells produce angiocrine factors, including VEGF, PGF and angiopoetins (Lash et al., 2010), to facilitate vasculogenesis and angiogenesis. When these factors are disrupted, blood vessel formation can be dysregulated and lead to pathological pregnancies (reviewed by Burton et al., 2009). Mesenchymal-derived mural cells are also incorporated into vessels from 4 pcw (Kaufmann et al., 2004). Mesenchymal ‘stromal’ cells, which exist throughout human placental development and maturation (see Glossary, Box 1), consist of mural subtypes including fibroblasts, pericytes and macrophages/Hofbauer cells (Robin et al., 2009; Crisan, et al., 2008), to form the ‘stromal core’ of villi. It is likely that these villous stromal cells are also derived from a common extra-embryonic mesoderm progenitor, before differentiation into a mesenchymal lineage, which is the case in mice and macaques.

As mentioned above, a subpopulation of allantoic mesoderm cells express Etv2 at E8.25, a marker of angioblasts (Lee et al., 2008). Angioblasts are important in vascular development because they give rise to both endothelial and hematopoietic progenitors and, by doing so, populate many EC populations de novo in the embryo (Lee et al., 2008). When Etv2 is deleted from the embryo, including the allantois, severe blood and vascular defects result (Lee et al., 2008). Specifically, in the chorio–allantoic placenta, Etv2-null embryos show an absence of blood vessels and ECs at E9.5 (Lee et al., 2008), indicating that resident allantoic angioblasts are the precursors to umbilical and placental ECs. Furthermore, the expression of Etv2 and activation of its downstream signalling cascades are important in establishing whether cells differentiate into ECs versus hemogenic lineages. In an experiment creating a range of Etv2 expression profiles in embryos (combining deletion of an Etv2 enhancer and crossing it with Etv2 null or wild-type allele), Sinha and colleagues showed that reducing Etv2 expression to 20% led to normal EC formation but perturbed erythropoiesis via a reduction in Tal1 expression (Sinha et al., 2022), a factor with an important role in hematopoietic differentiation (Pijuan-Sala et al., 2019). Using in silico analysis, Tal1 signalling was also predicted to be important in allantois mesoderm allocation (Kamimoto et al., 2023). Here, authors designed a genome regulatory network programme called CellOracle, which allows in silico perturbation of cell lineage-specifying transcription factors. Kamimoto and colleagues predicted that Tal1 knockout increases allantoic mesoderm at the expense of hemato–endothelial progenitors (Kamimoto et al., 2023). Overproduction of allantoic mesoderm is also seen in another mutant for Eed, a component of the PRC2 complex (Faust et al., 1995; Grosswendt, et al., 2020). Altogether, this suggests that there is fine genetic control of cell allocation to the allantoic mesoderm and hematopoietic lineages. Use of CellOracle (Kamimoto et al., 2023) to predict genome regulatory networks and allantois phenotypes could yield further interesting insights into important transcription factors at play during chorio–allantoic formation.

Profiling of human endothelial cells in scRNA-seq datasets shows that the embryonic dorsal aorta is already differentiated by CS13 (4 pcw) (Xu et al., 2023), while the fetal heart has arterial and venous populations by 13/14 pcw (McCracken et al., 2022). Placenta arterio–venous differentiation is likely to be matched with that occurring in the embryo and is, therefore, likely to occur much earlier than currently thought. Although not necessarily a surrogate of placental EC formation, the umbilical cord starts to form arterial vessels before CS9 and venous vessels from CS11 (∼3.5 pcw) (Kalisch-Smith, 2025), which is earlier than previously specified (O'Rahilly and Müller, 1987). In light of the new computational lineage trajectories discussed above, and umbilical differentiation being earlier than anticipated, it would appear more likely that placental and umbilical ECs are developing concurrently from connecting stalk EC progenitors, rather than developing independently before connecting to each other, as previously thought. Additional scRNA-seq datasets available from the first trimester human placenta (Suryawanshi et al., 2018; Vento-Tormo et al., 2018) will hopefully shed more light on early placental vascularisation in combination with synthetic embryo models.

Key to the continued formation of the placental vasculature in both mice and humans, is communication of ECs with the adjacent chorionic trophoblast. Although in other vascular beds, secreted pro-angiogenic signalling (e.g. VEGFA) comes from cardiomyocytes (heart), chondrocytes (bone), podocytes (kidney), neural tube and macrophages (Coultas et al., 2005), in the placenta, it is the chorionic trophoblast that provides these signals. In the mouse, the labyrinth trophoblasts progenitors and later SynT-II cells produce VEGF to initiate sprouting angiogenesis (EC production from existing blood vessels) via its receptor KDR (VEGFR2; Nahaboo et al., 2022; Marsh and Blelloch, 2020; Simons et al., 2016). Trophoblast paracrine signalling also has an important additional role: to match the development of the labyrinth vasculature with growth and remodelling of the maternal side of the placenta. Experiments in mice show that poor expansion of the spiral arteries and maternal blood spaces (sinusoids) can lead to reduced expansion of the fetal capillaries and impact the transport of nutrients to the fetus, ultimately causing fetal growth restriction (FGR) (Hamada et al., 2007; Kalisch-Smith et al., 2019). Furthermore, in humans it is well established that aberrant angiocrine signals from trophoblast cells can also lead to preeclampsia, a condition defined by poor spiral artery remodelling, hypertension at >20 weeks gestation and maternal endothelial dysfunction (reviewed by Dimitriadis et al., 2023), which also impacts the feto–placental vasculature (discussed below).

In mice, labyrinth trophoblast progenitors also contain corresponding ligands or receptors to those derived from allantoic ECs. For example, the allantois expresses VCAM1, whereas the chorion expresses its receptor integrin α4, thus permitting chorio–allantoic attachment (Kwee et al., 1995). Similarly, other receptor–ligand interactions are involved, including APELA (ELABELA)–APLNR required for angiogenesis (Freyer, et al., 2017; Ho, et al., 2017) and WNT–FZD (Nahaboo, et al., 2022; Parr et al., 2001). WNT signalling is required for SynT-II differentiation (Zhu et al., 2017), with allantois-derived Wnt2 deletion reducing labyrinth formation (Monkley, et al., 1996), and deletion of chorion-derived Wnt7b showing poor chorio–allantoic attachment (Parr et al., 2001). WNT–FZD5 signalling is also important for upregulation of Vegf expression in chorion trophoblast (Lu et al., 2013), forming a positive feedback loop. Recent discoveries from scRNA-seq of E9.5-E14.5 placental labyrinths has revealed known and previously unreported interactors between the chorion/labyrinth trophoblast/SynT-II with allantois ECs (Marsh and Blelloch, 2020). These include known interactors, such as VEGFA–KDR and VEGFA–FLT1, and previously unreported interactions to play a role in placental vascularisation, such as LGR5–RSPO3. RSPO3 is part of the WNT/β-catenin signalling pathway and its deletion is embryonic lethal due to defective chorio–allantoic attachment (Aoki et al., 2007). With the generation of these new scRNA-seq datasets from placental labyrinth development (Marsh and Blelloch, 2020; Liang et al., 2021), undoubtedly, more receptor–ligand interactions will be uncovered, which can be further tested in genetic mouse models. A challenge in the placental field in determining new interactions will be the reliance on available receptor–ligand libraries, which are curated based on known interactions in embryonic and adult contexts. Given the unique cell types within the placenta, creation of placental proteomic and protein–protein interaction datasets would be a good start to curate placental-specific libraries.

A major checkpoint in placental development is at the chorio–allantoic attachment phase: if not achieved this leads to embryonic lethality (Copp, 1995). Indeed, many mouse mutants exhibit poor/absent allantoic vascular formation, including defects in attachment upon perturbation of Tbx4 (Naiche and Papaioannou, 2003), Fgfr2 (Xu et al., 1998) and Mtrr (Wilkinson et al., 2021), as well as errors in initial sprouting of ECs into the chorion from mice null for Grb2 (Saxton et al., 2001), Fzd5 (Ishikawa et al., 2001; Lu et al., 2013), Notch1 (Krebs et al., 2000; Limbourg, et al., 2005), Hey1/Hey2 (Fischer et al., 2004) and Rbpj (Lu et al., 2019). Intriguingly, the Mtrr hypomorph also showed eccentrically located placentas in 12% of embryos, where allantoic and umbilical insertion was not in the centre of the chorion (Wilkinson, et al., 2021). The allantois can also vary in the space it uses to attach to the chorion, with the knockout of Apela showing reduced distal attachment at E8.5 (Freyer, et al., 2017). A different study showed that knockout of the same gene caused poor feto–placental angiogenesis and a thin labyrinthine layer later at E10.5 (Ho, et al., 2017), demonstrating that defects during early stages can have long-lasting impacts.

Continued labyrinth development and vascular branching requires further communication between the chorion and allantoic-derived ECs. Chorion-expressed Gcm1 cell clusters mark future branch points for the allantois to extend into the chorion (Anson-Cartwright, et al., 2000; Simmons et al., 2008). Gcm1 expression appears before chorio–allantoic attachment but requires allantoic contact for expression to be maintained; for example, through WNT–FZD-GCM1 signalling (Hunter, et al., 1999; Lu et al., 2013). Further labyrinthine branching also requires activation of cMet (expressed on labyrinth trophoblast progenitors) by allantoic mesenchyme-derived HGF (Ueno et al., 2013; Uehara, et al., 1995). In a large mouse knockout screen of 103 mutants, ∼40% were lethal between E9.5 and E14.5, and 95% of these had severe placental phenotypes at E9.5 (Perez-Garcia, et al., 2018). These observations demonstrate how crucial the early stage of placental development is for embryo–placental competence in later gestation, and that without sufficient formation of these allantoic-derived vessels, this likely results in embryonic lethality. For further allantois vascular mutants impacting feto–placental vascular formation, see Table 1 (Arora et al., 2012; Woods et al., 2018).

In mice, embryonic lethality also results from poor pericyte and smooth muscle recruitment from allantois-derived cells. Pdgfb⁺/ASMA (ACTA2; alpha smooth muscle actin)⁺ pericytes surround ECs and provide necessary structural and metabolic support via the release of growth factors. When Pdgfb or Pdgfrb is deleted, pericyte differentiation is impaired and the labyrinth vascular organisation is disrupted from E13.5 (Ohlsson et al., 1999; Looman et al., 2007). ASMA⁺/MYH11⁺ (SM-MHC) smooth muscle cells surround arterial and arteriole ECs of the placental labyrinth and other chorionic plate cells by E15.5 (Navankasattusas et al., 2008; Kalisch-Smith et al., 2022). Smooth muscle cells are yet to be extensively investigated for cell autonomous impacts in genetic mouse models. When these mural cells arise, and from which progenitors, is yet to be investigated.

One pertinent question to be answered in humans is whether chorio–allantoic defects occur in humans as they occur in mice, and whether these are a cause of early miscarriages. In mice, these defects are a common cause of lethality and a checkpoint for successful development. Even poor initiation of vascular branching early in mouse gestation has far-reaching consequences for the remainder of development and embryo competence. A recent study of miscarriages occurring at 7-10 pcw showed delayed embryo development, with embryos being four Carnegie stages earlier than those in an ongoing pregnancy, as estimated by crown–rump length (Pietersma et al., 2023). Whether the human placenta is equally affected during this early stage, and perhaps the cause of these miscarriages, however, is unknown. With new embryo models available, investigation of chorio–allantoic attachment and assessment of gene knockouts in human development feels closer than ever and would be inconceivable without this technological advance.

Investigation of placental defects during the first trimester from implantation are often assumed to be due solely to poor trophoblast formation and function, with little or no insight into placental vascular formation. To begin to explore this topic, we have recently carried out a systematic review of endothelial genes (gene-associated single nucleotide polymorphisms) from genome-wide association studies, involved in pregnancy disorders including miscarriage, stillbirth and congenital heart defects (CHD) (Kalisch-Smith et al., 2024). As cardiac and placental ECs share many of the same gene programmes, we assessed their likelihood of either heart or placental ECs being the cause of these pathologies. We first found that ∼7% of EC-associated genes were associated with miscarriage, 4% for CHD and 3% for stillbirth (Kalisch-Smith et al., 2024). Roughly half were more likely to be attributed to the placental endothelial expression (versus heart) for both miscarriage and CHD, compared with only 30% for stillbirth, with 70% being more likely to be attributed to heart-expression of EC genes (Kalisch-Smith et al., 2024). Further investigation into placental EC genes is now of utmost importance to understand their expression profiles, and their contribution to placental development and disease.

Although it is clear that allantoic tissue differentiates into the Efnb2⁺ umbilical artery by E8.5 and the Ephb4⁺ umbilical vein by E9.5 (Wang et al., 1998) in mice, it is unclear how – and when – the feto–placental ECs are formed. Blood flow can be seen by E10.5 in the umbilical cord, which suggests that the arterio–venous circulation is already connected and functional. Placental arterio–venous differentiation is complete by E12.5, shown by the presence of mature markers of placental arteries [e.g. Gja5 (Cx40) and Dll4] and veins (e.g. EMCN and APLNR) (Kalisch-Smith et al., 2022). In addition, Vegfc and Aplnr (Apj) are expressed in a subset of vessels at E9.5/E10.5 (Outhwaite et al., 2019; Freyer et al., 2017). Despite the robust expression of many arterio–venous genes in placental ECs, their roles in the placental vasculature are rarely investigated. To date, only two studies have investigated a placental phenotype when knocking out an arterial gene (Dll4; Gale et al., 2004, Duarte et al., 2004), while no studies deleting venous-associated genes have investigated the placenta.

Dll4 is a ligand to NOTCH receptors and is expressed in both embryonic arterial and sprouting ECs (Chong, et al., 2011). Haploinsufficiency of Dll4 shows degenerating placental vessels, particularly in the large stem arteries, and embryonic lethality by E10.5 (Gale et al., 2004; Duarte et al., 2004). Considering that chorio–allantoic attachment also occurs from E8.5 (Fig. 1), we could predict that the placental arteries bud from the umbilical artery, while the placental veins bud from the umbilical vein. However, contradictory to this theory, in the embryo, arterial ECs are not proliferative and are formed by alternate mechanisms. Although proliferation is not seen in major embryonic arterial ECs (with high expression of Unc5b, Dll4 and Efnb2), higher proliferation is seen in arterioles and venous capillaries (Hou et al., 2022). In models of injured adult arteries, arterial ECs must de-differentiate before proliferation and re-assembly into arteries (Das et al., 2019). An alternative way to generate arterial ECs, used by the sinus venosus (venous progenitor in the heart), is through the proliferation of venous ECs, which then differentiate into the coronary arteries (Red-horse et al., 2010; Su et al., 2018). Whether specific intermediates are required for placental EC development from the umbilical cord vessels is now of great interest. Furthermore, the cellular origin and exact timing of umbilical vein development is yet to be fully established.

A major disadvantage in the placental field is that scRNA-seq datasets are created on placental tissue in isolation, without incorporation with other embryonic/extra-embryonic tissues. Without these, it is very difficult to integrate datasets together without overlapping cell populations. To understand whether the placental vasculature is unique, for example compared with embryonic or yolk sac vascular beds, additional datasets are required. Currently, the transcriptional pathways used to create the placental vasculature, both de novo from mesoderm (i.e. vasculogenesis) as well as angiogenic growth of existing blood vessels, are similar to systems in both the embryo and yolk sac. These conserved factors and pathways for formation of mesoderm (e.g. T and Bmp4), differentiation to EC progenitors (e.g. Etv2 and Kdr) and angiogenesis (e.g. Tie2, Cdh5, Vegf, etc.) show that the placenta is equally affected by these processes (Table 1). That said, subtle differences between ECs of the placenta versus other organs are starting to emerge. For example, a study in human tissue from around 10-18 pcw has shown placental ECs do have a unique transcriptional profile compared with other organ vascular beds (Cao et al. 2020). Here, placental ECs express LHX6 (encoding for a unique cysteine-rich zinc-binding domain), LIN28B (encoding a protein that regulates mRNA translation and miRNA maturation) and MEOX2 (a transcription factor). These genes are yet to be investigated in detail and may provide the placenta with additional adaptability in adverse environments to control vessel formation or function. Subtle changes between embryonic and placental NRP1 and NRP2 expression have also been found (Kalisch-Smith et al., 2022). While embryonic NRP1 is expressed in arterial ECs (Chong et al., 2011; Erskine et al., 2017), placental NRP1 is expressed in arterial ECs at E12.5 but not E15.5 (Kalisch-Smith et al., 2022). Similarly, while embryonic NRP2 is expressed in venous ECs (Chong et al., 2011), in the placenta NRP2 is only expressed in a small subset of ECs (Kalisch-Smith et al., 2022). Additional transcriptional differences are likely, given that the placental arterio–venous system carries blood with different oxygenation to the embryo (e.g. placental arteries have deoxygenated blood, embryonic arteries have oxygenated blood). Furthermore, given that placental ECs are directly adjacent to trophoblast cells, ECs could adopt a different profile to signal with these trophoblast cells, either by direct contact or a paracrine manner. While mouse studies exploring placental endothelial heterogeneity versus other embryonic endothelial beds are yet to be performed or integrated with available datasets, an organotypic transcriptional signature is likely given the above outcomes and the high degrees of endothelial heterogeneity found in early mouse embryos (Hou et al., 2022) and from other adult organs (Wang et al., 2022).

It is well established in mice that both placental arterial and venous networks can change in response to genetic and environmental perturbations. The arterial tree is reduced in size by knockout of Unc5b (binds Flrt2 described later; Navankasattusas et al., 2008) or is enlarged by hypomorphic Gcm1 (Bainbridge et al., 2012), exposure to cigarette smoke (polycyclic aromatic hydrocarbons) (Detmar et al., 2008) and lower diversity of the maternal microbiome (Pronovost et al., 2023). Conversely, knockout of Cited2 reduces the expansion of both arterial and venous trees (Withington et al., 2006). The arterial circulation is also impacted in mouse embryos with CHD caused, by example, from maternal iron deficiency (Kalisch-Smith et al., 2021, 2022). If arteriole development is constricted, for example by reduced blood flow from the heart, blood flow may be reversed and cause embryonic lethality (Navankasattusas et al., 2008; Tai-Nagara et al., 2017 ). Although arterio–venous defects have not been well studied, in general, the feto–placental vasculature is reduced or disorganised in many mouse genetic knockout models that cause CHD, such as Ncx1 (Slc8a1; Cho et al., 2003), Cxadr (Outhwaite et al., 2019), Flrt2 (ligand for Unc5b mentioned above; Tai-Nagara et al., 2017) and excess glucocorticoids (made by knocking out Hsd11b2 expression, which inactivates glucocorticoids; Wyrwoll et al., 2016), among others (Maslen, 2018; Camm et al., 2018). In the cases of iron deficiency and Cxadr loss-of-function embryos (in which Cxadr has been deleted using Sox2-Cre, which targets epiblast derivatives, including allantoic blood vessels), the heart defects preceded the placental vascular defects. Conversely, in the Flrt2 mutant, placental vascular defects preceded the heart defects. Further, a small placental vascular bed mimicked by occlusion of the outflow tract (using experiments conducted in the chick) can cause a range of CHD defects depending on the level of constriction (Midgett et al., 2017). Curiously, mouse genetic knockouts that primarily impact placental SynT-I cells can also cause CHD (Perez-Garcia, et al., 2018; Radford et al., 2023). CHD could, therefore, be due to primary or secondary heart defects, because heart defects cause placental defects (Outhwaite et al., 2019; Kalisch-Smith et al., 2021, 2022; Radford et al., 2023), and placental defects cause heart defects (Barak et al. 1999; Adams et al., 2000; Perez-Garcia, et al., 2018; Ward et al., 2023; Radford et al., 2023; Fan et al., 2024).

It has not yet been well explored how mutations that affect SynT-I or SynT-II can elicit heart defects. A few possibilities could include knock-on impacts to the differentiation and development of adjacent SynT-II with the EC layer, which would contribute to heart defects through reduced flow. Another possibility, could be secreted trophoblast factors that make their way into the fetal circulation, impacting heart and embryonic development. An interesting avenue would be to examine how mutations that affect SynT-I/SynT-II impact the formation of the arterial and venous vascular trees. Given that trophoblasts are adjacent to both these vascular trees, it has not been considered whether the trophoblast is patterned differently, next to venous/arterial/capillary ECs, and how these interactions could reciprocally impact trophoblast development. Spatial transcriptomics experiments may offer some insight. In addition, given that labyrinth progenitors and SynT-II cells secrete VEGFA, a signal that confers both angiogenesis and arteriogenesis (at high levels) (Pontes-Quero et al., 2019; Luo et al., 2021), how does this spatially pattern and expand the arterial tree, capillary plexus versus the venous tree? It would also be interesting to understand what degree of placental vascular constriction confers CHD and other heart phenotypes such as coronary vessel development, myocardial proliferation etc. This may help us understand why embryos die at specific gestational ages. For example, coronary arteries recruit smooth muscle from E14.5, which, if affected, causes coronary arterial defects and embryonic lethality (reviewed by Smart et al., 2009; Lupu et al., 2020).

Similar to the mouse, human pregnancies with CHD often show placental defects, such as delayed villous maturation and poor perfusion (Matthiesen et al., 2016; Jones et al., 2015; Rychik et al., 2018; Courtney et al., 2020; O'Hare et al., 2023; Mahadevan et al., 2023). Women with preeclamptic pregnancies before 34 weeks of gestation, also have a 7-fold increased risk of their baby developing CHD (Boyd et al., 2017). Placental tissue from preeclamptic pregnancies exhibit feto–placental vascular defects including avascular villi, fetal vascular malperfusion and narrower umbilical cords, among other villous defects including villous hypoplasia, infarction, syncytial knots and perivillous fibrin deposition (Leavey et al., 2016; Benton et al., 2018). Villous hypoplasia is suspected to lead to poor vascular development and ultimately results in reduced placental weight and FGR (Fitzgerald, et al., 2012). These vascular defects are thought to be secondary to trophoblast defects, characterised by aberrant S-FLT and PGF levels – two important angiogenic proteins. Biphasic effects to both trophoblast and endothelium are not surprising considering the important angiocrine signals the trophoblast secretes (along with VEGF) to facilitate feto–placental vascular development as well as vascular remodelling on the maternal side of the placenta (Zhou et al., 2002). Feto–placental vascular defects including poor villous maturation, perfusion and reduced villous stromal cells are also common in pregnancies with FGR (Tun et al., 2019 ;Boss et al., 2023) and stillbirth (Wu et al., 2021; Tiwari et al., 2021; Stallmach et al., 2001). Indeed, 50% of stillbirths are also preceded by intrauterine growth restriction (IUGR) (Figueras and Gardosi, 2011), with 20-60% of stillbirths associated with placental abnormalities (Stillbirth Collaborative Research Network Writing Group, 2011; Varli et al., 2008; Korteweg et al., 2009). Feto–placental blood vessels can also be affected by maternal environmental impacts including iron-deficiency anaemia, gestational diabetes mellitus, smoking, IUGR and FGR (Mayhew et al., 2004).

Studies of vascular formation in vivo in humans are limited by technology and resolution of imaging modalities. Doppler ultrasound can give an overall measure of placental size and utero–placental blood flow in the first trimester scan (Mathewlynn and Collins, 2019), but is yet to estimate regions of the feto–placental tree. That said, first trimester placental volume has become an accurate predictor of conditions including preeclampsia with gestational hypertension (Hashish et al., 2014) and FGR (Papastefanou et al., 2018; reviewed by Srinivasan et al., 2021). Analysis of the feto–placental tree in vivo during early gestation, however, appears to be many years away.

Investigation into human placental arterio–venous vascular defects is currently limited. Current efforts on term placentas use vascular casts (Junaid, et al., 2017) or microCT (Byrne et al., 2021; James et al., 2021). Placentas from pregnancies with FGR show reduced arterial vessels but increased venous vessels (Junaid, et al., 2017). Another study showed that IUGR placentas had reduced arterial and venous branch radii (Saw et al., 2018). The causes of such defects could include poor proliferation or perturbed initial specification of cells to either population, or limited expansion of the arterial or venous trees from reduced/perturbed blood flow. Indeed, deletion of arterial Dll4 in mice reverts ECs to a venous identity (Duarte et al., 2004), which would perhaps explain an increase in venous vessels at the expense of arterial ones. Similarly, in mice, the placental arterioles appear to be more sensitive to blood flow changes i.e. as a result of heart defects (as previously mentioned in mouse models) and are more affected than the venous circulation in the placenta. The umbilical vessels may give clues to placental defects, as they connect to the respective placental arterial and venous vasculature, and abnormal cord insertion is associated with FGR, placental abruption and fetal demise (Moshiri et al., 2014). Further, 10% of stillbirths are associated with umbilical cord abnormalities (Stillbirth Collaborative Research Network Writing Group, 2011). In the context of IUGR, examination of the umbilical vein shows reduced size and structural alterations (Rigano et al., 2008; Peyter et al., 2014). Alternatively, the umbilical artery shows reduced flow in babies before delivery when the terminal villi are maldeveloped (Krebs et al., 1996). For a complete review of umbilical circulation and remodelling in mice and humans see Downs (2022) and Van Schoor et al. (2024), and for pathophysiology see Burton and Jauniaux (2018).

Conventionally, Sox2-Cre and Meox2-Cre have been used to target cells from epiblast derivatives, such as embryonic lineages and extra-embryonic mesodermal lineages including the allantois (but not, for example. yolk sac visceral endoderm), which makes it difficult to determine the causality in genetic knockout studies. Embryonic defects impacting the systemic vasculature and causing lethality could therefore be attributed to several organs and cell types including the heart, yolk sac and placenta. Sox2-Cre and Meox2-Cre approaches have been used in combination with trophoblast-lineage Cres (Cyp19a1 for pan-trophoblast, Tpbpa for junctional zone lineages or Sox2-FLP for pan-trophoblast lineages) to compare phenotypes from genetic deletions from epiblast-derived lineages versus trophoblast lineages (López-Tello et al., 2019; 2023; Radford et al., 2023).

Until recently, we have lacked genetic tools to research allantois lineages (or placental ECs) specifically in mice. However, as previously mentioned, recent studies have determined the constitutive Hoxa13-Cre allele to be allantois-specific, expressed highly in allantoic progenitors (Chen et al., 2022) along with other HOX genes (i.e. Hoxa10 and Hoxa11) at E8.25 (Ibarra-Soria et al., 2018; Scotti and Kmita, 2012). This has, therefore, been a very effective Cre allele for lineage tracing of allantois derivatives; ECs and mural cells (Chen et al., 2022). To target this Hoxa13 lineage during specific developmental windows of placental formation, a tamoxifen-inducible Cre was created (Hoxa13-CreERT2). However, using this Cre to target all placental vessels is not appropriate because Hoxa13 becomes restricted in expression to the umbilical artery and a subpopulation of the major placental arteries following differentiation of allantoic EC progenitors, and it is also expressed in limb buds (Liang, et al., 2021). Therefore, both of these systems may not be good models with which to generate placental EC-specific conditional knockouts. If umbilical ECs are the source of placental ECs, a constitutive Cre approach that targets all allantois derivatives will presumably perturb initial placental EC formation and cause chorio–allantoic attachment defects and embryonic lethality in every model by E9.0-E9.5, preventing later analysis. Conversely, an inducible knockout approach targeting only umbilical and placental stem arteries leaves capillary and venous EC understanding lacking.

In addition, Isl1-Cre has also been used to target allantois derivatives because it is expressed in extra-embryonic mesoderm (Zhu et al., 2024); however, Isl1 is also expressed in embryonic pharyngeal mesoderm and the second heart field of the developing heart (Cai et al., 2003), so cannot be used in a placental EC-specific manner. Therefore, additional genetic tools are now required to complement existing models to better progress our understanding of these processes. Additional genetic tools, such as intersectional genetics, may overcome the problems above and provide ‘placental-EC-specific’ gene deletion. For example, if a generic allantois-Cre, such as Hoxa13-Cre (Cre-lox), is crossed with an endothelial Dre (Dre-rox) line, such as Cdh5-Dre (Han et al., 2021) or Tie2-Dre (Pu et al., 2018), this will provide deletion by excluding cells not co-expressing Cre and Dre. Together these tools may provide us with more information on how placental defects can cause heart defects, amongst other embryonic vascular, and perhaps even lymphatic, outcomes if a systemic approach is pursued.

The feto–placental blood vessels originate from extra-embryonic mesoderm in mice and this is likely to be the case in humans. Arterio–venous differentiation occurs by mid-gestation, with early specification likely taking place around CS11-13 in humans, and by E10.5 in mice. New scRNA-seq datasets are likely to shed light on arterio–venous specification and differentiation and confirm the progenitor populations from which they derive. Defects in early placental EC formation cause embryonic lethality in mice; however, this is yet to be established in humans. Poor development of the feto–placental blood vessels and arterio–venous vasculature could have early origins in the first trimester and should be the focus of future studies.

Thank you to Prof. Shankar Srinivas, Prof. Nicola Smart and Prof. Sarah De Val (University of Oxford) for valuable comments on the manuscript.