The mouse allantois: new insights at the embryonic–extraembryonic interface

Abstract

During the early development of Placentalia, a distinctive projection emerges at the posterior embryonic–extraembryonic interface of the conceptus; its fingerlike shape presages maturation into the placental umbilical cord, whose major role is to shuttle fetal blood to and from the chorion for exchange with the mother during pregnancy. Until recently, the biology of the cord's vital vascular anlage, called the body stalk/allantois in humans and simply the allantois in rodents, has been largely unknown. Here, new insights into the development of the mouse allantois are featured, from its origin and mechanism of arterial patterning through its union with the chorion. Key to generating the allantois and its critical functions are the primitive streak and visceral endoderm, which together are sufficient to create the entire fetal–placental connection. Their newly discovered roles at the embryonic–extraembryonic interface challenge conventional wisdom, including the physical limits of the primitive streak, its function as sole purveyor of mesoderm in the mouse, potency of visceral endoderm, and the putative role of the allantois in the germ line. With this working model of allantois development, understanding a plethora of hitherto poorly understood orphan diseases in humans is now within reach.

This article is part of the theme issue ‘Extraembryonic tissues: exploring concepts, definitions and functions across the animal kingdom’.

1. Overview

Placentalia, one of three mammalian subgroups, are characterized by extreme dependence on their mother during gestation. Nestled within her uterus, the fetus relies wholly on the vast functional repertoire of the chorio-allantoic placenta, with whom it shares its genetic dowry, for survival and development. This major placenta is composed of the allantois-derived umbilical cord and the trophoblast-derived chorionic disc; once formed, the chorio-allantoic placenta separates the fetal and maternal circulatory systems. The umbilical component shuttles fetal blood to and from the chorionic disc, awash in maternal blood, thereby procuring vital nutrients and oxygen and eliminating metabolic wastes.

Despite its critical importance in fetal survival and development, the umbilical cord is one of the most understudied vascular organs in Placentalia, taking a back seat to both the embryo and trophoblast. Although the reasons are not clear, its precursor tissue, the allantois, seems to have been regarded as little more than a bag of unstructured mesoderm whose genesis into a vascular pipe could hardly be remarkable.

Over the past several decades, the mechanisms by which the umbilical cord develops from the allantois have slowly come to light in the mouse—and they are hardly mundane! The most unexpected finding is that development of the mouse allantois is wholly reliant upon interactions between the primitive streak and extraembryonic visceral endoderm. Their newly discovered properties and coordinated interactions not only create the posterior embryonic–extraembryonic interface, which encompasses the allantois, placental vasculature, and hindgut but they also challenge conventional wisdom, including long-held views on the limits of the primitive streak, the origin of mesoderm, rigid potency of visceral endoderm, and the whereabouts of the primordial germ cells (PGCs), or antecedents to the germ line.

2. Prequel to the current model of allantois development

When earnest study of the mouse allantois began several decades ago, facts were scarce. The general morphological sequence of allantois development in rodents had been described [1,2], and included the spatio-temporal whereabouts of the allantoic bud, its enlargement and extension toward the chorion, and union with the latter to create the chorio-allantoic placenta (figure 1a). The allantois appeared to issue from the posterior end of the embryo in register with its primitive streak, or body axis, whose physical boundaries were claimed to be limited to the embryo [7,8]. Brachyury (T) was involved in umbilical genesis, as mouse embryos homozygous for the T deletion exhibited a foreshortened and misshapen allantois [9]. Since the 1950's, the base of the allantois has been regarded as the temporary home to and site of segregation of the PGCs before they migrate to the gonads [10–15]. The establishment of post-implantation whole embryo culture in the rat [16] and its adaptation in the mouse [17] enabled fate mapping in living embryos which showed that at least part of the allantois is derived from primitive ectoderm, or epiblast, the founder population of the three primary germ layers [18,19]. Later, light, scanning and electron microscopy were systematically applied to the rat allantois [20] to describe its changing morphology and rudimentary vasculature which, in accord with previous observations [21], originated de novo rather than by endothelial invasion from the nearby yolk sac or fetus. Finally, with the advent of molecular biology and in situ hybridization, the c-myc proto-oncogene's mRNA was shown to be expressed throughout the base of the allantois [22]. But, how accurately to interpret that finding, as well as the results of a multitude of subsequent transgenic and knockout studies that affected allantois development and/or placentation? A systematic context, i.e. a detailed map of the allantois's cell types, their spatio-temporal origins and fate, assembly into an elongating projectile and its union with the chorion, vascular hemogenicity and relationship to the conceptus's major blood vessels, was needed, as none of this was known.

Figure 1.

(a) Schematic diagram summarizing development of the allantois according to general morphological descriptions prior to discovery of the streak's posterior extension into the exocoelom. Unless otherwise indicated, all views in this and other Figures are sagittal, showing the conceptus and its allantois through the axial midline. The red ellipse encircles the posterior embryonic–extraembryonic junction, highlighting the assumed posterior limit of the primitive streak within the embryo. Briefly, once formation of the exocoelom (xc) is nearly complete (NP approx. 7.0–7.5), the allantois (al) appears to emanate from the primitive streak. It enlarges, vascularizes, and elongates as a projectile toward the chorion (ch), fusing with it to become the chorio-allantoic placenta. (b) Current view of allantoic development vis-à-vis extension of the primitive streak into the exocoelom and contact with extraembryonic visceral endoderm. Modified and reprinted with permission [3]. Copyright 2020, Wiley. 1. Pre-allantoic bud stage. The primitive streak extends into the exocoelom where it is complexed with extraembryonic visceral endoderm (xve). The arrow in this and the other panels indicates the allantoic/yolk sac junction and at this stage, also, the site of placement and formation of the vessel of confluence resulting from the streak's extension. Also indicated is the extent of the primitive streak's reach in wild-type T⁺/T⁺, heterozygous T^C/T⁺ and homozygous mutant T^C/T^C embryos, with coloured dotted lines corresponding to genotypes. This is further explained in c. 2. Allantoic bud stage. Formation of the allantoic bud begins at the posteriormost extension of the primitive streak; once the bud is in place, streak-associated extraembryonic visceral endoderm is referred to as the ‘AX’. The AX-derived endodermal rod is inferred at this stage [4]. 3. Headfold-early somite stages. The extraembryonic extension of the primitive streak expands into a dense core, the allantoic core domain (ACD), under influence of the AX. The AX's rod-like extension has thread itself into the ACD. Axial visceral endoderm is segmented into three distinct domains: blood island-associated visceral endoderm, distal AX, which has given up its share of mesoderm, and the AX which, via contact with the streak, continues to undergo an EMT. 4. 7–9-somite stages. The ACD has retracted toward the hindgut as the latter formed. The AX has become the hindgut lip. The status of the allantoic rod is not known. (c) Placement of the vessel of confluence according to developmental stage and the status of Brachyury. Red asterisks indicate the stages when the VOC becomes misplaced within the embryo rather than remaining fixed within the extraembryonic region. The downward black arrows indicate the direction of displacement toward the embryo of the allantois-yolk sac junction over time. Reprinted with permission [5]. Copyright 2017, Elsevier. (d) The allantoic bud (al) appears near the posteriormost extension of the primitive streak (black arrow), at some distance from the embryonic–extraembryonic junction (dotted line) and presenting as an appendage issuing from extraembryonic visceral endoderm. Modified and reprinted with permission [6]. Copyright 2009, Elsevier. (e,f) Fate mapping the allantois-associated extraembryonic visceral endoderm (AX). The DiI-labelled whole mount images in (e,f) have not been published but are an accurate representation of those that have been [4]. (e) Whole mount view of a headfold-stage conceptus, and DiI label within the allantois-associated extraembryonic visceral endoderm (black arrow). (f) Ventral whole mount view of the hindgut lip. The labelled AX becomes the hindgut lip (white arrow). Inset: Photobleached histological section showing AX contribution (brown colour) to the full panoply of cells and tissues at the posterior embryonic–extraembryonic interface. Reprinted with permission [4]. Copyright 2017, Elsevier. Abbreviations for this and all following figures: A, anterior; ac, amniotic cavity; ACD, allantoic core domain; al, allantois; al-r, allantoic regenerate; am, amnion; AX, allantois-associated extraembryonic visceral endoderm; bi, blood island; ch, chorion; D, distal; da, dorsal aortae; dAX, distal allantois-associated extraembryonic visceral endoderm; DCM, dorsal cuboidal mesothelium; EHF, early headfold stage; em, embryo; ES, early primitive streak stage; hf, headfold (anatomical); hg, hindgut; hgl, hindgut lip; HF, headfold stages; ips, intraembryonic posterior primitive streak (synonymous with pps); LB, late allantoic bud stage; LHF, late headfold stage; LS, late primitive streak stage; n, node; NP, neural plate stages; ntc, notochord; OB, no allantoic bud stage; P, posterior; pac, proamniotic cavity; pbi, prospective blood island; pps, posterior intraembryonic primitive streak (synonymous with ips); Pr, proximal; ps, primitive streak; pxve, proximal extraembryonic visceral endoderm (i.e. beneath the level of the chorion); r, allantoic rod; s, somite pairs; T, Brachyury; tb, tailbud; ua, umbilical artery; VCM, ventral cuboidal mesothelium; vv, vitelline (yolk sac) vessels; xc, exocoelom; xve, extraembryonic visceral endoderm; y, yolk sac; ys, yolk sac.

Now, while the origin of the venous component of the umbilical cord is still largely obscure [23], how the umbilical arterial vasculature is created and patterned is finally coming to light alongside the origins, regional properties and shaping of its precursor tissue, the allantois. The mechanisms were hardly foreseen: the primitive streak and extraembryonic visceral endoderm engage in a highly choreographed pas-de-deux at the posterior embryonic–extraembryonic interface to create the nascent umbilical cord and its connection to the fetus. That new insight is the focus of this review article.

3. Origin of mesoderm

Genesis of the allantois involves formation of the primary germ layer, mesoderm, which underpins development of the entire circulatory system in both the fetus and adult. In the mouse, the primitive streak, or antero-posterior (A-P) embryonic body axis, is claimed to be the sole manufacturer of mesoderm [24]: epiblast translocates into this organizing structure and is transformed into mesoderm, both embryonic and extraembryonic. While fate mapping both the epiblast and primitive streak provided evidence for this claim [18,19,25–27], other sites of mesoderm formation, e.g. extraembryonic tissues, had never been directly examined.

By contrast, in humans, non-human primates, and even the armadillo, extraembryonic mesoderm (often called ‘mesenchyme’, a mesodermal derivative) is observed within early placental tissues in advance of formation of the primitive streak and closely associated with extraembryonic endoderm. In these Placentalia, mesoderm is thought to have both an extraembryonic and embryonic origin [28–32].

As a majority of mouse knockout lines lethal at or after mid-gestation are associated with faulty placentae [33], whose correct vasculature is critical for fetal–maternal exchange, accurate knowledge of mesodermal origins is essential for pinpointing the origin of these defects. At present, the origin of all rodent mesoderm is believed to be the epiblast [18,19,26], with an epithelial-to-mesenchymal transition (EMT) taking place within the primitive streak [34]. The streak then delivers mesoderm in orderly fashion to far-flung sites throughout the conceptus [25,27] where it is classified as embryonic mesoderm (lateral plate, paraxial and axial) and extraembryonic mesoderm. The latter is relevant to this review, as it includes the allantois and lining of the exocoelom, i.e. the extraembryonic cavity in which the allantois will develop into the umbilical cord [35]. Once formation of the exocoelom nears completion, the allantois appears, situated at the posterior embryonic–extraembryonic interface between the yolk sac and amnion, and ultimately becoming continuous with the embryo.

While the notion that all mesoderm of the rodent allantois comes from the epiblast/primitive streak is a long-held view, results of recent experiments show that this is not the case. Rather, as described in this review, the allantois is composed of mesoderm derived from two distinct sources: the primitive streak and visceral endoderm.

4. The primitive streak extends into the exocoelom

The story of the allantois begins with the primitive streak. For the past century, the primitive streak's physical limits have been assumed to lie within the embryo [7] (figure 1a). Within the embryonic region, the streak not only contributes mesoderm to the conceptus, as described above, but as the embodiment of the A-P axis found along the embryo's midline, it directs embryonic organogenesis via its polar coordinates (e.g. [36]). Given the field's intense focus on the embryo, a search for the streak's axial logic within the extraembryonic region, i.e. at the embryonic–extraembryonic junction where it might organize formation of a properly patterned umbilical bridge between the fetus and its mother, had, until recently, been overlooked.

Results of recent studies have shown that, as formation of the exocoelom nears completion, the primitive streak, defined by Brachyury [37], extends into it, juxtaposed to extraembryonic visceral endoderm (figure 1b-1). Analysis of T-curtailed mutants revealed that, just as T is required for axial elongation within the embryo [38], it is also required for the streak's posterior reach into the exocoelom [5]: two normal copies of T and the posterior streak fully extends into this extraembryonic space; one copy, and streak extension is stunted; no copies of T and the streak fails to extend appreciably into the exocoelom (figure 1b-1).

Although getting ahead of the story, it is worth noting at this juncture that other localization studies, though not appreciated at the time, clearly demonstrated the presence of T in the base of the allantois (fully described in §9), both its cognate mRNA and protein (e.g. [39]). Other gene products important in axial organization, including Hox [40] and the Wnts [41], were sighted there, as well. Light and electron microscopy at closely spaced developmental intervals further demonstrated the streak's extension into the exocoelom, where it was closely apposed to extraembryonic visceral endoderm [6].

5. The primitive streak designates the site of formation of the vessel of confluence

What is the significance of the streak's extension into the exocoelom? It is immense, as this extraembryonic segment of the primitive streak now triggers a cascade of highly orchestrated events at the embryonic–extraembryonic interface that culminate, by E13.5, in the correct patterning of the fetal–placental arterial vasculature to ensure survival and development within the mother.

At its farthest reach into the exocoelom, the primitive streak designates the site of formation of the vessel of confluence (VOC), where the conceptus's major arterial vessels will unite [42] (figure 1b-1; §17). The nascent VOC, which appears in advance of the yolk sac blood islands (figure 1b-2), becomes incorporated into the allantoic bud and exhibits a variety of gene products involved in the formation of hemangioblasts [43] and hemogenic endothelium [44].

Similar to its requirement in streak extension, Brachyury is also required in the biology of the VOC, specifically, for its correct placement and patterning along the A-P axis [5] (figure 1b,c). Two copies of T and the VOC forms at the streak's posteriormost extension, situated within the extraembryonic space (figure 1c, T⁺/T⁺). While one copy of T is sufficient to align the VOC with the body axis, stunted extension of the streak leads to the vessel becoming integrated into the embryo, and consequently, to downstream abnormalities in the fetal arterial blood vessels (figure 1c, T^C/T⁺; §18). Complete loss of T results in failure of adequate streak extension into the exocoelom and a mis-patterned and severely abnormal arterial confluence incompatible with survival (figure 1c, T^C/T^C; §18).

6. The primitive streak induces an epithelial-to-mesenchymal transition (EMT) in associated extraembryonic visceral endoderm

Once the streak has fully extended into the exocoelom and the VOC is in place, the allantoic bud appears (neural plate stages, approx. E7.0–E7.5), projecting from the wall of the visceral yolk sac into the exocoelom (figure 1b-2,d). Given the bud's association with visceral endoderm, it was tantalizing to hypothesize that the latter had a role in its creation.

Until recently, the functions attributed to visceral endoderm have been limited to histiotrophic nutrition [45], contribution to the embryonic gut [46], induction [47–49], and conferral of antero-posterior polarity on the primitive streak [50,51]. Reports of roles as a purveyor of mesoderm [28,32] and in formation of blood cells [52–55] were overlooked or ignored, perhaps because they were based mainly on morphology in non-rodent species without any fate mapping [56]; another possibility is that potency studies of yolk sac endoderm, i.e. visceral and parietal, in the mouse failed to show contribution to any tissues other than yolk sac endoderm [57,58]. However, that experimental approach relied on differences in glucose-6-isomerase within dissected tissues of mouse chimeras, a method of limited sensitivity [59].

Thus, to find out whether visceral endoderm contributes to the embryo's complement of mesoderm, streak-associated extraembryonic visceral endoderm was directly fate-mapped at increasing developmental stages [4] using the lipophilic dye, DiI, followed by photobleaching and histological analysis after whole embryo culture [49,60] (figure 1e,f). Analysis of labelled descendants revealed that streak-associated visceral endoderm is a mesendodermal tissue, contributing not only to the hindgut, as would be expected [46], but unexpectedly, and in abundance, to extraembryonic mesoderm, including the allantois and its major components, i.e. its proximal dense core domain (§9), its ventral proximal wall (§8), loose mesoderm, and the allantoic rod (§10), as well as to placental blood vessels, including the VOC (§§ 5, 17, 18), and the umbilical and omphalomesenteric arteries (§§ 17, 18) (figure 1f). No other segment of visceral endoderm exhibited this bipotency.

During its period of delamination and unlike any other extraembryonic visceral endoderm, this small segment of axial extraembryonic endoderm is also rich not only in indicators of pluripotency, including alkaline phosphatase (AlkPhos) activity [10]; Plate 1, Fig. 2 of that publication) and, at headfold through 6-somite pair stages, OCT-3/4 [61] (figure 2a), but especially, of mesendodermal potency, e.g. T [6,63] (figure 2b) and FOXa2 [62] (figure 2c). It also contains cells exhibiting PECAM1 [4] and Runx1 [66], foretelling contribution to hemogenic endothelium of the placental arterial blood vessels. Thus, streak-associated extraembryonic visceral endoderm is a mesendodermal progenitor pool for the fetal–placental interface.

Figure 2.

(a) Localization of OCT-3/4 (brown colour) to the AX (arrowheads) and ACD (asterisk; the ACD is described in §9 and further in figure 3a). Modified and reprinted with permission [61]. Copyright 2008, Elsevier. (b) Localization of T (brown colour) to the AX (arrowheads) and ACD (asterisk; §9 and figure 3a). Unpublished transverse immunohistological section through EHF-stage conceptus (approx. E7.75) shows T within the AX and the ventral disposition of the ACD (§9, figure 3a), and is taken from the same embryo reported in figure 1i,j [6]. (c) Localization of Foxa2 (brown colour) to the AX (arrowheads). Modified and reprinted with permission [62]. Copyright 2017, Elsevier. (d) Grafts of the AX, ACD and AX + ACD into host conceptuses. Arrows indicate 1: AX grafts that did not maintain contact with the primitive streak showed formation of blood cells and paucity of mesoderm production. 2: AX + ACD fully recapitulated the embryonic–extraembryonic interface. 3: ACD grafts that lost contact with the AX lost polarity and the dense core. 4: ACD + AX grafts fully recapitulated the embryonic–extraembryonic interface. Modified and reprinted with permission [4]. Copyright 2017, Elsevier. (e–h) Shaping the allantois via Hedgehog signalling. Included is a brief description of allantois vasculogenesis. See text for details. Modified and reprinted with permission [3]. Copyright 2020, Wiley. See abbreviations in figure 1.

To find out whether contact with the streak was required for delamination of the endodermal epithelium into mesoderm, visceral endoderm was separated from it and grafted (figure 2d). Loss of contact with the streak resulted in production not only of just a small amount of mesoderm but also of blood-like cells [4,56] (figure 2d), as predicted by the morphological studies cited above in humans [52–55]. Control grafts of the streak and visceral endoderm together remarkably recapitulated the posterior fetal–placental interface (figure 2d) (§15). While the signals involved in putative induction are currently speculative (§16), the paucity of mesoderm, along with other parameters whose discussion is beyond the scope of this review, provided evidence that contact with the streak induces an EMT in associated extraembryonic visceral endoderm [4].

On the basis of these results, the allantoic bud is initially composed of visceral endoderm-derived mesoderm. Retrospective analysis suggests that only when the bud has enlarged enough to make contact with the intraembryonic primitive streak at the embryonic–extraembryonic interface does the streak finally pass on mesoderm to it [6,18,19,25–27] and, importantly, as described in §9, only to the proximal allantoic flanks, not to the midline or distal regions.

In the light of these discoveries, many defects at the posterior interface are open to re-evaluation. Here, for example, Cripto would be worth re-considering. Cripto is expressed in the primitive streak and nascent mesoderm, but not in the visceral endoderm [67]. Abrogation of Cripto led to gross attenuation of the primitive streak and a smaller-than-normal allantois [68]. As all mouse mesoderm, including that of the allantois, was assumed to be derived solely from the primitive streak, it was concluded that the mutant streak must not have delivered the appropriate amount of mesoderm to the allantois. However, in light of visceral endoderm's major contribution to the allantois, it is now possible to reconsider that conclusion and speculate that associated visceral endoderm may not have been stimulated to release its share of mesoderm to create the allantois, either because the mutant Cripto streak did not extend far enough into the exocoelom or, if it did enter the exocoelom, its inducing potential was defective.

7. Allantois-associated extraembryonic visceral endoderm (AX) becomes the hindgut lip, a next-generation mesendodermal progenitor pool

Once the allantoic bud forms, streak-associated visceral endoderm is now referred to as allantois-associated extraembryonic visceral endoderm (‘AX’ for short) (figure 1b-2, 1d). Delamination of the AX continues in the direction of the embryo, its mesoderm contributing to an expanding allantois and to the mesendodermal components of the embryonic–extraembryonic interface (figure 1b-3, f) [4]. At about 4-somite pairs (approx. E8.25), approximately 12 h after the foregut appears anteriorly, the AX becomes and/or is incorporated into the hindgut lip [4] (figure 1b-4,f), visible as a slight invagination in the visceral endoderm, just beneath the allantois [24].

The hindgut lip exhibits similar potency proteins/activities as those found in the AX (§6); direct fate mapping of the hindgut lip revealed a major role as a next-generation mesendodermal progenitor pool, contributing to most, if not all mesendodermal tissues at the allantois/hindgut interface [4].

Thus, anatomically, the allantois and gut are intimately juxtaposed (figure 1b-4). This is an important point because many umbilical defects are associated with abnormalities of the gut [69,70]; the visceral endodermal link between the allantois and hindgut offers a means by which to understand these important defects.

8. Shaping the allantois

Conversion of extraembryonic visceral endoderm into mesoderm initially requires relatively low levels of Patched-1 (Ptch-1), Hedgehog's major receptor [71], and Sonic Hedgehog (SHH), one of its major ligands [72], both of which were previously identified in this region within extraembryonic visceral endoderm [44] (figure 2f). As visceral endoderm delaminates, its fully liberated mesodermal cells lose SHH [4]; accumulation of mesoderm then separates the visceral endoderm from the primitive streak and the EMT ceases (figure 2g). Where the endoderm is now distanced from the streak, levels of Ptch-1 increase, resulting in cessation of the EMT and formation of a gap, thereby helping to shape the allantois into its iconic projectile (figure 2g,h) [3,4].

The unique axial segment of high Ptch-1 (referred to as distal allantois-associated extraembryonic visceral endoderm, or dAX, once it has lost contact with the streak) is found nowhere else in extraembryonic visceral endoderm [44]; it was the major clue that Hedgehog signaling was involved in the EMT and morphological grooming the allantois. This is not surprising, as Hedgehog activity plays major roles in a variety of biological processes along the embryo's body axis [73], as well as in formation of blood vessels [48,74–76], and in the EMT [77]. Resolution of the EMT via de-repression of Ptch1 is reminiscent of Hedgehog's role in tumour metastasis [3].

The aforementioned discussion considers shaping the allantois's ventral surface, a major proximal component of which has been referred to as the ventral cuboidal mesothelium (VCM), for the characteristic shape of its closely apposed and relatively impermeable cells [66] (§§ 17, 21). But what about its dorsal surface? It differs from the ventral one in several ways. The dorsal surface is more porous, uniquely localizes alpha-4-integrin, and is highly blebbed [20,66,78,79]. As the allantois reaches toward the chorion, the blebs make transient contact with the amnion [78,79]. However, the function of the allantoic blebs is a matter of speculation, with possible roles including contact guidance to the chorion as the amnion expands [78,79], haematopoiesis, binding to the chorion, and as a reservoir of allantoic tissue to aid chorio-allantoic union [66,80]. Finally, the origin of the cells that comprise the dorsal cuboidal mesoderm (DCM) as it is now called because of its distinct properties [66] (§§ 17, 21; figures 4h and 5a) is not clear; while the AX contributes to the ventral surface [4], contribution to the dorsal one was not noted.

Figure 4.

(a) Schematic diagram of headfold-stage allantois divided into four fate-mapped regions and the underlying intraembryonic posterior primitive streak; indications inside each region reveal contributions to the conceptus. Reprinted with permission [3]. Copyright 2020, Wiley. (b) Left panel: Whole mount view of a post-culture donor conceptus whose allantois and ACD were completely removed; the donor ACD was grafted into the conceptus on the right. In the absence of the ACD, the donor's regenerate allantois (al-r) is unable to elongate. Right panel: Whole mount view of post-culture grafted host conceptus whose allantois had been removed and replaced with the donor ACD graft, showing restoration and elongation of the allantoic regenerate (al-r). Modified and reprinted with permission [6]. Copyright 2009, Elsevier. (c) Histological section of chimeric T⁺/T⁺ ↔ T^C/T^C allantois, showing restored elongation (but lack of fusion with the chorion) of mutant T^C/T^C allantois that received a wild-type T⁺/T⁺ ACD. Modified and reprinted with permission [6]. Copyright 2009, Elsevier. (d) Localization of MIXL1 (brown colour; arrowheads within allantois, arrow in AX) to the ACD and AX. Modified and reprinted with permission [81]. Copyright 2014, Elsevier. (e) Frontal (rear) histological section of Mixl1^−/− mutant showing overgrowth of the allantois. The base of the allantois (dotted line) is this author's interpretation, as it is continuous with the amnion. Reprinted with permission [82]. Copyright 2002, Development. (f) Origin of the ACD founder population. Leftmost drawing indicates clonal labelling of cells within Zones I and II (left to right, respectively, outlined in red); rightmost pair of schematic drawings indicates that cells from Zones I and II, respectively, ended up clustered in the base of the allantois after 24 h of whole embryo culture, and thus, appear to indicate the ACD's stem cell population, a portion of which stays in place after labelling (figure 3d,h). Modified and reprinted with permission [26]. Copyright 1991, Development. (g) Activated Caspase-3 localization (arrow) within a unique axial border cell separates the distal AX from the yolk sac blood island visceral endoderm. Asterisk indicates the allantois–yolk sac junction. Reprinted with permission [4]. Copyright 2017, Elsevier. (h) Schematic diagram illustrating the allantois's major blood vessels and mesothelial domains identified by porosity tests, and their possible role in branching vascularization and protection from branching within the allantois. DCM, VCM, dorsal and ventral cuboidal mesothelium, respectively, have limited porosity. See text for details. Asterisk, VOC. Reprinted with permission [66]. Copyright 2011, Wiley. (i) The allantois of this specimen is wedged between the visceral yolk sac and amnion, and thought to be the normal disposition of these tissues in vivo [78], thereby facilitating directed extension to the chorion and the EMT. BMP4 immunostaining (brown colour, arrowheads) is abundant in extraembryonic visceral endoderm, especially allantois-associated extraembryonic visceral endoderm (AX) that is undergoing the EMT, and in distal allantois-associated extraembryonic visceral endoderm (dAX) that is completing and/or has completed the EMT (2-somite stage, approx. E8.25). BMP4 is also observed throughout the ventral component of the allantois that has just undergone shaping via Hedgehog signalling. Inset: A similar immunohistological section taken from the same embryo which shows, at a higher magnification, a freshly delaminated cell (arrow) from the AX, near the BMP4-positive allantois-yolk sac junction. The status of BMP4 in.the allantois and surrounding visceral endoderm was originally described ([93]; Figure 7 of that publication). The antibody used in the latter, obtained from Santa Cruz Biotechnology, was applied following preparation of histological sections. The antibody used to obtain the data in this panel was from Abcam (Ab39973, 0.4 mg ml⁻¹, diluted 1/200) and carried out first in whole embryos followed by embedding in paraffin and sectioning on the microtome, as previously described [61]. Thus, two independent antibodies from different manufacturers and under different conditions produced similar results. The horizontal line in the main panel is the embryonic–extraembryonic junction. K. Downs 2015, unpublished data. See abbreviations in figure 1.

Figure 5.

(a) Summary of current view of the allantois’s architectural features. See text for details. Reprinted with permission [3]. Copyright 2020, Wiley. (b) Modification of the standard fate map of the zygote [84], incorporating definitive endoderm [4,46] and extraembryonic mesoderm (red arrows), both derived from visceral endoderm [4]. Reprinted with permission [4]. Copyright 2017, Elsevier. See abbreviations in figure 1.

9. The elongating allantois acquires a dense core—the allantoic core domain—and axial polarity

As described in the previous sections, the primitive streak designates the site of formation of both the vessel of confluence and the allantoic bud; in addition, contact with the primitive streak induces an EMT in visceral endoderm to create the bud. But does the streak play other roles in the biology of the allantois? The answer is a resounding yes.

Once the allantoic bud and VOC are in place, the extraembryonic primitive streak expands into a dense T-positive ventral core [6,63,85] (figures 2b and 3a). Here, within the base of the allantois, the primitive streak is now referred to as the ‘allantoic core domain’ (ACD) as by comparison with the distal allantoic region, its cells are characteristically packed closely together [6], even exhibiting E-cadherin [66], a property of epithelial-like cells [86], and which is also found within the primitive streak [34].

Figure 3.

(a) Immunostained section showing localization of T within the base of the allantois of a late headfold stage conceptus (approx. E8.0); this region was since named the allantoic core domain (ACD) (asterisk). Modified and reprinted with permission [63]. Copyright 2006, Elsevier. (b,c) Explanted allantoises on tissue culture plastic, neural plate/late allantoic bud stage (approx. E7.5), in the presence of allantois-associated extraembryonic visceral endoderm (white arrow) (b), and its absence (c). In the absence of the AX, the ACD does not seem to form and the allantois is greatly reduced. Reprinted with permission [6]. Copyright 2009, Elsevier. (d) Schematic diagram, frontal (rear) view, of the posterior side of the conceptus, summarizing the results of DiI labelling the axial midline of the ACD, intraembryonic posterior primitive streak, and the allantoic flanks, both left and right. Reprinted with permission [6]. Inset shows actual result of DiI labelling the axial midline of the ACD, with cells remaining at the site of labelling (white arrowhead), as well as extending further posteriorly (distally) through the axial midline of the allantois. Modified and reprinted with permission [6]. Copyright 2009, Elsevier. (e) Frontal whole mount fluorescence of conceptus whose anteriormost region of the primitive streak, the node (n, white arrow), had been DiI labelled. As with the ACD, cells at the site of labelling remained in place while extending a midline file of cells, here, further anteriorly, to create the notochord (ntc). Modified and reprinted with permission [60]. Copyright 1994, Development. (f) Schematic diagram of late headfold-stage conceptus depicting bifurcation of extraembryonic visceral endoderm (arrow) into the allantois, creating the allantoic rod (r) (arrowhead). Modified and reprinted with permission [64]. Copyright 1974, Company of Biologists (Development). (g) Frontal (rear) view reconstruction of the relationship of the ColIV-positive allantoic rod to the ACD within the allantois. Dotted line represents the lower limit of the allantoic rod; the white asterisk indicates the proximal component of the ACD (figure 4a). Reprinted with permission [4]. Copyright 2017, Elsevier. (h) Ventral whole mount view showing simultaneously DiI-labelled node (n) and ACD (white asterisk) post-culture; DiI remains in place at the site of labelling while sending forth the notochord and allantoic rod further anteriorly and posteriorly, respectively. The gap between the node and notochord (the ‘head process’) has been previously explained [3]. Reprinted and modified with permission [3]. Copyright 2020, Wiley. (i) Frontal (rear view) histological section of an early primitive streak-stage human conceptus (approx. E18.5) showing the pre-umbilical cord consisting of endodermal (‘allantois’ in humans) and mesodermal (‘body stalk’ in humans) components. A discussion of terminology in humans and mouse can be found in [3]. Modified and reprinted from [65], with acknowledgement to the Carnegie Institute of Washington. See abbreviations in figure 1.

From the time it first appears at the headfold stage, the ACD's length remains constant through 4-somite pairs (approx. E7.75–E8.25) [6], after which it wanes commensurate with the timing of hindgut formation. Whether the ACD regresses into the posterior region or disappears altogether is not, at present, known and thus, whether the streak is involved in delamination of the hindgut lip into mesoderm (§7) is not known either.

Several lines of experimentation showed that visceral endoderm is involved in expansion of the extraembryonic extension of the primitive streak into the ACD. In the first, allantoises were explanted in the presence or absence of the AX. In isolation, allantoic explants hardly proliferate, but in the presence of associated visceral endoderm (at late bud stage or later, approx. E7.25 onward), they are very large (figure 3b,c). In the second example, ACDs explanted without associated AX were round (figure 2d) and exhibited scanty cells, with no indication of a densely packed core [4]. Third, by contrast and as described below, grafts of the ACD and AX together maintained the dense core and allantoic polarity (figure 2d).

Concomitant with formation of the ACD, the allantois acquires intrinsic A-P polarity. Acquisition of polarity is undoubtedly essential for patterning the umbilical vasculature and aligning it with that of the fetus and yolk sac. But how was polarity identified? Labelling an axial point within the ACD revealed descendant cells that both remained in place, as well as a labelled population that formed a file which extended posteriorly from the axial point of labelling [6] (figure 3d). In marked contrast, descendants of labelled allantoic flank cells—i.e. on either side of the ACD—remained as a cohort, translocating en masse to the mid- allantoic flanks, with no evidence of any cells remaining behind or moving to the midline (figure 3d). The proximal mesodermal flank cells are likely derived from the intra-embryonic primitive streak whose descendants never contributed to the axial embryonic or allantoic midline but were found only in the allantoic flanks and/or locally dispersed in the embryo (figure 3d).

In that way, the ACD—and not the intraembryonic posterior streak—is analogous to the streak's anterior end, where labelling the node revealed a cell population that remained in place while a file of descendant cells marched further anteriorly, through the midline, to form the notochord [60] (figure 3e). Thus, intrinsic A-P polarity is established within the headfold-stage allantois; polarity is not dependent upon the intraembryonic primitive streak but rather, on the presence of the extraembryonic primitive streak/ACD internal to the allantois.

What is the origin of the allantois's antero-posterior polarity? Given that the primitive streak defines the A-P axis, it stands to reason that the mature ACD is the source. On the other hand, an origin within associated extraembryonic visceral endoderm cannot be ruled out. For example, distal embryonic visceral endoderm determines A-P polarity in the mouse [51]. Moreover, as described above, when the ACD is grafted without visceral endoderm, the resulting allantois is spherical and lacks polarity [4] (figure 2d). By contrast, when grafted with visceral endoderm, the fetal–placental connection is recapitulated and organized, the polarity of its components intact (figure 2d) (§15). The possible role of visceral endoderm in allantoic polarity is discussed more fully in the next section.

10. Visceral endoderm creates an axial rod-like structure

Decades ago, a rod-like extension emanating from visceral endoderm was mysteriously included in a series of schematic drawings, though without any evidence or reference [3,7,64] (figure 3f). Whatever the reasons for depicting the posterior interface in this manner, now a half century later, they seem to be correct. Streak-associated extraembryonic visceral endoderm sends forth a rod-like structure that threads itself through the ACD, forming an axial column of cells through the allantoic midline [4] (figure 3g). As assays to identify the rod required that it be patent and elongated, its precise timing of entry into the allantois is not clear, though it probably entered the allantois at the neural plate stages (approx. E7.0–E7.5) and became part of the ACD by the headfold stage (approx. E7.75) [4]. The endoderm-derived allantoic rod uniquely localizes collagen type IV (COLIV) and is thus distinguished from the PECAM-1-positive umbilical artery which runs alongside it [4].

On the basis of these observations, perhaps the endoderm, and not the primitive streak, is the source of the allantois's intrinsic A-P polarity (§9); this remains to be seen. Whatever the origin of these polar coordinates, the primitive streak is satisfyingly symmetric, capped at each end by dense structures, and sending forth rod-like extensions anteriorly and posteriorly, extending A-P identity further in each direction [3,7] (figure 3h).

Finally, the visceral endoderm-derived outcropping into the allantois is reminiscent of the human body stalk/allantois [65] (figure 3i). At last, the mouse allantois falls in line with the pre-umbilical cords of all other Placentalia where it has been suggested that the endodermal component induces the formation of blood cells [87]. Thus, in addition to directing placement of the umbilical artery along the axial midline (§17), the endoderm-derived rod may be responsible for conferring hemogenicity onto the umbilical artery and VOC [44], and for the allantois's definitive erythro-myeloid potential [88,89] (§17), though this remains to be experimentally tested.

11. The primitive streak's allantoic core domain and extraembryonic visceral endoderm build the posterior embryonic–extraembryonic interface

The ACD is plentiful in proteins indicative of pluripotent stem/progenitor cells, including AlkPhos/Tissue Non-Specific Alkaline Phosphatase (TNAP) [10,11,90,91], OCT-3/4 [61,92,93] (e.g. figure 2a), STELLA (DPPA3) [15,94] and PRDM1 (BLIMP1) [95,96]. To elucidate the role of the ACD in the biology of the allantois, the latter was subdivided into four regions and each one was fate mapped [94] (figure 4a). The T-positive ACD encompassed two regions: a distal one that contributes only to the allantois and a proximal one that contributes not only to the allantois but also to all three primary germ layers at the embryonic–extraembryonic interface [94]. Contributions to the germ layers were similar to those of the posterior embryonic primitive streak with a few exceptions: the proximal ACD supplied the embryo with blood cells and contributed cells to the amnion, while the embryonic primitive streak provided cells to the somatopleure and ventral ectodermal ridge. As shown previously with smaller grafted cell clumps [97], the other allantoic regions contributed only to the allantois, and rarely to the fetus. Of note, and contrary to previous reports of allantois/hindgut specificity [15], protein localization together with fate mapping the ACD and its component STELLA population showed that STELLA-bearing cells were not confined to the hindgut, but populated derivatives of all three primary germ layers ([94]; see also [98] and §20 for discussion of this region vis-à-vis the so-called PGCs).

Thus, the proximal ACD contains populations of progenitor cells that, together with those of the embryonic streak, build the posterior embryonic–extraembryonic interface.

12. The allantoic core domain, and possibly component endodermal rod, are required for allantoic elongation

Given the later discovery of the endodermal rod within the ACD [4], previous results that attributed an elongation function solely to the ACD [6] are subject to re-interpretation. For example, microsurgical removal of the entire allantois resulted in stunted allantoises that failed to reach the chorion [6,83] (e.g. figure 4b). Given that the ACD was missing in the allantoic regenerates, the results suggested that the ACD—or perhaps now, re-interpreted, the ACD/rod complex—is responsible for allantoic elongation.

A role in elongation is supported by grafts of wild-type ACDs into the base of host conceptuses whose own allantois had been removed (figure 4b). The grafted wild-type ACD (or now, ACD/rod complex) completely restored the host's allantois, with the chimeric allantoic regenerate reaching all the way to the chorion to fuse with it. In another scenario, wild-type ACDs were grafted into homozygous T-curtailed mutant allantoises which were missing the dense core domain [63] (retrospectively, the ACD [6] and allantoic rod [4]); chimeric allantoic regenerates reached the chorion, exhibiting mutant host-derived proximal allantoic flanks (figure 4c), consistent with fate mapping (figure 3d) and a donor-derived midline extension and distal region (figures 3d, 4c). T-mutant cell contribution to the chimeric allantoic regenerate was not unexpected, as T-curtailed mutants and wild-type embryos showed no differences in early contribution to the allantois [85].

Thus, the ACD, possibly with its component rod, is required for allantoic elongation.

13. Growth control of the allantois

In the chick, whose developmental programme during gastrulation shares many properties with the mouse [99], the primitive streak behaves as one of two major growth centers [24,100]. To test this hypothesis vis-à-vis the ACD, the distal allantois was removed from embryos, leaving the ACD (now, in retrospect, the ACD/rod complex) intact and exposed within the exocoelom. Following whole embryo culture, the result was surprising: rather than lag behind the allantoises of unoperated controls, regenerated allantoises exhibited catch-up growth, fusing with the chorion at the same time as the controls [6]. These results suggested that the ACD contains an intrinsic growth control mechanism to ensure that the allantois makes its timely rendez-vous with the chorion (§19).

Which genes are involved in this internal timing mechanism of growth control? One obvious possibility is Mixl1, whose gene products localize to both the AX and the ACD [81,101,102] (figure 4d). The embryonic primitive streak is foreshortened in Mixl1^−/– mutants, while the extraembryonic component of the streak appears to be expanded, resulting in an unusually enlarged allantois (figure 4e). Another candidate for growth control is c-myc, whose expression in the base of the allantois [22] set into motion systematic enquiry of the biology of the allantois. Although c-myc's role in mammalian development is still not altogether clear, its loss does affect vascular and hematopoietic development, resulting in mid-gestational morbidity [103], as would be expected of faulty placentation [104]; in addition, results in Drosophila suggest tantalizing roles in growth control [105,106]. Thus c-myc might regulate growth factors/cytokines required for vascular and hematopoietic development [106]. Another possible candidate for growth control is E-cadherin [107] and which, as discussed above (§9), is found in the ACD.

Because the AX is required for both expansion of the ACD and contribution to it (§9), a combination of genetics with classical methods of embryology highlighted in this review illustrates how a multi-pronged approach is essential to understanding the role of any gene product in development, as use of genetic mutants alone is not sufficient.

14. Origin of the allantoic core domain founder population

Can at least part, if not all, of the ACD's founder population be traced farther back in time than when the streak enters the exocoelom at the neural plate stages? A retrospective analysis of clonal fate maps [26] suggests that ACD founder cells are derived from proximal anterior epiblast (Zones I and II, figure 4f) of the early streak-stage conceptus (approx. E6.75–E7.0). They move anisotropically toward the streak and, once in the base of the allantois, remain there as a clustered cell population (figure 4f), reminiscent of those cells that remained in place after labelling the ACD (figure 3d). Thus, at least part, if not all of the extraembryonic primitive streak's founding population may be traced to proximal anterior epiblast at early streak stages.

Apropos of clonal fate mapping, this method can be used not only to address normal fate but also genetic control of that fate. In the case of the allantois, a perfect example was its use in Polycomb-group gene, eed (embryonic ectoderm development) mutants [108]. Fate mapping distal components of the embryonic primitive streak in homozygous eed mutants revealed that defective distribution of its mesoderm resulted in an expanded allantois and allantoic T-domain [109]. Thus, it is important to keep in mind that, in some cases, growth abnormalities of the allantois (§13) might also involve more far-flung sites that, as shown by clonal fate mapping, affect morphogenetic movements.

15. Primitive streak (ACD) and visceral endoderm (AX) are sufficient to recapitulate the fetal–placental connection

The functions of the ACD and AX are mutually dependent: in the absence of the AX, the primitive streak does not expand into the ACD and an elongated allantois is not created [4,6] (figure 2d); in the absence of the extraembryonic component of the primitive streak/ACD, the AX undergoes a limited EMT (figure 2d). When grafted together, however, the ACD and associated visceral endoderm self-organize into a complete version of the nascent fetal–placental connection (figure 2d) (§§6 and 9), forming a normal-sized structure that exhibits most, if not all of the properties of the posterior embryonic–extraembryonic interface: a dense core domain, an elongated allantois, visceral endoderm-derived rod, placental arterial blood vessels that encompass the vessel of confluence and umbilical and omphalomesenteric arteries, and the AX/hindgut lip, all properly organized along the axial midline [4].

Given that both the ACD and AX are sources of putative pluripotent/mesendodermal progenitor cells, the interactions between these tissues may be exploited in future to create an invaluable in vivo organoid model, i.e. a three-dimensional structure created from populations of stem cells that can be used to understand signals required for differentiation of pluripotent cells and mesendoderm [110]. Perhaps primitive streak/visceral endoderm ‘organoids’ could even be formed in suspension culture, obviating the need for grafting and whole embryo culture.

16. Defining three distinct domains of axial extraembryonic visceral endoderm

Thus, during genesis of the allantois, axial visceral endoderm is segmented into at least three distinct regions [4] (figure 2g,h): blood island-associated axial extraembryonic visceral endoderm does not undergo an EMT, in accord with its role in induction of the blood islands [47,48]; the high-Ptch1 dAX, no longer in contact with the streak, is inert, contributing nothing to the conceptus; and the low Ptch1 AX, in contact with the streak, contributes to the panoply of mesendodermal cell lineages at the fetal–placental interface until it becomes the hindgut lip.

How are the borders between these axial segments defined? Intriguingly, at the first signs of EMT resolution, a unique axial ‘border cell’, resembling an apoptotic cell and positive for anti-activated caspase-3 (CASP3), appears between the yolk sac blood islands and the high Ptch1 dAX [4] (figures 2g,h and 4g). Although this cell physically separates the yolk sac blood island endoderm from the dAX, its role is not known. One possibility is that the CASP3 border cell might be involved in stabilization of the visceral endoderm as it delaminates, perhaps involving compensatory proliferation [111]. In this case, continuous death and replacement of the CASP3 border cell would be a signal for nearby cells to proliferate and fill in the gaps made by exiting EMT cells, thereby maintaining the visceral endoderm. However, there was little correlation between the border cell and apoptosis, as defined by trypan blue uptake [4]. Another possibility is that the border cell is involved in non-apoptotic processes rather than in destruction [112–117]. Specifically, it might work together with Notch1. Notch has been implicated in separating signalling domains during development and it also intersects the Hedgehog pathway, especially under conditions of low oxygen [118], as would be found in the mammalian uterus [119]. Moreover, Notch1 is part of an evolutionarily conserved intercellular signalling pathway that regulates interactions between contiguous cells, as well as apoptotic cell death [120]. Intriguingly, although not highlighted in that study, Notch1 can be observed anecdotally to localize to a single punctate site near the allantois-yolk sac junction within extraembryonic visceral endoderm (Fig. 1A of [121]). Moreover, Notch1 knockouts lead to unrestricted arterial vessel formation [122] and haemorrhaging at the fetal–umbilical connection [123]), similar to the phenotype observed when Hedgehog is abrogated (§17). The appearance of the single CASP3 axial cell is under the control of Hedgehog signalling, as it was absent in Hedgehog-inhibited specimens [4]. To the best of this author's knowledge, the CASP3 cell is the only example thus far discovered in mammals where a unique single cell separates two distinct developmental domains.

The AX domain is defined by contact with the primitive streak; whatever signals induce the EMT likely originate within the latter [4]. Although Brachyury has not been implicated in Hedgehog signalling, it is involved in the regulation of key mesendodermal genes [124]. In human cells, downregulation of T leads to downregulation of SNAI2 and upregulation of E-CAD, both of which were found in the ACD/AX domain [4], while in the mouse, downregulation of T leads to downregulation of N-CAD and FOXa2, both of which were also documented in the ACD/AX domain [4,62].

17. Vasculogenesis within the allantois

As in the yolk sac and fetus, allantoic blood vessels are formed de novo, by vasculogenesis [42,125]. During elongation and shaping, the allantois is now supplied with adequate mesoderm to form three distinct vasculatures: (i) a distal branched vascular network, (ii) the proximal unbranched umbilical artery, and (iii) the vessel of confluence (VOC).

Allantoic angioblasts were identified by both morphology and expression of Flk1 [42,126,127], diagnostic of all blood vessels [128,129], and later supported by a variety of other gene products typical of endothelial cells [130–132]. Angioblasts first appear in the distal allantois at the headfold stage (approx. E7.75–E8.0) (figure 2g). Given the timely need for establishment of a vascular conduit between the fetus and its mother, establishing a branched vascular network at the site closest to maternal blood is an expedient biological strategy. Angioblasts continue to appear with distal-to-proximal (posterior-to-anterior) polarity down the length of the allantois [42]. Endothelialization and branching follow closely behind (figure 2h). Distal-to-proximal polarity of vasculogenesis is also maintained in the allantois in explant culture [130], providing further evidence that, by the headfold stage, axial information is intrinsic to the allantois (§9).

By contrast, the umbilical artery forms in the axial midline of the proximal third of the allantois, alongside the allantoic rod [4] (figure 2h); the umbilical artery is unbranched to ensure mistake-proof focal union with the vessel of confluence (figures 1b-4, 4h).

The vessel of confluence, which is positioned by the primitive streak (§5) (figure 1b), becomes incorporated into the allantois at its junction with the yolk sac; despite that junction translocating toward the fetus as the allantois undergoes shaping, the VOC remains associated with it (figure 1b). Presumptive VOC hemangioblasts gradually endothelialize, requiring FGFR1 [5,42]. The VOC is highly conserved, having now also been identified in humans, pigs and rabbits [5].

Although Brachyury localizes to allantoic blood vessels [5,63], it is not required for angioblast formation and endothelialization [85] but rather for axial placement of the VOC and umbilical artery, and for the patency of the allantoic rod and umbilical artery [4]. Patterning the umbilical artery along the allantoic rod is reminiscent of patterning the embryo's paired dorsal aortae in other amniote species, which is guided by the notochord at the other end of the primitive streak [133–135]. Other T-box genes have been implicated in allantoic vasculogenesis [131,136], but their functions are not yet clear.

Not surprisingly, given its role in mediating the EMT and shaping the allantois, Hedgehog is required for overall vascular patency of the major arterial vessels [4]. In the absence of Hedgehog signalling at early stages, the allantois does not form a projectile and the nascent arterial vessels are disorganized and unresolved [4]. At later stages, Hedgehog is required for individualization of the posterior arterial placental vasculature, especially the vessel of confluence [4].

Is allantoic vasculogenesis accompanied by primitive erythropoiesis as it is in the yolk sac? The rodent allantois does not exhibit blood cells until post-confluence with the yolk sac when blood cells can freely enter the allantois from elsewhere [20]. Removing the allantois prior to confluence and then grafting them free-floating into the exocoelom showed that allantoic vasculogenesis is not accompanied by primitive erythropoiesis [42]. However, when cultured in isolation prior to confluence and in the presence of haematopoiesis-stimulating factors, similar-staged allantoises exhibited definitive erythro-myeloid potential [88,89]. We know now that, at the time these isolates were made, visceral endoderm had made substantial contributions to the allantois [4]; given visceral endoderm's hematopoietic potential, shown by grafts that have lost contact with the streak (§3; figure 2d), as well as by morphological evidence in humans (§6), that of the allantois might come from visceral endoderm [56]. This remains to be seen. Moreover, as the placenta is a source of hematopoietic cells [137,138] whose founding population's whereabouts are still unknown, an origin within visceral endoderm could be easily addressed in a variety of experimental settings that do not require specialized whole embryo culture [56].

Finally, it is tantalizing to speculate that the characteristic patterning of the nascent umbilical vasculature involves the distinct surface properties of the allantois [66]. The mesothelial surface of the distal two-thirds is highly porous, while the proximal third is protected by epithelial-like cells that limit passage of exocoelomic factors (§8). Thus, vascular growth factors within the exocoelomic cavity might enter the porous distal region and stimulate endothelial branching, while in the proximal region such factors would be blocked [66] (figure 4h). Recently, FGFR2 was found to localize throughout the distal allantois, and remarkably, was excluded from the proximal region [5]. Loss of FGFR2 results in allantoic vascular defects [139]. Perhaps, via porous mesothelium, FGFR2 receives exocoelomic factors that stimulate creation of the branched network. Unfortunately, the content of the mouse exocoelom has been little characterized [140,141].

18. The fetal–placental arterial confluence and downstream consequences

During allantoic vasculogenesis, the yolk sac's omphalomesenteric artery is forming nearby, and the posterior ends of the embryo's paired fetal dorsal aortae merge into a single segment as they reach the base of the allantois. Although its precise origin is unknown, the omphalomesenteric artery may originate within the yolk sac blood islands, proceeding proximally toward the VOC [42]. By 4-somite pairs (approx. E8.25), this major artery is visible at the allantoic-yolk sac junction, near the VOC [42] (figure 2h).

By 4–6 somite pairs (approx. E8.25–E8.5), the vessel of confluence brings together the umbilical and omphalomesenteric arteries, and fetal dorsal aortae (figure 1b-4). Union involves branching of the VOC, requiring FGFR1 [5], and then proceeds sequentially, with the omphalomesenteric artery first to join, followed by the umbilical artery, and lastly by the now-fused posterior end of the paired fetal dorsal aortae [5].

Once the arterial confluence is established, a series of major remodelling events takes place within the VOC that, because of Brachyury having fixed the latter's position at the allantois/yolk sac junction, remains associated with the hindgut. Beginning at approximately E8.75 (approx. 10-somite pairs), the tailbud expands, causing an anterior shift of the base of the allantois which brings the VOC with it. This causes the dorsal branches of the VOC to extend ventrally and to curve around the expanding hindgut. Traditionally, the origin of these branches has been obscure, probably because their study took place after arterial union [142] and/or because the imaging tools used at earlier stages were not sensitive enough to detect the vessel of confluence [23], due to its small size [5]. As a result of discovery of the VOC, we now know that the ‘re-curved portions of the dorsal aortae’ [143,144] or the ‘medial umbilical roots' [142] are derived from the VOC [5].

Between approximately E10.0 and E13.5, the medial umbilical roots regress to become the caudal mesenteric artery in the mouse (the inferior mesenteric artery in humans) [142]. The site of regression of the medial umbilical roots further designates the site of formation of the lateral umbilical roots from the aorta, which take over the connection between the aorta and the umbilical artery [142]. Ultimately, the lateral umbilical roots become the common iliac arteries which divide the flow of blood from the aorta into each hindlimb [142].

Not surprisingly, the success of this entire series of remodelling events is wholly dependent upon T-mediated extension of the primitive streak into the exocoelom, which began six days earlier (figure 1b). Failure of the streak to extend into the exocoelom and designate the appropriate axial level of formation of the vessel of confluence has catastrophic domino effects on the placement and/or formation of the common iliac arteries [5].

Finally, the allantois becomes the mature umbilical cord. Contrary to the situation in the neonate and adult, the embryonic arterial blood supply carries de-oxygenated blood away from the fetus to the chorion while the venous supply picks up oxygen via maternal blood within the chorion and delivers it to the fetus. At birth, rapid umbilical closure forces re-routing of the fetal blood supply [145] and the umbilical cord is severed from the fetus, leaving behind only a trace of its former pre-eminence in the form of the belly button, or naval. Inexplicably, for so important an organ, not a single monograph systematically documents later stages of umbilical development in any species.

19. Chorio-allantoic union

By approximately E8.25, when the major arterial blood vessels are coalescing in the proximal region of the allantois, chorio-allantoic union is beginning in the distal region (figure 1b-3,b-4). Why is this event relevant to the embryonic–extraembryonic interface? For several reasons. First, chorio-allantoic union takes place at a central location on the chorion beneath the site of entry of the maternal spiral artery [146]. Control of this striking alignment might involve Brachyury, as T was unexpectedly identified not only in the allantois but also in extraembryonic ectoderm and its derivative, chorionic ectoderm [63,147] (figure 3d). The ACD-rescued chimeric T⁺T⁺ ↔ T^C/T^C allantois touched, but did not fuse with, the chorion (§12) (figure 4c), suggesting that T is required in the central chorionic dome to direct fusion there; this would create an axially patterned vasculature that extends from fetus to mother via the embryonic–extraembryonic interface. Second, chorio-allantoic union occurs within a narrow timeframe and is dependent upon the developmental maturity of the allantois [79]. Catch-up growth in +ACD allantoic regenerates (§13) provided evidence for an intrinsic timing mechanism within the base of the allantois, i.e. at the embryonic–extraembryonic interface, which controls appropriately timed union, discussed below. Third, at the time of chorio-allantoic union, the ACD is regressing/disappearing and the allantois is differentiating; it makes sense that it should lose its stem cells.

At the time of chorio-allantoic union, the allantois has enlarged through a combination of proliferation [20,148] and abundant hyaluronic acid [149], whose hygroscopic nature facilitates swelling of a distal hydropic space to ensure timely contact with and rapid spreading over the chorion. Chorio-allantoic union is gradual, occurring over a specific timeperiod (approx. E8.25–E8.75; 4–8-somite pairs). While union involves the mesothelial surfaces of both tissues [79], it is dependent upon the developmental maturity of the allantois rather than on the maturity of the chorion [79]. The expression/localization patterns of VCAM-1 and its co-receptor, alpha-4-integrin, required for chorio-allantoic union [150–152], reflect the spatio-temporal requirements of this event: VCAM-1 is gradually expressed in the distal two-thirds of the allantois, reaching its maximal intensity by 4–8-somite pairs when chorio-allantoic union occurs, while its co-receptor, alpha-4-integrin, is constitutively found in chorionic mesothelium [78].

The chorio-allantoic placenta is one of two placentae in Placentalia (the other being the chorio-vitelline or yolk sac placenta); once formed, its function takes precedence throughout the remainder of gestation. Now supplied with fetal blood on one side and bathed in maternal blood on the other, the chorio-allantoic placenta separates the fetal and maternal circulatory systems so that all exchange must occur across this interface. In that way, and during its brief lifespan, the chorio-allantoic placenta serves as the major source of the fetus's nutritional needs and the sole route of disposal for fetal waste.

20. The allantois and primordial germ cells

As noted in §2, for well over the past 70 years, the biology of the allantois has been ignored in favour of studies of so-called ‘PGCs’, which are claimed to have an extragonadal origin in mammals. This theory has been around for well over a century, beginning with claims that morphology could identify cells that resembled gonadal gametes in far-flung sites across the embryo as it underwent organogenesis [153]. Because no one had actually traced these so-called extragonadal PGCs to the gonads, this notion was ultimately challenged during the period when histochemical staining became the sought-after means by which to identify specific cell types. Striking alkaline phosphatase (AlkPhos) activity was observed in the allantois-associated yolk sac in humans [154] and in the base of the allantois and nearby visceral endoderm in mice [10,11,91]. As the allantois fused with the chorion, AlkPhos activity was lost there while at the same time, the hindgut gained it. The ‘migratory hindgut PGCs' then moved into the associated dorsal mesentery, ‘travelling’ along them and ending up in the gonads [10,91]. Despite still the lack of fate mapping in situ to demonstrate that AlkPhos-positive allantoic cells actually populate the gonads, an extragonadal origin for the mammalian germ cell lineage became firmly embedded as dogma in the scientific community.

Further, with the advent of molecular biology, i.e. the ability to clone genes of interest and identify their cognate mRNAs in situ, the field extended the notion that gene activity identified cell lineage. Thus, Oct-3/4 became a new ‘marker’ of the PGC lineage, as its expression pattern was found to be similar to that of AlkPhos activity and indeed, overlapped it [93]. Oct-3/4's promoter was later hooked up to ß-galactosidase (LacZ) [155], or Green Fluorescent Protein (GFP) [156], becoming a reporter strategy for identifying so-called PGCs, and thereby by-passing tedious hybridization in situ. In addition, posterior AlkPhos positive cells could be transformed into ‘embryonal germ’ (EG) cells in culture [157,158] (see also §21), paving the way for the isolation of additional molecular ‘markers' of PGCs, e.g. STELLA (DPPA3) [15], FRAGILIS (ITIFM3) [15] and PRDM1 (BLIMP1) [96]. Not surprisingly, given their discovery within the posterior AlkPhos-positive cell population, these genes' expression patterns overlapped the AlkPhos trajectory and are often used instead of AlkPhos to identify so-called PGCs.

However, AlkPhos-related gene products neither co-localize to every AlkPhos-positive ‘PGC’ nor to each other; neither are their localization patterns restricted to the so-called PGC trajectory, i.e. allantois and visceral/hindgut endoderm [10,62,91,94,159,160]. Like AlkPhos activity and its cognate gene product, TNAP (tissue non-specific alkaline phosphatase), also found in abundant posterior sites, including extraembryonic tissues [10,90], cells bearing STELLA, FRAGILIS and PRDM1 localize to derivatives of all three primary germ layers throughout the posterior region, and in distinct subpopulations [62,94,159,160]. Importantly, fate mapping the allantois's ACD—which, retrospectively, is the site where ‘PGCs’ are claimed to segregate from the soma [13]—showed that part of its STELLA population ends up contributing to the allantois, and the other, to the entire embryonic–extraembryonic interface [94,160]. On the basis of these varied expression data throughout the posterior region, how can the field claim that any one of these gene products, or even a battery of them, identifies a unique PGC lineage and distinguishes it from somatic contributions? They cannot.

False equivalence between gene expression and cell lineage has been extensively documented and presented [98], gene by gene, for all of the ‘PGC markers’, above, as well as others, e.g. NANOG [161]; to reiterate that history here is a monumental undertaking that goes far beyond the scope of this review. ‘Off-piste PGCs’ identified by gene expression outside of the trajectory are likely to be stem cells that build those organs in which they are found [98,160], especially as the period between E7.5 when so-called ‘PGCs’ first appear in the allantois through E10.5 when the gonads form is an intense period of organogenesis throughout the conceptus.

While it is understandable that, early in the history of germ cell biology, scientists wanted and even needed to exploit whatever means, albeit limited, were at their disposal to explore their theories, germ cell biologists are now deeply caught in a tangled web, breaching time and again a fundamental principle of embryology: gene expression and cell lineage do not form an obligate ancestral relationship [162]. In the absence of independent lineage tracing to the gonads, allantois ‘PGCs’, hindgut ‘PGCs’ and/or those of the dorsal mesentery cannot be guaranteed the imprimatur ‘primordial germ cells'. Rather, at present, the only certainty is that cells in the base of the allantois, associated visceral endoderm, and hindgut lip contribute to the posterior embryonic–extraembryonic interface of the mouse gastrula, together establishing an organized framework upon which the fetal–maternal circulatory continuum is created.

Knowing the whereabouts of the stem cells that build the fetal–placental interface predicts that morphological defects in the allantois will be accompanied by defects in those tissues which harbour its stem cell populations (what the field calls ‘PGCs’), i.e. the primitive streak and visceral endoderm. Accordingly, a correlation was found between defects in the allantois, primitive streak and/or visceral endoderm, and loss, mislocalization, and/or reduced numbers of so-called PGCs, identified by AlkPhos and/or associated gene products, above, for a large number of mutations (Table 1.1 of [98]): Bmp2 [163,164], Bmp4 [12,165], Bmp8b [166], Cdx2 [167,168], Eed [109], Foxa2 (Hnf3b, Tcf-3b) [169–171], Lhx1 (Lim1) [171,172], Otx2 [12,172,173], Smad1 [174–176], and Smad5 [173,177,178].

Re-examination of each of these gene products in light of the biology of the allantois will be an important future assignment, but for now, only two of them are highlighted here: Bmp4 because its requirement in so-called ‘PGC’ development [12,15,163,179] has become a guiding principle in the field of germ cell biology [180], and Smad1, because its loss affects both visceral endoderm and the allantois, and it is required for expression of Bmp4.

BMP4 is found in the embryonic primitive streak, the allantois's distal regions and mesothelial surface, extraembryonic ectoderm [12,83,165], and extraembryonic visceral endoderm [83] (figure 4i) which, because of its limited span of expression and/or sensitivity of the reporter used, was not noted in the other studies [12,165]. Homozygous Bmp4 mutants exhibit a defective primitive streak, scanty mesoderm, and both the allantois and so-called PGCs, as judged by AlkPhos activity, are missing [12,165]. Heterozygous Bmp4 mutants exhibit many fewer AlkPhos-positive cells, but development, including that of the allantois, appears normal (though neither the gonads nor reproductive fitness of the adult heterozygotes were assessed; nevertheless, normal development in heterozygotes is consistent with entire loss of TNAP, which does not affect development [90]). In chimeras, wild-type reporter Bmp4^+/+ ES cells did not colonize extraembryonic ectoderm or, importantly for this discussion, visceral endoderm of Bmp4^−/− mutants; and in cases where the entire embryonic region was populated by wild-type cells, neither the allantois nor its AlkPhos-positive cells were rescued.

On the basis of these data, the authors offered several possible explanations for the role of Bmp4 in the posterior region. The one that seems to have stuck over the years is a dose-dependent requirement for BMP4 signalling within extraembryonic ectoderm to create PGCs and/or the allantois (e.g. [15]). However, in light of what we now know about the biology of the allantois and its creation, is there an alternate explanation? After all, extraembryonic ectoderm has no known role in the biology of the allantois. I think that the answer is ‘yes’: BMP4 is more likely required in allantois-associated extraembryonic visceral endoderm to create the allantois and its ACD stem cell pool, especially as it is found in endoderm during the critical stages of bud formation (approx. E7.25–E8.0) [83]). That observation, along with key results of the wild-type Bmp4^+/+ ES cell ↔ Bmp4^−/− chimeras suggest two possible scenarios. In the first, visceral endoderm lacking BMP4 could not induce the wild-type extraembryonic primitive streak to enlarge, the result being loss of the ACD and its AlkPhos stem cells. In a second scenario, the wild-type streak could not induce mutant Bmp4^−/− visceral endoderm to delaminate into the allantoic bud and produce the signals required for the streak to expand into the ACD and its stem cells (§9). In both cases, the result would be a significantly diminished or absent allantois lacking the ACD and its complement of AlkPhos-positive cells. Thus, BMP4 is likely required in visceral endoderm to create the allantois and primitive streak's ACD and its component stem cells.

Supporting a role for BMP4 in visceral endoderm and formation of the allantois is SMAD1, whose loss simultaneously affects extraembryonic visceral endoderm, the allantoic bud, and so-called ‘PGCs’ [176], i.e. allantois AlkPhos-positive cells. Smad1^−/− mutants lacked BMP4 in visceral endoderm with the result that extraembryonic visceral endoderm was ‘erratically’ folded. As in wild-type Bmp4^+/+ ES cell ↔ Bmp4^−/− chimeras, above, the wild-type ES cells of Smad1^+/+ ES^+/+ ↔ Smad1^−/− chimeras colonized the epiblast of Smad1^−/− mutants, but not their extraembryonic visceral endoderm. The result was a small bud resembling the allantois, and diminished numbers of AlkPhos cells. Here, the authors concluded that Smad1 was required for function of the visceral endoderm during gastrulation—however, they did not explain how visceral endoderm could affect development of the allantois. I suggest that, just as in the BMP4 scenario, above, the data are consistent with failure of the visceral endoderm to form the allantoic bud and induce expansion of the extraembryonic component of the primitive streak into the ACD and its component stem cells. On a final note, and highlighting a role in visceral endoderm, BMP4 signalling directs XEN cells, which are derived from primitive endoderm [181], to acquire an extraembryonic visceral endoderm identity [182].

21. Conclusion and future directions

Results discussed in this review are summarized, according to developmental stage [183], in table 1. In addition, the current architecture of the allantois and surrounding region is presented in figure 5a, alongside a revision [4] of the standard fate map of the zygote [84] (figure 5b). It is now clear that the base of the allantois and its associated visceral endoderm are where some of the most important functions having to do with establishing the vital fetal–placental connection take place. Yet, the posterior embryonic–extraembryonic interface of the mammalian conceptus has been overlooked for decades in favour of study of the embryo alone. As early as the blastocyst stage, embryonic and extraembryonic tissues have been shown to be mutually dependent on each other for growth and development [184,185]. We now know that mutual dependency extends to post-implantation development, and especially to the posterior embryonic–extraembryonic interface.

Table 1.

Timeline of major events in development of the mouse allantois at the posterior embryonic–extraembryonic interface. ACD, extraembryonic streak-derived allantoic core domain; A-P, antero-posterior; AX, allantois-associated extraembryonic visceral endoderm; EMT, epithelial-to-mesenchymal transition; HGL, hindgut lip; IPS, intraembryonic posterior primitive streak; L-R, left-right; LUR, lateral umbilical roots; MUR, medial umbilical roots; PS, primitive streak; VOC, vessel of confluence; XPS, extraembryonic primitive streak; XVE, extraembryonic visceral endoderm.

embryonic day post-coitum (E) (approximate)	developmental stage	major event(s)
∼E7.0	neural plate/no bud	extension of PS into exocoelom
		posteriormost limit of XPS designates and aligns site of formation of VOC
		anterior proximal epiblast cells move anisotropically toward the base of the allantois
∼E7.25	neural plate/early bud	XPS induces associated XVE to undergo EMT
		allantoic bud forms and is composed of XVE-derived mesoderm
		VOC angioblasts appear within allantoic bud at al/ys junction
		XVE extends allantoic rod into the allantois
∼E7.5	neural plate/late bud	EMT continues in axial AX
		allantoic bud enlarges
		VOC enlarges
		XVE-derived allantoic rod enters allantois
∼E7.75–E8.25	headfold – 4 s	XPS expands into T-positive ACD under the influence of associated AX
		AX continues to release mesoderm
		allantois shaping begins
		allantoic bud becomes continuous with embryo
		anterior proximal epiblast cells arrive in the base of the allantois and become part of the ACD
		allantoic ColIV-positive rod lengthens via ACD, forming an axial midline structure
		allantois acquires A-P polarity and L-R asymmetry
		IPS adds mesoderm to allantoic flanks
		ACD and IPS build embryonic–extraembryonic junction
		allantoic vasculogenesis
		elongation and shaping of allantois
∼E8.25–E8.75	4–8 s	chorio-allantoic union
		mitotic index slows
		distal allantoic angioblasts endothelialize and branch, penetrating chorion upon union
		unbranched umbilical artery visible alongside allantoic rod
		omphalomesenteric artery appears near allantois/yolk sac junction
		embryonic dorsal aortae reach the base of the allantois
		VOC unites the major arterial vessels
		ACD retracts/disappears toward hindgut
		AX becomes the HGL
		HGL becomes mesendodermal progenitor pool
∼E8.75–E10.0	early organogenesis	formation of MUR from VOC
∼E10.0–E13.5	mid-organogenesis	successive formation of LUR and common iliac arteries

These are exciting times. The posterior embryonic–extraembryonic interface is no longer a black box; rather, knowledge of its biology opens up an entirely new mammalian region to the creative thinking of the next generation of developmental embryologists to explore a vast arena of developmental questions, including continuity of antero-posterior polar coordinates between the fetus and its mother; the origin of placental blood cells, looking more closely at visceral endoderm and its hematopoietic potential [56]; further exploration into the origin of mesoderm, perhaps including the poorly understood sinuses of Duval [2], which are composed of extraembryonic endoderm that creates the intraplacental yolk sac where robust Hedgehog signalling is also found [44]; and the underpinnings of a variety of orphan diseases, many of which involve the umbilical cord, yolk sac and gut [3,69,70]. The results featured here can also be used to garner new insight into how formation and organization of the arterial confluence affects fetal and long-term adult development. For example, many mouse mutants have demonstrated that defects in placentation and the heart go hand-in-hand [152,186]. If the umbilical cord were examined more closely at birth rather than discarded as it is in most birthing rooms, it could act as a potentially life-saving sentinel that picks up subtle post-natal heart defects [187]. The problem, though, has been what to look for? The answer is still unclear, but hopefully a monograph of umbilical development is forthcoming that, when placed in the context of new understanding of the mouse allantois and fetal–placental interface, will provide clues which foretell the quality of the post-natal circulation, especially the heart.

In light of new architectural and functional features at the embryonic–extraembryonic interface, many more genes are worth re-evaluating in this region [188]. For example, while the Wnt signalling pathway affects development of the primitive streak, its role in the allantois is still unknown [41]. But, as with all mutational studies, given the rapid changes that occur at this interface, any new scrutiny must involve careful spatio-temporal localization of cognate gene products at closely spaced intervals, and in multiple orientations in histological sections; otherwise function cannot be assigned. For example, not only does BMP4 in visceral endoderm provide an alternative explanation for its role in the allantois and associated pool of stem cells, but T's unexpected and widespread localization in extraembryonic tissues, including the allantois, visceral endoderm and extraembryonic ectoderm, necessitates re-consideration of previous conclusions concerning T's function at the posterior interface when it was believed that T was restricted to the embryo [189].

Finally, accurate knowledge of the origin of the germ line is important for devising treatments for infertility; moreover, because the germ line recapitulates the entire organism, such cells undoubtedly hold secrets for understanding normal potency and regeneration, as well as the abnormal processes of tumorigenesis and metastasis. Despite lack of fate mapping, it has been become acceptable to claim, on the basis of gene expression alone, that the allantois and associated endoderm serve as temporary sites of residence for the germ line antecedents (e.g. most recently [190]). However, we now know that these same sites harbour stem cells which build the fetal–placental interface. Perhaps the time has come to consider that the origin of the mammalian germ line is, at present, unknown. While it could originate within extraembryonic tissues in humans, mice, and other species where similar gene expression has been found [154,191–197], not a single experiment shows lineage continuity independent of gene expression between extragonadal sites and the gonadal gametes. Rather, the data currently in hand argue that allantois/hindgut stem cells in the mouse are conserved across species to build the fetal–placental interface common to all of them.

Based on this new knowledge of the posterior embryonic–extraembryonic interface, challenges to PGC dogma over the past decade [56,62,94,95,98,159,160] might be gaining traction. With the advent of powerful methods of single-cell analyses of gene expression, scientists in the field of transcriptomics have admitted that what appear to be homogeneous cell populations are frustratingly heterogeneous and that their transcription patterns change markedly over time [198]. Most importantly, they duly recognize non-equivalence between gene expression and cell lineage and have concluded that the former, gene expression, cannot be used to trace the latter, cell lineage [199]. Rather, an indelible mark that can withstand generations of cell division and which is distinct from gene expression must be used. CRISPR technologies that employ a sort of cellular barcode look highly promising [200]. But even then, cell lineage tracing is not going to be easy—the major block in the case of the mouse gastrula is that whole embryo culture is stage-limited, and to only a few days at most, probably owing to increasing reliance on a functional placenta [201].

Therefore, a holy grail in mammalian biology must surely be the creation of a means by which to trace, in vivo, a particular cell or group of cells over a long period of time and through major events such as gonadogenesis. One possibility that might be revived is transplantation of early post-implantation mouse conceptuses from their own implantation site to the deciduum of another pregnant mouse [202]. At low frequency, this tour-de-force line of experimentation was successful, as approximately E6.0–7.0 embryos underwent apparent normal development and placentation through approximately E12–14. One could envision removing embryos at approximately E7.5–E8.0, injecting into the base of their allantois an indelibly labelled population of cells, and placing these chimeras into the deciduum of a pregnant foster mother for further development. Once the gonads form, these can be analysed for the presence of the labelled cells and their descendants. Overcoming the low frequencies of success in such experiments may involve ensuring less uterine crowding and better placental connections between the conceptus and its mother, which nowadays might be rectified through implanting endothelial growth factor-soaked beads within the donor's placental tissue prior to transplantation. This is, at present, speculation, but given the original proof of principle [202], achieving this goal is not beyond the realm of possibility.

In the meantime, despite recent challenges to the origin of the mammalian germ line, more has been gained than has been lost from decades of investigation of ‘allantois PGCs’. Posterior cells of the embryonic–extraembryonic interface can, with extensive treatment in vitro, contribute to the germ line [157]. Consequently, a plethora of interesting and potentially important genes and signalling networks with functions in stem cell potency, epigenetics, and differentiation relevant to gametogenesis have since been discovered. But potency is not equivalent to fate. Therefore, it is tantalizing to speculate that the posterior stem cells which build the embryonic–extraembryonic interface represent a tabula rasa which may be used in future studies of embryonic potency, unraveling the genetic control of differentiation into many types of cells.

Data accessibility

This article has no additional data.

Authors' contributions

K.D.: conceptualization, data curation, formal analysis, funding acquisition, investigation, methodology, project administration, supervision, validation, visualization, writing—original draft, writing—review and editing.

Conflict of interest declaration

I declare I have no competing interests.

Funding

I received no funding for this study.