Investigation into a role for the primitive streak in development of the murine allantois

A major goal in developmental studies is to discover how embryonic cells differentiate into a vast array of cell types. In the mouse, it is thought that many critical decisions take place just before and during gastrulation,when the primary germ layers are established(Beddington, 1983).

Pre-eminent in germ layer formation is the primitive streak, a localized thickening in the midline of the epiblast(Bonnevie, 1950; Batten and Haar, 1979). Appearance of the streak defines the posterior end of the future fetus and thus, its anteroposterior axis. The streak is also thought to be where epiblast is transformed into endoderm and mesoderm, both of which are then directed to appropriate sites in the conceptus(Jolly and Ferester-Tadié,1936; Snell and Stevens,1966; Poelmann,1981; Tam and Beddington,1987; Lawson et al.,1991; Kinder et al.,1999).

Results of both tissue (Beddington,1982; Copp et al.,1986; Tam and Beddington,1987) and clonal (Lawson et al., 1991; Lawson and Pedersen, 1992) fate mapping experiments demonstrated that proximal epiblast, located at the embryonic/extraembryonic junction, ingresses into the nascent posterior streak, where it de-epithelializes to emerge as mesoderm (Batten and Haar,1979; Bellairs,1986). A large fraction of this mesoderm will be displaced into the extraembryonic region, and lines the exocoelomic cavity(Bonnevie, 1950; Lawson et al., 1991; Lawson and Pedersen, 1992). Shortly thereafter, additional extraembryonic mesoderm accumulates at the posterior angle between the amnion and yolk sac and becomes the allantoic bud(Bonnevie, 1950; Snell and Stevens, 1966)(reviewed by Downs, 1998). Thus, the posterior primitive streak is the allantois' immediate site of origin.

The allantoic bud enlarges into the exocoelomic cavity by a combination of proliferation (Ellington,1985; Downs and Bertler,2000), continuous deposition of mesoderm from the streak(Tam and Beddington, 1987; Downs and Bertler, 2000), and distal cavitation (Ellington,1985; Brown and Papaioannou,1993; Downs,2002). During enlargement, two morphologically distinct cell populations are established within the allantois(Ellington, 1985; Downs et al., 1998): an outer layer of `mesothelium', and an inner core of vascularizing mesoderm, defined initially by a plexus of endothelial cells. Formation of the endothelium occurs de novo within the allantois, beginning in its distal region with the appearance of FLK1 (KDR – Mouse Genome Informatics)-containing angioblasts (Downs et al.,1998). Specification of angioblasts and their morphogenesis into endothelial tubules then proceeds proximally to the base of the allantois,where nascent allantoic blood vessels amalgamate with those of the yolk sac and the fetus to create a vascular continuum throughout the conceptus(Downs et al., 1998). Ultimately, the allantoic vascular plexus will be remodeled into an umbilical artery and vein (Kaufman,1992).

Although the timing of appearance of the mesothelium is not known, it has been proposed, on the basis of limited light microscopic data, that distalmost allantoic cells begin to flatten at approximately the neural plate stage(Downs et al., 1998). By 4-somite pairs, the allantoic projection is enveloped in a morphologically distinct mesothelium (Downs et al.,1998), the distal portion of which contains vascular cell adhesion molecule (VCAM1) (Gurtner et al.,1995; Kwee et al.,1995; Downs et al.,2001; Downs,2002), required for chorioallantoic union(Gurtner et al., 1995; Kwee et al., 1995). Aside from VCAM1, the only other gene whose protein product has been clearly demonstrated in outer allantoic cells at all stages examined is Ahnak(Kingsley et al., 2001; Downs et al., 2002), although its significance there is not yet known.

On the basis of its ultimate morphology and function, the allantois must eventually contain at least two other cell types: vascular smooth muscle(Takahashi et al., 1996),common to all major arteries and veins, and mesenchymal cells that will provide a connective tissue matrix for the umbilical blood vessels. However,neither of these cell types has been identified in the developing allantois,nor is it known whether they differentiate from allantoic mesoderm or are contributed to the allantois by the chorion after these two tissues unite.

Intriguingly, unlike yolk sac mesoderm, whose transformation into blood islands is dependent upon contact with adjacent endoderm(Wilt, 1965; Miura and Wilt, 1970; Belaoussoff et al., 1998),allantoic mesoderm grows into the exocoelomic cavity as a physically isolated projection. It complexes with the chorion only after allantoic endothelium,mesothelium and chorio-adhesive cells have appeared(Downs and Gardner, 1995; Downs et al., 1998; Downs, 2002).

Thus, given the allantois' relative independence in the exocoelom, it was not clear how the allantois might differentiate. We envisioned at least two strategies. First, the allantois itself might contain intrinsic cues that direct its own differentiation. Second, as the posterior primitive streak is both the site of production of allantoic mesoderm and also initially continuous with it, we postulated that the streak might be integral to allantoic differentiation, specifying allantoic mesoderm as it formed therein,and/or once the allantoic bud had appeared.

Results of several studies support roles in differentiation for both of these tissues. When distal allantoic halves were placed into the exocoelomic cavity of synchronous hosts, donor allantoises united with the hosts' chorion at the appropriate time, despite lack of continuity with the proximal allantoic region and the posterior primitive streak(Downs and Gardner, 1995). Moreover, when allantoises were explanted and cultured in isolation, they invariably vascularized (Downs and Harmann, 1997; Downs et al.,1998) in a stereotypic distal-to-proximal sequence(Downs et al., 2001). Together, these observations suggested a mechanism of differentiation intrinsic to the allantois.

That the streak participates in differentiation of allantoic mesoderm was anticipated in at least two ways, each supported by experimental analysis. In the first, results of heterotopic transplantations suggested that the streak might specify proximal epiblast into particular allantoic cell types while it was being transformed into extraembryonic mesoderm. When nascent posterior mesoderm, the immediate precursor of extraembryonic mesoderm, including the allantois (Tam and Beddington,1987; Lawson et al.,1991; Kinder et al.,1999), was removed from the primitive streak and transplanted to the distal region of hosts, it exhibited limited developmental potential,failing to colonize the somites of the ectopic site(Dunwoodie and Beddington,2002). Moreover, when proximal allantoic mesoderm, having just emerged from the streak, was transplanted into a similar ectopic region(Downs and Harmann, 1997), it contributed only to blood vessels, and not to somitic mesoderm. Together,these results suggested that allantoic endothelial cells might be specified within the streak.

Results of a second set of observations suggested that the streak might influence allantoic differentiation after the bud had formed, for example,through factors emanating from a localized signaling center. Allantoic cells closest to the streak might remain initially undifferentiated, whereas those cells farther away would escape the streak's influence and differentiate[proposed by Downs and Harmann (Downs and Harmann, 1997)]. This possibility is consistent with the observation that differentiation of the allantois into endothelium and chorio-adhesive cells begins in the allantois' distal region(Downs and Harmann, 1997);although endothelium is eventually observed throughout the allantois(Downs et al., 1998), VCAM1's domain is restricted to the distal region at all time points examined(Gurtner et al., 1995; Kwee et al., 1995; Downs et al., 2001; Downs, 2002). Thus, distance from the primitive streak might dictate where chorio-adhesive cells arise and are maintained.

Results of several studies have suggested that the posterior streak is indeed a site within which decisions are made that specify cell differentiation and fate, distinguishing extraembryonic from embryonic mesoderm and the future germ line. Intriguingly, the protein products of some homeobox genes, i.e. those transcription factors critical in the regulation of cell proliferation, differentiation, migration, organogenesis and pattern formation during embryogenesis (reviewed by Deschamps and Meijlink, 1992),appear to be involved, as described below.

First, that extraembryonic and embryonic mesoderm may be distinguished in the streak was suggested by expression studies of the homeobox gene, Evx1, and by analysis of genetic knockouts of the Polycomb-group homeobox gene, embryonic ectoderm development, or Eed. Throughout gastrulation, localization of Evx1 was limited to cells near and within the streak (Dush and Martin,1992). Together with results of fate-mapping, which demonstrated that different types of mesoderm emerge from different axial levels of the streak (Tam and Beddington,1987; Lawson et al.,1991; Lawson and Pedersen,1992; Kinder et al.,1999), it was suggested that Evx1 might play a role in establishing the downstream regulatory cascade required for specification of mesodermal cell fate. In a separate study, absence of the Eed gene product led to overproduction of extraembryonic mesoderm, whereas embryonic axial and paraxial mesoderm were missing in the mutants(Schumacher et al., 1996). Together, these observations support the notion that the posterior streak is a region where mesodermal type is specified.

Second, in addition to distinguishing extraembryonic from embryonic mesoderm, the posterior primitive streak is also the site where primordial germ cells (PGCs) are allocated(Chiquoine, 1954; Ozdzenski, 1967; Copp et al., 1986; Ginsburg et al., 1990; Lawson and Hage, 1994; Lawson et al., 1999; Saitou et al., 2002). A recent model has proposed that localized cell signaling within the posterior streak distinguishes extraembryonic mesoderm from nascent germ cells(Saitou et al., 2002). The distinction is thought to be achieved through suppression of Hox gene expression in some cells, thereby allowing them to escape a somatic, or extraembryonic mesodermal, cell fate and retain pluripotency, becoming PGCs. However, the precise nature of these signals is not known.

Thus, to discriminate between roles for the allantois and the primitive streak in differentiation of allantoic mesoderm, we exploited several observations. First, microsurgical removal of allantoises during their pre-fusion phase had previously demonstrated that the streak continuously contributes mesoderm to the allantois, as reflected in the formation of allantoic regenerates over a 20-hour time period(Downs and Bertler, 2000). We reasoned that, if allantoic cell types were specified within the streak, one or more of them might be missing in the regenerates. Alternatively, if the regenerates appeared relatively normal, then we might conclude that differentiation occurred by cues intrinsic to the allantois. A combination of experimental approaches could be used to test this: morphology and the presence of FLK1 within the regenerates' core would identify endothelial cells; morphology and localization of cytokeratins to the outer surface of the regenerates might identify mesothelium; and the functional chorio-allantoic union assay (Downs and Gardner,1995) together with VCAM1 in outer distal cells of the regenerates would indicate the presence of chorio-adhesive cells(Downs et al., 1998; Downs, 2002). Second, regional specificity of VCAM1 might be established by the streak once the allantoic bud had formed. To test this, whole allantoises could be microsurgically removed and their polarity reversed within the exocoelom. Loss of contact with the streak might induce proximal mis-expression of Vcam1, thereby identifying a role for the streak in localizing chorio-adhesive cells to the distal allantoic region.

Results of this study provide evidence that specification of endothelium,mesothelium and chorio-adhesive cells does not occur within the streak. Nevertheless, they do support a role for the streak in establishment, and possibly maintenance, of the Vcam1 expression domain once the bud has formed. Further, ontogeny of mesothelium does not begin in the distal region,as previously posited (Downs et al.,1998); rather, a variety of junctional complexes are visible throughout the allantoic periphery as soon as the bud appears in the exocoelom.

On the basis of these and previous results, a model of allantoic development is proposed that involves information intrinsic to the allantois,as well as signaling from the primitive streak once the allantoic bud has formed.

Light-reversed conditions (dark period: 13.00-1.00) maintained all mice used in this study. Matings between F1 hybrids (C57Bl/6×CBA) (Jackson Laboratory) produced an F2 generation that was used as standard `wild-type'material in most experiments. For transplantation experiments, lacZ/+labeled tissue was obtained by mating F1 hybrid females with ROSA26 males of F1 genetic background and homozygous for two copies of the ubiquitously expressed lacZ transgene(Friedrich and Soriano, 1991). Dissection, staging, whole embryo culture and morphological scoring were previously described (Downs and Davies,1993; Downs and Gardner,1995; Downs and Harmann,1997; Downs et al.,1998; Downs and Bertler,2000; Downs et al.,2001; Downs,2002). Most experiments involved conceptuses encompassing all pre-fusion stages, defined from the time of the appearance of the allantoic bud through six-somite pairs, when fusion with the chorion was completed(Downs and Gardner, 1995)(approximately 7.25-8.5 days postcoitum, dpc). In all experiments, operated conceptuses were cultured alongside stage-matched unoperated ones, the latter of which verified culture conditions appropriate for gene expression,endothelialization and chorioallantoic union. Except for experiments involving AlexaFluor594-conjugated Concanavalin A, described below, no significant differences in embryo size or morphological differentiation were noted. After scoring, conceptuses were prepared for histology and immunostaining(Downs and Harmann, 1997; Downs et al., 1998; Downs et al., 2001; Downs, 2002; Downs et al., 2002), described below.

Allantoic regenerates were created throughout pre-fusion stages by aspirating whole allantoises through a mouth-held microcapillary after puncturing the anterior yolk sac (Downs and Gardner, 1995; Downs and Bertler, 2000) and culturing operated conceptuses for variable time periods. The base of the allantois had previously been defined as the point of insertion of the allantois into the yolk sac and amniotic mesoderms(Ozdzenski, 1967; Downs and Harmann, 1997). Complete removal of the allantois was verified by visual inspection before culture.

For reversed polarity experiments, whole lacZ/+ donor allantoises were introduced into wild-type host conceptuses whose own allantoises had been removed by aspiration. Where control donor allantoises were placed in normal orientation within the hosts' exocoelom (i.e. distal tips toward the chorion),their proximal ends were labeled by brief (10 seconds) immersion in DiI′/DiIC18(3)(1,1′-dioctadecyl-3,3,3′-tetramethyindocarbocyanine perchlorate,MW 933.88, Molecular Probes, Eugene, OR, USA: D-282) as previously described(Downs et al., 2001). Where donor allantoises were introduced into the hosts' exocoelom with reversed polarity (i.e. proximal ends toward the chorion), their tips, rather than their proximal ends, were labeled. Orientation of donor allantoises was monitored immediately before, during and at the end of the culture period either in the light microscope, as the dye was visible as a pink color on the surface of the cells, or in the inverted compound microscope with brief rhodamine excitation (G2-A filter cube, excitation 535/50, emission 590;Chroma, Rockingham, VT, USA); proximodistal orientation was scored with respect to the anteroposterior and left-right body coordinates. In some experiments, wild-type allantoic tissue confirmed results obtained with lacZ/+ allantoises, as X-gal could obscure the immunostain in the latter.

Allantoic subregions were excised from whole allantoises at all prefusion stages with the aid of glass scalpels(Beddington, 1987) and either introduced into the exocoelom of host conceptuses whose own allantois had been removed, or cultured in isolation either directly onto tissue culture plastic or free-floating, as previously described(Downs and Harmann, 1997; Downs et al., 1998; Downs et al., 2001).

X-gal staining identified lacZ/+ tissue after fixation in 4%paraformaldehyde, as previously described(Downs and Harmann, 1997; Downs et al., 1998; Downs et al., 2001; Downs, 2002), with the exception that exposure to X-gal was for 2 or 6 hours. Bouin's fluid was used to fix all conceptuses that did not involve X-gal staining. All fixed material was embedded in paraffin wax and sectioned at a thickness of 6 μm. All intact allantoic profiles presented in this study are sagittal views.

Antibodies against FLK1, VCAM1, BMP4 (Santa Cruz Biotechnology, Santa Cruz,CA, USA) and cytokeratins (DAKO Corporation, Carpinteria, CA, USA) were used in indirect immunohistochemistry. As the literature provided no guidance concerning expression of specific cytokeratins in the allantois, antibodies against a wide spectrum were used. Antibody binding was performed for variable times at room temperature or 4°C, and ready-to-use reagents detected the antibody-antigen complexes (streptavidin-horseradish peroxidase, Vector Laboratories, Burlingame, CA, USA; diaminobenzoate (DAB), DAKO). Pre-binding control peptides to anti-FLK1 and anti-VCAM1 served as negative controls and were previously described (Downs et al.,1998; Downs et al.,2001; Downs,2002); increasing concentrations of control peptide to anti-BMP4(Santa Cruz Biotechnology), which was raised against the N-terminal region of the protein, confirmed antibody specificity. Histological sections were also incubated in the absence of antibody in all immunolocalization experiments,including those for cytokeratins for which no control peptide was commercially available. However, antibodies against cytokeratins strongly reacted against trophoblast giant cells, in agreement with previous findings(Hashido et al., 1991; Jaquemar et al., 2003),thereby providing an internal control for antibody specificity. X-gal-stained material was counterstained in nuclear fast red; most other material was stained either in hematoxylin/eosin (H/E) or hematoxylin alone.

Lectin-conjugated compounds have previously been used to fate-map cells which border a cavity (Tam and Beddington,1987). After a brief microspin to remove particulate,AlexaFluor594-conjugated Concanavalin A (absorbance, 590 nm; emission, 617 nm;2.5-5 mg/ml sterile PBS; Molecular Probes) was gently blown into the exocoelomic cavity of pre-fusion conceptuses via a mouth-held microcapillary in eight separate experiments. After 1 minute of exposure, labeled exocoeloms were repeatedly rinsed with the aid of a mouth-held microcapillary for 3-5 minutes, after which allantoises were removed. ConA labeling of both the mesoderm lining the exocoelomic cavity and the entire surface of isolated allantoises was confirmed (G2-A filter cube, described above, mercury lamp had logged 0-100 hours). After 8-12 or 20-24 hours of culture, allantoic regenerates and unoperated labeled and unlabeled control allantoises were examined by peeling away yolk sacs, cutting out allantoises with a glass scalpel, and viewing them, the yolk sacs and the amnions in both whole-mount and squashed preparations under fluorescence.

To confirm that labeled cells both proliferated and contributed label to their progeny, ConA-labeled headfold- and early somite-stage yolk sacs were isolated, rinsed and treated with a mixture of trypsin and pancreatin(Downs and Harmann, 1997) for 2 minutes on ice, after which the enzyme was inactivated in dissection medium. The mesodermal and endodermal cell layers were then separated via aspiration through a small diameter microcapillary(Beddington, 1987). The cell layers were placed separately into sterile Eppendorf tubes, and further enzyme-treated (15 minutes on ice) to obtain single cell suspensions, which were briefly spun down, resuspended in dissection medium, and plated onto 8-well chamber slides (Nunc Permanox; Nunc, Naperville, IL, USA) containing 0.4 ml culture medium. After 5 hours, floating cells were removed, the adherent cells were fed, counted and cultured for 20 hours more, at which time the medium was again replaced, and the adherent labeled cells counted. In addition, we verified that AlexaFluor ConA did not inhibit cell motility by plating labeled aspirated allantoises (two experiments, n=10)directly onto tissue culture plastic(Downs et al., 2001) and demonstrating the presence of labeled mesothelial outgrowth at 8 and subsequently 20 hours.

In eight experiments, all 21 unlabeled, unoperated conceptuses were normal at the end of the culture period and negative for ConA. Of 27 control conceptuses whose allantois had been labeled but not removed from the exocoelom, one fetus (3.7%) was abnormal in that the ConA-labeled allantois had fused with the yolk sac rather than with the chorion. Of the 72 conceptuses whose allantoises had been labeled, removed and subsequently regenerated, none were abnormal, although one conceptus (1.4%) was dead; a subset of these (n=22) was measured in the dissection microscope (see next section) and their average length compared with that of unlabeled regenerated allantoises (n=6). In a final set of control experiments,allantoises were removed prior to labeling exocoelomic cavities (two experiments, eight conceptuses, two fetuses of which were abnormal after culture) to ascertain the probable labeling pattern if yolk sac/amniotic mesodermal cells crawled over the primitive streak during regeneration, as well as to verify persistence of label in the regenerates.

The length of some 20- and 6-hour allantoic regenerates and unoperated allantoises was measured in the dissection microscope immediately after culture, by means of an eyepiece reticule, and plotted(Fig. 2A,B) or reported in Results. The average length and standard error of the mean (s.e.m.) of histologically prepared and sectioned allantoises and their VCAM1-negative regions were obtained from the longest three sagittally oriented sections after photography, computer scanning (Adobe Photoshop) and printing(Fig. 4).

Conceptuses were dissected and staged as described above. Embryos at the neural plate (early and late bud), headfold (early and late), and 4-somite pair stages were immersion fixed in Karnovsky's fixative (2% paraformaldehyde,2.5% glutaraldehyde in 0.1 M NaPO₄ buffer, pH 7.4) for 2 hours at 4°C, after which they were washed in rinsing buffer (0.1 M NaPO₄) and postfixed in 2% osmium tetroxide (OsO₄)buffered in 0.1 M NaPO₄ for 1 hour at room temperature. Postfixed samples were washed in rinsing buffer and dehydrated in a graded series of ethanols for 7-12 minutes at room temperature, with a final dehydration in propylene oxide as a transition solvent. Dehydrated samples were infiltrated between 20-65°C in a 1:1 mixture of Embed-812 and Spurr's Low Viscosity embedding media (Electron Microscopy Sciences, Fort Washington, PA, USA) that included propylene oxide. Semi-thin (3 μm) serial sections of sagittally oriented allantoises were collected and dried on slides. Target sections were selected and remounted on polymerized resin Beem capsule stubs in preparation for thin sectioning. Ultrathin sections (70 nm) were collected on pioloform-coated 2×1 mm aperture copper electron microscope grids(Electron Microscopy Services, Fort Washington, PA, USA). Ultrathin sectioning(100-500 nm) was performed on a Reichert-Jung Ultracut E ultramicrotome. The sections were post-stained in uranyl acetate and lead citrate. The samples were viewed and documented on a Philips CM120 at 80 kV.

The Student's two-way t-test (Mini-Tab, equal variances assumed,Confidence Interval=95%) determined significant differences(P<0.05) between treatment categories wherever relevant in this study.

Results of earlier studies suggested that allantoises microsurgically removed from the conceptus prior to their union with the chorion would regenerate (Downs and Gardner,1995). Later, allantoic regeneration was exploited to determine when and how much mesoderm was added to the allantois by the streak(Downs and Bertler, 2000). Findings revealed streak activity at all pre-fusion stages, although mesoderm contribution decreased with increasing developmental age. On the basis of these observations, we hypothesized that allantoic regenerates would be a powerful experimental means by which to investigate the effect of streak activity on differentiation of allantoic mesoderm. The rationale was that, if any of the regenerates lacked certain cell populations, then specification of the missing cell type(s) might have occurred within the streak at or slightly later than the time of allantoic removal.

To ensure that regenerates formed at all pre-fusion stages, whole allantoises were removed, exposing the subjacent posterior primitive streak(Fig. 1A-E). After 20 hours in culture, operated conceptuses were examined in the dissection microscope and regenerates were measured. Regeneration was found at all developmental stages,but only a few regenerates grew far enough to unite with the chorion(Fig. 2A). The average length of `unfused' regenerates was remarkably similar between the neural plate and 5-somite pair stages (501.5±21.7 μm, n=54), but significantly shorter (P<0.000) than unoperated controls(924.7±17.6 μm, n=44). The average length of regenerates formed from the 6-somite pair stage was significantly shorter (244.6±66μm, n=6) than the other regenerates (P=0.002, Student's two-way t-test with 5-somite pair regenerates). Thus, on the basis of these findings, we conclude that, during pre-chorionic fusion stages, the primitive streak is active at all time points, producing regenerated allantoises.

To establish whether endothelium, mesothelium and chorio-adhesive cells were present in all of the 20-hour regenerates, histological and immunohistochemical analyses were performed. All regenerates were completely enveloped by mesothelium, and all contained evidence of a vascularizing core that extended down their length (Fig. 1F), in accord with the morphology of normal ex vivo counterparts at equivalent developmental ages (≥7-somite pairs)(Downs et al., 1998).

FLK1 was found in the endothelial cells of 20-hour regenerates, and never in the mesothelium (Fig. 1G)(Downs et al., 1998). Similarly, VCAM1 was appropriately localized to the distal two-thirds of all allantoic regenerates, which included the mesothelium and some core cells(Fig. 1H)(Downs et al., 2001; Downs, 2002). Double immunohistochemistry revealed normal cell-cell relationships in the regenerates: VCAM1-positive cells were associated with, but distinct from, the FLK1-positive core vasculature, and absent from the proximal allantoic third(Fig. 1I-L)(Downs, 2002).

Of the 77 allantoic regenerates of Fig. 2A, a small number (22%) had undergone union with the chorion. In these, microscopic inspection revealed that the early steps of placentation were occurring: the chorio-allantoic fusion surfaces appeared to be breaking down, and the FLK1-positive allantoic vasculature was penetrating chorionic ectoderm (data not shown), as previously described(Downs, 2002). Of the remaining regenerates, none grew far enough to unite with the chorion. To discover whether these were nonetheless functionally normal, lacZ/+unfused regenerates were aspirated at 20 hours, placed into the exocoelom of pre-fusion wild-type host conceptuses whose own allantoises had been microsurgically removed, and the operated hosts were cultured to just beyond 6-7-somite pairs, when all allantoises have normally united with the chorion(Downs and Gardner, 1995; Downs, 2002). In two experiments, 4/4 grafted X-gal-stained regenerated allantoises exhibited convincing union with the hosts' chorion (data not shown), suggesting functional adhesive capabilities.

Why some regenerates grew far enough to unite with the chorion could be explained by prolonged collapse of the exocoelomic cavity, possibly because the yolk sac puncture did not heal in a timely manner. As a consequence of reduced exocoelomic volume, they could have been brought into closer proximity to the chorion, thereby initiating union.

Thus, not only do regenerates demonstrate appropriate morphological and molecular properties, but they are also functionally competent, capable of uniting with the chorion if given the opportunity.

The normal morphological, molecular and functional profiles of 20-hour regenerates suggested that the primitive streak does not specify allantoic mesoderm prior to its becoming the allantoic bud. Although outer cell morphology and expression of Vcam1 alluded to the presence of mesothelium in the regenerates, flattening of distal outer cells cited as a provisional criterion for the onset of formation of mesothelium at the neural plate stage (Downs et al.,1998), was subjective and non-rigorous. Thus, electron microscopic analysis was used here to determine when and where junctional contacts,indicative of an epithelium, were present on the outer surface of the allantois.

The number of contacts between outer cells was scored in ultrathin sections of allantoises spanning the neural plate/early bud through 4-somite stages(Table 1). Three types of electron dense structures, including spot desmosomes, adhesive plaques and presumptive tight junctions (Batten and Haar, 1979), were identified at sites of contact at all stages(Fig. 3A-D), with neither the distal nor proximal halves favored (Table 1). Moreover, these structures were not always found in flattened cells. Finally, at the neural plate and headfold stages, the only junctional contacts observed in cells of the core were with outer cells(Fig. 3A), suggesting that the otherwise junction-free nascent core mesodermal cells shared ultrastructural similarities with migrating mesoderm rather than with epithelium(Batten and Haar, 1979).

Thus, the outer layer of allantoic mesoderm epithelializes as soon as the bud appears in the exocoelomic cavity. To obviate dependence upon the electron microscope for detecting mesothelium in short-term regenerates (next section),immunohistochemistry was performed against cytokeratins, found in many polarized cells. Cytokeratins were observed in the outer cells of the allantois as early as neural plate stages(Fig. 3E,F). However, by the headfold stage, cytokeratins were no longer confined to outer cells, but were also found in core allantoic cells (Fig. 3G), possibly in angioblasts undergoing endothelialization into nascent blood vessels.

To verify that differentiation took place with normal kinetics in allantoises post-microsurgical removal, 6-hour regenerates were created and analyzed for morphology and gene expression. As a prelude to this analysis, we first established the distance from the primitive streak at which VCAM1 appeared in ex vivo conceptuses (Fig. 4). VCAM1 was visible in distal mesothelial cells only when allantoises were at least 220.7 μm (±5.5 μm) long (data not shown; 1-somite pair; n=4). At 1-somite pair and all stages thereafter, the average length of the VCAM1-negative region was fixed at approximately 220.2 μm (±5.9 μm, n=34)(Fig. 4). This length did not differ significantly in conceptuses cultured for 20 hours (193.1±16.5μm, n=9; P=0.08; data not shown), confirming that culture conditions did not alter correct topographical expression of Vcam1. The length of the VCAM1-negative region in fused regenerates was 192.4 μm(±12.8 μm; n=12; data not shown), similar to the unoperated cultured controls, whereas in unfused regenerates it was 149.5 μm(±8.0 μm; n=14; data not shown), significantly shorter than the controls (P=0.01). This difference may be attributable to the fact that once allantoises are anchored to the chorion, they may `stretch'during subsequent enlargement of the exocoelomic cavity and lengthen; in the absence of anchoring to the chorion, `stretching' would not occur in the unfused regenerates. Thus, after long-term culture, unfused regenerates might ultimately appear slightly collapsed and shorter.

In the next set of experiments, 6-hour regenerates were created(Fig. 2B, Fig. 5). Their average length before fixation was 177.6±19.2 μm (n=23; Fig. 2B). As we found that Bouin's fluid and subsequent histological processing resulted in tissue shrinkage of 14.8% ±2.2% (n=14 fresh and subsequently fixed specimens, neural plate – 5-somite pair stages; data not shown), gene expression in the 6-hour regenerates was predicted to resemble that found in fixed ex vivo early headfold stages (Fig. 4).

Thus, after histological processing, several subsets of these 6-hour regenerates were then examined for gene expression and morphology. The average length of the first set of regenerates after fixation was 116.4 μm(±10.6 μm; 18 sections; n=6 specimens), with the longest one measuring 210.7±3.5 μm and the shortest one 88 μm. Although all of the unoperated controls contained distal VCAM1, in accord with their average length of 312.0±13.7 μm(Fig. 5C) (1-5-somite pairs, 13 sections; n=5 specimens, data not shown), none of the regenerates contained this protein (Fig. 5A,B). Thus, these results supported a minimal length requirement of 220 μm for Vcam1 expression.

The presence of mesothelium was then confirmed in a second set of 6-hour regenerates by antibody staining against cytokeratins(Fig. 5D). As the average length of fixed 6-hour regenerates was similar to normal headfold-stage allantoises, described above, it was therefore not unexpected that cytokeratins were found in the regenerates' outer and inner cell populations,as reported in the previous section for expression of cytokeratins in ex vivo headfold-stage allantoises (Fig. 3G).

FLK1, which was previously observed in the distal core allantoic region between the late bud/neural plate and 2-somite pair stages(Downs et al., 1998), also showed distal staining in 6-hour regenerates created between the neural plate-and 2-somite pair stages; the primitive streak was also appropriately negative(Fig. 5E,F)(Downs et al., 1998). Not surprisingly, those regenerates created after 2-somite pairs did not exhibit regional polarity of FLK1 (Fig. 5G,H), as it was known that, by the time of microsurgery, the primitive streak region had become FLK1-positive, thereby making distal polarity of Flk1 expression in these later regenerates impossible to discern (Downs et al.,1998).

Thus, taken together, our results revealed that allantoic regenerates behave like intact allantoises, no matter when they are formed: mesothelium was present, VCAM1 identified the chorio-adhesive cell type at a fixed and reproducible distance from the primitive streak, and endothelial cell formation began in the distal region during regeneration. On the basis of these observations, we conclude that specification of mesothelial,chorio-adhesive and endothelial cells does not occur within the primitive streak.

Although the aforementioned observations provide compelling evidence that information intrinsic to the allantoic bud is responsible for re-establishing appropriate differentiation in the regenerates, an equally plausible mechanism could involve the adjacent yolk sac and/or amnion. Specifically, we hypothesized that, after removal of the allantois, the exposed primitive streak `heals' by contribution of mesoderm from the yolk sac and amnion. The allantoic regenerate would then grow into and become coated in cells of yolk sac and/or amniotic origin (Fig. 6), which would then reorganize allantoic development. The molecular cue responsible could be the gene product of BMP4 as, in its absence, allantoises do not form or they are highly reduced in size and abnormal in morphology (Winnier et al.,1995; Lawson et al.,1999; Fujiwara et al.,2001). Moreover, both the yolk sac and amnion contain BMP4(Lawson et al., 1999).

To investigate whether yolk sac and/or amniotic mesoderm and their associated BMP4 were involved in regeneration of allantoises, we first confirmed the presence of BMP4 in yolk sac and amniotic mesoderm at all time points (data not shown), in accord with previously defined BMP4 promoter-driven lacZ expression(Lawson et al., 1999). In the allantois, BMP4 was observed in many outer cells at neural plate stages(Fig. 7A, and data not shown),whereas the core was negative (data not shown). By the headfold stage, BMP4 was present in most cells throughout the distal allantoic third, as well as in much of the entire allantoic periphery to a depth of 1-3 cells(Fig. 7B, and data not shown from transverse sections of the allantois). This pattern remained relatively unchanged (Fig. 7B-E, and data not shown) until 6-somite pairs, at which time BMP4 was present throughout the allantois, and included both putative mesothelial and nascent endothelial cell populations (Fig. 7F, and data not shown), in agreement with an anecdotal in situ mRNA analysis of the allantois at approximately the same stage(Mahlapuu et al., 2001). Between 7- and 16-somite pairs, these global patterns were little changed, and included robust levels of BMP4 at the chorio-allantoic fusion junction(Fig. 7G,H). Twenty-hour regenerates exhibited widespread BMP4 protein similar to that found in intact allantoises at ≥7-somite pairs (Fig. 7I).

To discover whether extrinsic BMP4, i.e. that from the yolk sac and/or amniotic mesoderm, was involved in restoring normal differentiation to the allantoic regenerates, we examined allantoises after regeneration from the headfold stage at 1-1.5-hour intervals between 0 and 6 hours(Fig. 7J-M). Although no evidence for BMP4 was found in the exposed primitive streak at time 0 hours(Fig. 7J), or in regenerating allantoises at 1.5 hours (Fig. 7K), BMP4 was clearly visible in peripheral and distal core cells of the regenerates at 3, 4 and 6 hours(Fig. 7L,M, and data not shown), thus resembling normal expression patterns. However, these results did not distinguish between BMP4 of extrinsic origin from BMP4 produced within the allantois itself.

To address this, the adjacent yolk sac and amniotic mesoderms were fate-mapped by labeling the mesodermal lining of the exocoelomic cavity with AlexaFluor-conjugated Concanavalin A, which selectively bindsα-mannopyranosyl and α-glycopyranosyl residues of glyco-conjugates on the cell surface. First, to verify that ConA-labeled cells proliferated and passed their label on to progeny cells, single cell suspensions of the mesodermal and endodermal components of one of the tissues, the yolk sac, were plated onto tissue culture plastic (see Materials and methods). After 20 hours in culture, endodermal cells remained floating, and did not adhere to the plastic, in agreement with findings by others (R. Gardner, personal communication). In contrast, both labeled and unlabeled yolk sac mesodermal cells adhered to the plastic and had a vacuolated appearance(Fig. 8A-D). In three experiments, AlexaFluor ConA-labeled cells doubled in number over the 20-hour culture period (Table 2). We also confirmed that ConA-labeling did not inhibit mesothelial cell motility,by plating labeled allantoises (Downs et al., 2001) (Fig. 8E) and observing outgrowth at 8(Fig. 8F) and 20 hours thereafter (data not shown). Moreover, the explants looked normal, in that they were well vascularized by 20 hours and morphologically similar to unlabeled controls, as previously described(Downs et al., 2001). Finally,use of ConA in these circumstances further revealed that allantoic outgrowth was mesothelial in origin.

After labeling exocoelomic mesoderm(Fig. 9A,B), allantoises were then removed and examined to verify that labeling was complete(Fig. 9C). Operated conceptuses were cultured for 8-12 or 20-24 hours alongside controls(Table 3). By 20 hours, most of the fluorescent label was localized to the distal region of intact labeled allantoises (Fig. 9D,E), in agreement with fate mapping studies which suggested that nascent allantoic mesoderm is displaced distally as fresh mesoderm is added by the streak(Lawson et al., 1991; Lawson and Pedersen, 1992; Downs and Harmann, 1997; Kinder et al., 1999). Although it might be argued that proximal cells died in the presence of ConA, we consider this unlikely, as the allantoic explants of Fig. 8E,F were covered in labeled mesothelium at the end of the explant period.

In all six control conceptuses in which allantoises had been removed prior to labeling, label persisted in experimental regenerates, and was found laterally, rather than distally (Fig. 9F,G). Thus, we concluded that, if yolk sac and/or amniotic mesoderm crawled over the primitive streak, labeled cells might be expected in a lateral line, spanning the proximal to distal region of the regenerate. Of 72 experimental regenerates (Table 3), 23 were not scorable (see legend to Table 3). Of the remaining 49 regenerates, 28 were completely negative for AlexaFluor ConA (data not shown),whereas the other 21 regenerates contained just a few (2-10) positive cells(e.g. Fig. 9H,I) whose localization, though lateral, was not spatially consistent amongst the regenerates. Where regenerates were ≥266.8 μm long, all were well-vascularized and exhibited an outer layer of mesothelium (data not shown); where they were ≤200.1 μm long, all regenerates bore an outer layer of mesothelium, but they may or may not have contained an obvious vascular plexus (data not shown). The length of a small set of fresh 20-hour regenerates (403±47 μm; n=22) whose exocoelomic cavities had been ConA-labeled was compared with that of unlabeled regenerates(500±54 μm; n=6) from the same experiments (n=3),and found not to be significantly different (P=0.31). Moreover, four regenerates fused with the chorion (Table 3), confirming their functional competence despite the presence of ConA.

On the basis of these findings, we conclude that labeling the exocoelomic cavity of headfold- and early somite stage conceptuses with ConA has no detrimental affect on embryonic growth, embryonic differentiation or on allantoic regeneration. Although a very small number of yolk sac/amniotic cells may crawl over the exposed primitive streak and become incorporated into the lateral surface of the regenerates, failure to identify them in the majority of regenerates strongly suggests that differentiation in allantoic regenerates is not re-specified by a healing process that involves the adjacent yolk sac, amnion or BMP4 contained therein. Moreover, these results accord with previous ones, which investigated the fate of presumptive PGCs thought to reside in the base of the allantois(Ozdzenski, 1967); results of those studies demonstrated that cell movement at the posterior embryonic/extraembryonic junction is typically away from the streak, rather than toward it, the consequence of which is that re-entry of allantoic primordial germ cells into the fetus probably does not occur(Downs and Harmann, 1997; Anderson et al., 1999).

As described above, the 220 μm-long VCAM1-negative region remained fixed throughout allantoic development, suggesting either that the allantois acquires intrinsic information that limits the domain of VCAM1, or that the streak suppresses VCAM1 in the proximal allantoic region closest to it.

To investigate a role for the streak in suppressing VCAM1, whole donor allantoises were removed and placed into the exocoelomic cavity of hosts,either in normal or reversed orientation(Fig. 10), and cultured for 8 hours. We reasoned that, if the streak suppresses gene expression or translation of VCAM1 transcripts in cells closest to it, then the proximal region would exhibit VCAM1 upon release from contact with the streak.

Of 14 normally oriented grafts, 57% had fused with the chorion, 29% were lying flat across the chorion, 7% were free-floating in the exocoelom, and 7%had become wedged inside the yolk sac. Whatever the status of the allantoises,however, all 14 contained an uninterrupted layer of mesothelium and all displayed the VCAM1-negative proximal region, whereas the distal region was VCAM1-positive (Fig. 10A,C,D). Of the 28 reversed-polarity grafts, 50% exhibited proximal regions in contact with the chorion, 43% were lying across the chorion, and 7% were free-floating in the exocoelom. Despite their varied orientation, all looked identical to the intact controls, in that all contained mesothelium, and all proximal regions were negative for VCAM1 relative to the distal third(Fig. 10B). A separate set of experiments confirmed FLK1 throughout the length of the allantois, in accord with normal profiles in intact allantoises(Fig. 10H).

Allantoises were then removed and, based on measurements of the VCAM1-negative region (Fig. 4),cut into distal and proximal subregions. To ensure that the proximal regions fell to within at least one standard deviation of the average 220 μm VCAM1-negative region (data not shown), these were never longer than 150μm. Proximal and distal allantoic subregions were then placed separately into the exocoelomic cavity of individual hosts, and cultured for either 8 or 24 hours.

All 8-hour proximal regions were negative for VCAM1(Fig. 10E). Seven out of 24 proximal regions were free-floating in the exocoelom and 3/24 were tethered to the host's regenerated allantois (data not shown); all had formed a complete outer rind of mesothelium (Fig. 10E). Fourteen out of 24 proximal regions had undergone union with the chorion (data not shown), exhibiting mesothelium at the unfused edge and,where they had made contact with the chorion, cells appeared to be breaking down (Downs, 2002). Of eight control distal regions, all were appropriately VCAM1-positive(Fig. 10F), with levels of VCAM1 approximating those of the intact controls (compare Fig. 10F with 10G). Both 8-hour distal and proximal regions were positive for FLK1(Fig. 10I,J).

By 24 hours, 21/21 distal regions maintained VCAM1, with levels similar to intact allantoises (Fig. 11A). However, in contrast to the 8-hour explants, all 21/21 proximal regions had become VCAM1-positive throughout the mesothelium and core(Fig. 11B,C). One proximal region was tethered to the tip of the host's allantoic regenerate(Fig. 11B), and one had fused with both the chorion and the yolk sac(Fig. 11C). All the others had united with the chorion alone and behaved like distal subregions, with mesothelium breaking down at the fusion junction (data not shown). FLK1 localization was unchanged, being present in both the proximal and distal regions (data not shown). Further support for these observations was obtained by culturing subregions in suspension outside of the conceptus. Ten out of 10 distal and 10/10 proximal regions exhibited VCAM1(Fig. 11D,E), whereas all 4/4 whole control allantoises were VCAM1-positive in both the outer mesothelial circumference and the core (Fig. 11F).

These findings suggest that a signal(s) emanating from the primitive streak or the allantoic mid-region suppresses proximal VCAM1 during normal allantoic development, but has no effect on formation of mesothelium or expression of Flk1. Once this putative factor(s) becomes inactive or depleted between 8 and 24 hours in the proximal isolates, VCAM1 is either induced or derepressed.

To distinguish between signaling from the primitive streak and/or the allantoic midregion in inhibiting proximal VCAM1, distal allantoic thirds were excised from whole allantoises, maintaining continuity between the mid- and proximal regions (`mid/proximal'). In the same experiments, whole allantoises,distal and mid/proximal regions were cultured individually within the exocoelomic cavity of host conceptuses. As expected, distal thirds were always strongly VCAM1-positive (Fig. 11G). Moreover, not only was VCAM1 found throughout explanted whole allantoises (Fig. 11I),but it was also found throughout the mid/proximal subregions, although only two of these had remained free-floating in the exocoelom(Fig. 11H). Failure of the allantoic midregion to suppress proximal VCAM1 was confirmed in three additional experiments of isolated explants cultured in suspension for the same time period (data not shown).

On the basis of these observations, we conclude that the primitive streak limits the expression domain of VCAM1, confining the chorio-adhesive cell type to the distal allantoic region. In contrast, the streak is neutral with regard to formation of the mesothelium and endothelial-based vasculature.

The goal of this study was to investigate the mechanism(s) by which the murine allantois differentiates into its major cell types. On the basis of findings reported here, we conclude that both the allantois itself and the primitive streak cooperate in normal allantoic development. Once the allantoic bud has formed, intrinsic factors appear to play a major role in specification of its endothelium, mesothelium and chorio-adhesive cells, whereas the posterior primitive streak may act to restrict the chorio-adhesive cells to the distal allantoic region by suppressing VCAM1 in proximal allantoic mesoderm.

That the allantois contains intrinsic factors responsible for its own differentiation was demonstrated by studying its regeneration. A combination of morphological and immunohistochemical approaches were used to demonstrate the normal kinetics of allantoic regeneration, including the invariable presence of mesothelium as defined by cytokeratins, a FLK1-positive endothelial plexus, and a defined VCAM1-positive distal region, indicative of the presence of chorio-adhesive cells. In addition, despite their smaller size, 20-hour regenerates were found to be competent to unite with the chorion. Repair of the exposed primitive streak from the adjacent yolk sac and amnion, possibly stimulated by BMP4 contained therein, was not consistently involved in allantoic outgrowth or in the re-establishment of proper morphological and molecular patterning in the regenerates. However, we cannot discount the possibility that BMP4, which was found in the normal allantoic bud and all of the regenerates (Fig. 7), may originate within the allantois and thereby contribute to its differentiation. Moreover, although cell age appears to play a major role in the timing of differentiation once the bud has formed, as discussed in the model below, the period during which proximal epiblast is converted into mesoderm within the streak plays no discernible role in specification of allantoic cell types.

These observations strongly suggest that, as the allantois becomes visible in the exocoelom, it acquires information that will direct its own differentiation.

Although the streak does not specify the chorio-adhesive cells, it does appear to be involved in restricting their domain within the allantois by suppressing expression of Vcam1 in the proximal allantoic region.

The base of the allantois, which corresponds with the posteriormost level of the streak, has previously been defined as the point of insertion of the allantois into the yolk sac and amniotic mesoderm(Ozdzenski, 1967) (reviewed by Downs, 1998). This definition has guided the present and all previous studies involving microsurgical removal of the allantois (Downs and Harmann, 1997; Downs et al.,1998; Downs and Bertler,2000; Downs et al.,2001; Downs,2002). Accordingly, measurements on histologically prepared allantoises revealed that VCAM1 was consistently localized to the distal region, beginning at a distance of 220 μm from the posterior limit of the primitive streak (Fig. 4). In isolation, proximal regions of allantoises were negative during the first 8 hours of culture, but by 24 hours they contained VCAM1 and behaved like distal tips in their ability to unite with the chorion(Fig. 11). Co-culture experiments revealed that suppression of Vcam1 expression in the proximal region of the allantois was not because of its being in contact with its mid-region (Fig. 11).

On the basis of these data, we conclude that, although the primitive streak does not specify chorio-adhesive cells during formation of allantoic mesoderm,it does suppress VCAM1, either at the level of transcription or translation,in the proximal allantoic region once the bud has formed. Whether the mechanism of suppression involves direct cell contact and/or localized signaling from the streak remains to be determined.

On the basis of limited light microscopic data, we had previously supposed that formation of mesothelium began in the distal allantoic region(Downs et al., 1998). Thus, an unexpected finding in this study was the presence of this layer as soon as the bud appeared (Fig. 3). Moreover, mesothelium was found in all of the regenerates, as well as in explanted whole allantoises and allantoic subregions. These observations argue that mesothelium forms as a result of cues intrinsic to the allantois.

What those cues are is unknown, although it is tempting to speculate that they involve cells being on an outer surface. In other words, differentiation of mesothelium may share similarities with that of trophectoderm of the blastocyst, where differentiation of blastomeres along this lineage seems to depend upon being located externally(Tarkowski and Wroblewska,1967). Thus, mesothelium of the allantois may differentiate by an inside-outside mechanism, as a result of formation of asymmetrical cell contacts between outer and inner cells. Indeed, in our limited study, outer and immediately subjacent inner cells were linked by junctional structures(Fig. 3A) that were not found anywhere else in the allantoic core during early stages. Intriguingly, the giant phosphoprotein, Ahnak, which is found near desmosomal contacts on epithelial cells (Hashimoto et al.,1995; Masunaga et al.,1995), was recently described on the allantoic surface(Kingsley et al., 2001). Ahnak was present there as early as the neural plate stage(Downs et al., 2002). Like uvomorulin which, along with one of its key modulators, protein kinase C(Winkel et al., 1990; O'Sullivan et al., 1993), has been implicated in blastomere polarization(Fleming et al., 1994), Ahnak is under the control of protein kinase B/C(Hashimoto et al., 1995; Sussman et al., 2001), and may thus localize to sites of cell-cell contact as outer cells epithelialize into mesothelium.

On the basis of these and previous findings, we propose a model of differentiation of allantoic mesoderm(Fig. 12). All of the data point to roles for intrinsic cues that may control allantoic cell age, as well as roles for cell position, and signaling from the posterior primitive streak.

In the first step of allantoic development, nascent mesoderm exits the posterior primitive streak and is deposited into the exocoelom as the allantoic bud (Beddington,1982; Beddington,1983; Copp et al.,1986; Tam and Beddington,1987; Kinder et al.,1999) (Fig. 12A,light pink). Outer allantoic cells acquire junctional complexes, possibly based on their exterior position, forming mesothelium(Fig. 3, Fig. 12A, light blue circles). In addition, allantoic mesoderm acquires intrinsic factors, possibly in the form of homeobox-encoded proteins, which may then trigger a cascade of downstream events controlling place- and time-dependent expression of morphoregulatory genes (Jones et al.,1992). Although systematic examination has been limited,expression of Hox genes encompasses several scenarios: it may begin in the allantois and spread to the streak (e.g. Hoxb8/Hox2.4)(Deschamps and Wijgerde, 1993; Deschamps et al., 1999; Forlani et al., 2003), it may begin in the streak and spread to the allantois (e.g. Hoxa3/Hox1.5)(Gaunt, 1988), or it may span both tissues (e.g. Hoxc8/Hox3.1)(Gaunt, 1988; Bieberich et al., 1990). Also,in some cases, Hox genes may be expressed only in the streak where they define an apparent border with the allantois (e.g. Evx1)(Dush and Martin, 1992; Crossley and Martin, 1995). Thus, the presence of a large number of Hox family members in both the allantois and primitive streak at critical times in differentiation of allantoic mesoderm make them prime suspects in regulation of the genetic and molecular regulatory pathways that control differentiation into the umbilical cord and chorio-allantoic placentation.

Twelve hours after the bud appears, FLK1-positive angioblasts differentiate in the distal allantoic region, where mesoderm is oldest(Downs and Harmann, 1997; Kinder et al., 1999)(Fig. 12B, dark pink color). Several hours later, distal angioblasts coalesce into identifiable endothelial tubules (Downs et al., 1998)(Fig. 12E,F). The sequence of angioblast differentiation and endothelial tube formation then proceeds down the length of the allantois with striking regularity(Fig. 12E,F)(Downs et al., 1998), strongly suggesting that cell age plays a major role in this event. The importance of cell age is further underscored by uniform cell proliferation along the allantoic projection (Downs and Bertler,2000); constant cell turnover could ensure that differentiation of each allantoic region takes place in turn. Moreover, whole explanted headfold-stage allantoises invariably vascularized with distal-to-proximal polarity according to a precise temporal program(Downs et al., 2001), and proximal allantoic subregions, which are initially relatively negative for FLK1 (Downs et al., 1998),also acquired abundant FLK1 in isolation(Fig. 10J) (this study).

Thus, a series of downstream genetic events would be activated until a final gene product triggers the expression of Flk1 in the distal allantoic region, where mesoderm is oldest. Consequently, cell age, possibly the result of an internal timing mechanism initiated by Hox gene products in the bud (Fig. 12A), and cell position, ultimately coincide. An example of a candidate regulatory Hox gene is Hoxb5, whose product has been implicated in controlling expression of Flk-1(Wu et al., 2003), although its localization to the allantois has not yet been reported.

By 8.0 dpc, FLK1 is found in the distal two-thirds of the allantois(Downs et al., 1998)(Fig. 12E,F). Together with cell proliferation and cavitation, sustained activity of the streak results in lengthening of the allantois. Then, at a distance of approximately 220 μm from the streak, VCAM1 becomes visible in distalmost mesothelial cells(Fig. 12C, light green color). Thus, chorio-adhesive cells have formed.

Results of this and previous studies suggest a model of chorio-adhesive cell formation that, like endothelium, involves both cues intrinsic to the allantois and an internal timing mechanism (reviewed by Downs, 1998). In addition,signaling from the primitive streak appears to be involved (Figs 10, 11). Saitou et al.(Saitou et al., 2002) have suggested the presence of a localized signaling center within the posterior streak. Factors emanating from the posterior streak could be extended to the allantois, inhibiting expression of Vcam1 in proximal mesoderm. A downstream target might be the homeobox-containing forkhead transcription factor, Foxf1, as in Foxf1 mutants Vcam1's expression domain is expanded to include the proximal allantoic region(Mahlapuu et al., 2001).

In this step, we propose that factors emanating from the streak continue to suppress VCAM1 to a distance of 220 μm. Endothelium has spread all the way down to the base of the allantois (Fig. 12D-F), at which time, the allantoic vasculature amalgamates with those of the yolk sac and fetus (Downs et al., 1998; not shown).

The VCAM1-negative proximal region may, in future, winnow out genes that regulate chorio-adhesive cell formation. For example, Tbx4nullizygous allantoises failed to grow far enough to fuse with the chorion and did not exhibit VCAM1 (Naiche and Papaioannou, 2003). In light of findings here, absence of VCAM1 may be due to the foreshortened mutant allantois' failure to escape the streak's suppressive influence, rather than to defective genetic interaction between Tbx4 and Vcam1.

Intriguingly, despite a conspicuous absence of VCAM1, 8-hour proximal regions nonetheless often made contact with the chorion. Although the extent to which non-VCAM1- and VCAM1-containing proximal allantoic regions penetrate the chorion awaits further study, this observation is consistent with transgenic knockouts. These revealed that 50% of Vcam1 nullizygous embryos undergo chorio-allantoic union before dying at approximately 11.5 dpc(Gurtner et al., 1995; Kwee et al., 1995). Potential explanations for this intriguing observation have already been discussed(Gurtner et al., 1995; Kwee et al., 1995; Downs, 2002).

Why VCAM1 is normally absent from the proximal allantoic region is not clear, but given that a major allantoic function is to unite with the chorion and form the umbilical vasculature, repression might ensure that the proximal region cannot fuse with the chorion. If it, like the distal region, spread along the chorionic surface (Downs,2002), compression rather than elongation of the umbilical cord might occur. Although short cords are extremely rare, they are usually associated with fetal malformations(Benirschke, 1998).

Finally, although three cell types have thus far been characterized in the allantois, it is not known whether vascular smooth muscle cells arise from the allantois itself or are contributed to the allantois from the chorion. According to recent results (Yamashita et al., 2000; Motoike et al.,2003), vascular smooth muscle cells may originate from FLK1-positive precursor cells.

The function and origin of other cell types, known to exist only through gene expression patterns, remain unclear or unknown. For example,VCAM1-positive cells are closely associated with the FLK1-positive endothelium in the distal and not the proximal core, but their significance in allantoic development has not been elucidated (Downs et al., 2001; Downs,2002). One possibility is that, once the chorio-allantoic fusion surfaces break down, VCAM1-positive distal cells bridge the connection between the FLK1-containing endothelial cells and the chorionic ectoderm(Downs, 2002), binding toα4-integrin present in the chorion(Bowan and Hunt, 1999). In addition to core VCAM1, α4-integrin has been identified in a small cell population in the proximal region of the allantois, which is continuous with the amnion (Downs, 2002). These cells may be involved, through integrin-mediated cell motility, in expanding the amnion during fetal growth.

The authors are grateful to Professor Richard Gardner for his suggestion of use of AlexaFluor-conjugated Concanavalin A for fate mapping yolk sac and amniotic mesoderm as well as for his most helpful comments on the final draft of the manuscript, to Professor Tom Fleming for a valuable discussion on trophectoderm formation, to Randy Massey and Ben August of the University of Wisconsin-Madison Medical School Department of Anatomy's EM Facility for expert preparation of conceptuses for electron microscopy, to Dr Jacqueline Deschamps for help with Hox terminology, and to the anonymous reviewers for their very insightful and constructive criticisms. Grants HD036847 and HD042706 from the National Institutes of Health (K.M.D.) and a Hilldale Undergraduate Student Fellowship (E.H.) supported this study. Portions of this work will be submitted by Kimberly Inman in partial fulfillment of the Ph.D. requirements at the University of Wisconsin-Madison. Humane animal care was overseen by an institutional review board according to guidelines set by AALAC.