Investigation of cell lineage and differentiation in the extraembryonic endoderm of the mouse embryo

Primitive endoderm cells appear to differentiate on the blastocoelic surface of the inner cell mass (ICM) and, according to Nadijcka & Hillman (1974), can usually first be discerned by electron microscopy before the zona pellucida is lost, late on the fourth day of development. By the middle of the fifth day they can be distinguished from the remaining primitive ectoderm cells in dissociated ICMs, thereby enabling one or more cells of either type to be injected into genetically dissimilar host blastocysts (Gardner & Rossant, 1979). Following transplantation, primitive endoderm cells were found to colonize only one of the ICM derivatives into which host conceptuses were dissected, namely the extraembryonic endoderm of the visceral yolk sac. Primitive ectoderm cells, in contrast, contributed to all foetal and extraembryonic fractions of ICM origin that were analysed, except the visceral yolk-sac endoderm. Indeed, the precision with which the two classes of cells partitioned in host embryos suggested that both had acquired a relatively stable state of differentiation by the late blastocyst stage (Gardner & Rossant, 1979).

However, several studies on isolated ICMs grown in vitro have led to the conclusion that the primitive ectoderm retains the option of forming extra-embryonic endodermal cells after endodermal differentiation has taken place (Pedersen, Spindle & Wiley, 1977; Atienza-Samols & Sherman, 1979; Dziadek, 1979). This is based on the finding that if the layer of endoderm which differentiates initially on the surface of such ICMs is destroyed by immunosurgery, the purportedly endoderm-free cores of primitive ectoderm cells left behind can generate a new one. Dziadek (1979) found, in addition, that the regenerated endoderm differed from the original layer in displaying a much higher proportion of cells that reacted with an antiserum directed against alpha-foetoprotein, a visceral endoderm cell marker (Dziadek & Adamson, 1978). This led her to propose that two distinct phases of extraembryonic endoderm differentiation take place normally during development; formation of an initial parietal cell layer being succeeded by production of a second visceral layer from the underlying ectoderm (Dziadek, 1979). Such a cell lineage scheme clearly differs from one suggested earlier in which visceral and parietal endoderm were presumed to share a common origin from a single pool of primitive endoderm cells (see Gardner & Papaioannou, 1975; Gardner, 1978a, Fig. 1).

The blastocyst injection experiments outlined earlier provided no evidence for the existence in mature ICMs of cells capable of contributing to both visceral endoderm and primitive ectodermal derivatives that one might expect according to Dziadek’s (1979) hypothesis (Fig. 1). However, they were uninformative regarding the alternative possibility of a common origin for parietal and visceral endoderm, because only the latter of the two extraembryonic endodermal layers was analysed for chimaerism. Hence, more extensive cloning of 5th day ICM cells in blastocysts has been undertaken in which the resulting conceptuses were recovered earlier in gestation than previously, thus enabling inclusion of the parietal endoderm among the fractions analysed. These studies are reported in the present paper, together with the results of additional experiments in which the developmental potential of established parietal versus visceral endoderm cells from post-implantation embryos was also examined by blastocyst injection. A preliminary investigation had shown that visceral endoderm cells from both embryonic and extraembryonic regions of 6th day egg cylinders could colonize the visceral yolk sac endoderm of host conceptuses, albeit at a low frequency (Rossant, Gardner & Alexandre, 1978). However, parietal endoderm was neither used for injection nor included among the fractions analysed in these earlier transplantations of post-implantation cells.

Two stocks of mice were used, both of which had been derived from the PO (Pathology Oxford) random-bred albino strain. The two stocks were homozygous for different alleles at the glucose phosphate isomerase (Gpi-1) locus on chromosome 7, one being GPi-1^a/Gpi-1^aand the other Gpi-1^b/Gpi-1^b (Carter & Parr, 1967; De Lorenzo & Ruddle, 1969). The former was used to provide host blastocysts and pseudopregnant recipients, and the latter donor blastocysts and post-implantation embryos. All mice were maintained in controlled lighting and provided with food and water ad libitum. Most of the Gpi-1^a/Gpi-1^a stock and all Gpi-1^b/ Gpi-1^b mice were exposed to light between 07.00 and 19.00 h each day. Oestrous females were selected by vaginal inspection (Champlin, Dorr & Gates, 1973) before onset of the dark period, paired with fertile or vasectomized males, and inspected for vaginal plugs the following morning. The day on which the plug was found was recorded as the first day of pregnancy or pseudopregnancy. In addition, some matings were arranged between Gpi-1^a/Gpi-1^a mice kept in a room which was in darkness between 14.00 and 23.00 h each day. These were used to provide more mature blastocysts that are assumed to be approximately 6 h ahead of those from females mated in the more conventional lighting conditions. The former are referred to hereafter as advanced and the latter as standard, host blastocysts.

PB1 medium (Whittingham & Wales, 1969) was used for the recovery, storage at room temperature and manipulation of embryos and, in addition, for post-operative culture of the majority of injected blastocysts. It differed from the original formulation in containing glucose (1 g/1) in place of lactate and foetal calf serum (10% v/v) in place of bovine serum albumin. Occasionally, injected blastocysts were cultured in a-medium (Stanners, Eliceiri & Green, 1971) supplemented with 10 % (v/v) foetal calf serum because it supported more rapid re-cavitation than PB1, and thus increased the chances of identifying the location of the transplanted cells. One of two media (that of Whitten, 1971, in earlier experiments, and that in Table 6.5 of Biggers, Whitten & Whittingham, 1971, in later experiments), from which calcium salts had been omitted, was used for the culture of donor tissue between its exposure to proteolytic enzymes and its final dissociation. Cultures were usually set up either in microdrops of medium under liquid paraffin (Boots Pure Drug Co., U.K.) in 60x15 mm plastic culture dishes (Falcon, Oxnard, U.S.A.) or in glass cavity cells, and maintained at 37 °C in an appropriate gas phase. Injected blastocysts cultured in PB1 were sometimes kept in hanging drops in manipulation chambers (Puliv, Leitz, W. Germany) during incubation to facilitate detailed inspection during recovery.

Both standard and advanced host blastocysts were recovered between 15.45 and 18.00 h on the 4th and donor blastocysts between 13.00 and 15.45 h on the 5th day of pregnancy in all except one series of injections. The exception involved transplantation of earlier primitive ectoderm cells, for which advanced host blastocysts were recovered between 12.15 and 12.45 h and donor blastocysts between 09.15 and 09.45 h on the 4th and 5th day, respectively. The uterine horns of 5th day pregnant females were distended with medium prior to flushing, to aid release of the implanting blastocysts. Despite this measure, the number of blastocysts obtained per female on the afternoon of the 5th day was consistently lower than on the 4th day. Post-implantation donor embryos were obtained by dissection, with watchmaker’s forceps, of decidua from females killed between 14.30 and 16.50 h on the 6th or 7th day of pregnancy.

Initially, ICMs were recovered microsurgically from 5th day blastocysts (Gardner & Johnson, 1972), incubated for 10– 15 min in 0· 25% (w/v) pronase (Calbiochem, Grade B) in Dulbecco ‘A’ phosphate-buffered saline (PBS, Oxoid, U.K.) at 37 °C, and then cultured in calcium-free medium for a further 30– 45 min prior to dissociation. Since the yield of viable cells was rather variable, the following protocol was adopted in the majority of experiments. The mural trophectoderm of donor blastocysts was first torn open with a pair of siliconized (Repelcote, Hopkin & Williams, U.K.), sharp-tipped glass needles controlled by Leitz micromanipulators. The opened blastocysts were then exposed to a mixture of 0· 25 % (w/v) trypsin and 2· 5 % (w/v) pancreatin (both from Difco Laboratories, U.S.A.) made up in calcium-magnesium-free Tyrode saline (approximately pH 7· 7) for 20– 25 min at 4 °C. Thereafter, following a brief rinse in PB1, the blastocysts were micromanipulated again with glass needles in order to separate ICM tissue from investing trophectoderm and also, in some cases, primitive ectoderm from primitive endoderm (Gardner, 1982). The isolated ICMs or primitive ectoderms were incubated for 15 min in 0· 25 % pronase in PBS at 4 °C, rinsed in calcium-free medium, and then cultured for up to 45 min in the latter at 37 °C. Finally, the ‘loosened’ cell masses were transferred via a wide-bore pipette to a hanging drop of PB1 in a manipulation chamber, and dissociated by repeated aspiration through a siliconized pipette whose tip had been heat polished down to an inside diameter of 30 μm or less in a De Fonbrune microforge (Beaudouin, Paris). Virtually no cell death was seen using this procedure, which yielded suspensions composed principally of single cells and cell pairs.

For obtaining visceral endoderm cells, the embryonic or extraembryonic regions of 6th and 7th day embryos were first isolated as described by Rossant et al. (1978). They were then incubated in the trypsin-pancreatin mixture described above for approximately 15 min at 4 °C (Levak-Svajger, Svajger & Skreb, 1969), followed by 0· 25% pronase in PBS for a further 8 min at room temperature. Finally, each fragment was rinsed and then cultured in calcium-free medium for 30– 45 min before being restored to PB1 and dissociated by pipetting.

Difficulties were experienced in obtaining sufficient numbers of viable parietal endoderm cells from 6th day embryos. Therefore, only 7th day embryos were used as sources of these cells in the experiments reported in this paper. Intact embryos were first cut in two with very fine iridectomy scissors (Weiss, U.K.) below the level of insertion of Reichert’s membrane and the egg cylinder held by sucking its cut surface against a flame-polished pipette. Reichert’s membrane was then pulled off the egg cylinder by gripping the nipple-like extension at the distal tip with watchmaker’s forceps. Once isolated, Reichert’s membrane was incubated at 37 °C in pronase for 15 min followed by calcium-free medium for up to 30 min, and then pipetted to release the parietal endoderm cells. It was found in later experiments that enough viable cells were released spontaneously if isolated membranes were incubated in trypsin and pancreatin at 37 °C for 30– 40 min.

The rough appearance of endoderm cells (Gardner & Rossant, 1979) was used to distinguish them from ectoderm cells in disaggregates of 5th day ICMs and of embryonic and extraembryonic fragments from post-implantation embryos. Using a modified Leitz micromanipulator assembly, one or more cells of a particular type were injected into each of a series of host blastocysts by a technique that is described fully elsewhere (Gardner, 1978b). Operated blastocysts were cultured for up to 90 min prior to inspection and then transplanted to one or both uterine horns of females on the 3rd day of pseudopregnancy.

Recipients were killed between 9 and 12 days after they had received injected blastocysts, and conceptuses dissected as follows. First, Reichert’s membrane was dissected away from the edge of the chorioallantoic placenta, and trophoblast plus decidual tissue removed from its outer surface. The visceral yolk sac was then separated from the base of the placenta, and the foetus, amnion and umbilical cord removed from inside it. Next, the decidua basalis was separated from the chorioallantoic placenta and discarded. The placenta, Reichert’s membrane and attached parietal endoderm cells, trophoblast (minus as much decidua capsularis as possible), and foetus plus amnion and umbilical cord, were handled as four separate fractions. They were rinsed in PBS, placed in separate wells of a microtest plate (Nunc, Denmark), diluted with a small volume of distilled water, and then frozen. Visceral yolk sacs were incubated in the trypsin-pancreatin solution described earlier for a minimum of at 4 °C, after which the endoderm was separated from the mesoderm in PB1. The separated layers were then treated as described above for the other fractions.

In some cases, visceral yolk sacs and/or Reichert’s membranes were cut in two with iridectomy scissors prior to freezing so as to provide separate proximal (placental) and distal (ab-placental) fractions for analysis. The technique for separation and visualization of the glucose phosphate isomerase allozymes was as described previously (Gardner, Papaioannou & Barton, 1973; Gardner & Rossant, 1979), except that the 3 mm-thick starch gels were sliced in two following electrophoresis, and the cut surface of each slice stained separately through an agar overlay. Proportions of the two allozymes in a sample were estimated visually from the electrophoretograms by two independent observers.

Cells for injection were obtained either by dissociating microsurgically isolated primitive ectoderm or, as previously, by selecting ‘smooth’ cells from disaggregated ICMs (Gardner & Rossant, 1979). A single cell was transplanted into each blastocyst in one experiment, and either pairs or groups of up to six cells, all or most of which were loosely attached to one another, in the remainder (Tables 1 and 2). Donor cells typically adhered to the centre of the blastocoelic surface of host ICMs, against which they were invariably injected. Standard rather than advanced 4th-day blastocysts were used as hosts in all experiments except those involving transplantation of cells from blastocysts recovered on the morning of the 5th day.

The rates of normal development and numbers of chimaeric conceptuses obtained in the different series of injections are presented in Table 1. Table 2 provides details of the distribution and extent of contribution of the donor cells in each of the 46 chimaeras that were obtained in these experiments.

A minor contribution of donor cells to the placental fraction was seen in 19 cases (Table 2), but is an uninformative observation because the embryonic part of this organ is evidently composed of cells of trophectodermal, primitive ectodermal and primitive endodermal derivation (Duval, 1892; Snell & Stevens, 1966; Gardner & Papaioannou; 1975). Hence, this fraction will not be considered in the following appraisal of the distribution of donor cells.

Forty-one of the chimaeras exhibited the B type of pattern of chimaerism established earlier for primitive ectoderm cells (Gardner & Rossant, 1979) in which donor cells were confined to the foetal fraction and/or visceral yolk-sac mesoderm (all except S4, S29, S39, S40 and S41 in Table 2). Furthermore, the donor contribution approached or exceeded 50 % in one or both of the fractions in more than half these chimaeras. Indeed, donor allozyme activity alone was detected in both the visceral yolk-sac mesoderm and foetal fractions of one chimaera (S32, Table 2). Enzyme activity was absent from the visceral yolk-Sac mesoderm of three conceptuses and too low for unequivocal scoring in a further two. The reason for this is not clear, since no problem was encountered with the corresponding endodermal fractions that had the same period of exposure to trypsin and pancreatin.

Three of the five chimaeras that did not conform to the B pattern contained donor cells in the trophoblastic fraction as well as in the foetal fraction and visceral yolk-sac mesoderm (S4, S39 and S41 in Table 2). The extent of the donor contribution to the trophoblast is almost certainly underestimated because of unavoidable dilution of this fraction with contaminating maternal decidual cells. Donor cells were confined to the parietal and visceral yolk-sac endoderm in one of the remaining two exceptional chimaeras (S29 in Table 2). The other displayed a very minor contribution of donor cells to the parietal endoderm fraction in addition to major contributions to the foetal and visceral yolk-sac mesodermal fractions (S40). Since this last chimaera did not show evidence of trophoblastic chimaerism, the parietal contribution cannot be attributed to contamination of the latter fraction with cells from the former. All five exceptional chimaeras were produced in experiments in which more than one cell was injected into each blastocyst. The distribution of chimaerism seen in S4, S39, S40 and S41 (Table 2) may therefore be attributable to inadvertent inclusion of one or more trophectoderm or primitive endoderm cells among those transplanted rather than to lack of restriction in potency of primitive ectoderm cells. The pattern observed in S29 is very difficult to explain, unless one assumes that an attached group of primitive endoderm cells had been misclassified as primitive ectoderm cells. Notwithstanding, no case of chimaerism specifically limited to the visceral yolk-sac endoderm, or to this tissue plus the corresponding mesoderm and/or foetal fraction was encountered, even among the 12 conceptuses colonized by donor cells recovered earlier on the 5th day of development (S10-S12 and S38-S46 in Table 2).

Donor cells with a ‘rough’ appearance were selected, and injected either centrally against the surface of the ICM of host blastocysts, or peripherally at the junction between ICM and mural trophectoderm. Both advanced and standard 4th day blastocysts were used as hosts, the former typically differing from the latter in showing signs of endodermal differentiation and, in some cases, loss of the zona pellucida as well. The majority of host blastocysts were injected with single cells, the remainder each receiving an attached pair of cells (Table 3).

As shown in Table 3, 63 of the 117 normal conceptuses derived from blastocysts injected with single cells were found to be chimaeric, representing a cloning efficiency of the donor cells of 54 %. The foetal fractions and/or visceral yolk-sac mesoderm were analysed in addition to both the parietal and visceral yolk-sac endoderm in 50 of these chimaeras (nos. R1– R48, including R2A and R15A, in Table 4). In all except three of the latter the progeny of the transplanted cell appeared to be confined exclusively to the extraembryonic endoderm. Two out of three of the exceptional chimaeras (nos. R2A and R15A in Table 4) displayed patterns of colonization appropriate for primitive ectoderm rather than primitive endoderm, and may therefore represent cases in which the donor cell had been misclassified. The third (no. R31 in Table 4) exhibited a high proportion of donor allozyme activity in the endoderm layer of the visceral yolk sac, and a much lower proportion in the corresponding mesodermal layer (Fig. 2 F). This conceptus was recovered on a day in which unusual difficulty was experienced in separating the two layers of visceral yolk sacs following cold enzyme treatment. The absence of donor allozyme activity from the foetal fraction of this chimaera is consistent with the conclusion that its presence in visceral mesoderm was due to contamination of the latter with visceral endoderm cells. None of the 28 chimaeras in which the trophoblast was analysed exhibited colonization of this fraction (Table 4). A modest donor contribution was evident in 2 out of 23 cases in which the chorioallantoic placenta was examined. However, as noted earlier, these particular findings are not instructive because of the composite nature of the embryonic part of this organ.

Two further points are evident from Tables 3 and 4 concerning the single-cell injections. First, in 8 of the 61 chimaeras in which the extraembryonic endoderm had been colonized, progeny of the donor cell were detected unequivocally in both its parietal and visceral layers (chimaeras R14, R15, R16, R17, R26, R41, R48 and R74 in Table 4; and Fig. 2A– E). Equivocal results were obtained in two further cases (R29 and R30 in Table 4) because of low enzyme activity in the parietal endoderm. Secondly, chimaerism was confined to the parietal endoderm in most cases, despite all but 12 conceptuses having developed from blastocysts in which the donor cell had been placed against the centre of the free surface of the ICM (Table 4). Indeed, only two chimaeras (or three if R31 is included) appeared to have donor cells restricted to the visceral yolk-sac endoderm. A marked bias is evident in the central injection into both standard and advanced host blastocysts, 23 out of 29 in the former series (excluding R2A and R15A, but including R31) and 16 out of 20 in the latter showing donor cells in the parietal endoderm only. It is clear from Table 4 that, in some of the cases in which the position of the donor cell could be established following postoperative culture, it had not remained at the site of injection (see R72-R80 in Table 4). However, even when the cell did retain a central location (see chimaeras R20-R22, R29-R37, and R68-R71, in Table 4), its progeny was found exclusively in the parietal endoderm in the majority of cases. Peripheral injection of the donor cell did not altogether preclude its making a contribution to the visceral endoderm (see chimaeras R14, R41 and R48 in Table 4).

Pairs of ‘rough’ cells which had resisted separation during dissociation of donor ICMs were selected for transplantation, and injected centrally into either standard or advanced host blastocysts. As expected, the frequency of chimaerism was higher than in the single-cell injections, and the donor contributions were also usually greater (Tables 3 and 4). In the 10 chimaeras whose visceral yolk-sac mesoderm was analysed in addition to both parietal and visceral yolk*sac endoderm, this fraction was devoid of donor allozyme activity. The distribution of chimaerism with respect to parietal versus visceral yolk-sac endoderm once again showed a strong bias towards the former fraction. Furthermore, the 10 chimaeras in this series in which distal and proximal halves of the extraembryonic endodermal layers were analysed separately (chimaeras R49-R58 in Table 4) support the impression gained from the eight single-cell injection chimaeras treated similarly (chimaeras R41– R48 in Table 4), of non-random distribution of donor cells within the parietal endoderm itself. Thus, proximal parietal endoderm was chimaeric in 14 and distal parietal endoderm in all 18 cases in which the two regions were analysed separately (Table 4C). In none of the 14 in which both regions were chimaeric did the donor contribution to the proximal obviously exceed that to the corresponding distal half. Contributions were approximately equal in six conceptuses (R42, R44, R47, R48, R50 and R54, Table 4), distal exceeding proximal in the remaining eight (R45, R51, R52, R53, R55, R56, R57 and R58, Table 4).

Only two of the conceptuses whose proximal and distal visceral yolk-sac endoderm regions were analysed separately exhibited chimaerism in this layer (R48 in the single and R51 in the cell-pair injections, Table 4), and it was confined to the former region in both cases. Finally, visceral endodermal chimaerism exceeded parietal in only 3 (R14, R48 and R63, Table 4) of the 11 conceptuses obtained in both the single-cell and cell-pair injections in which both layers were chimaeric. Levels of visceral and parietal chimaerism were approximately equal in two cases (R17 and R74, Table 4), parietal exceeding visceral in the remaining six (R15, R16, R26, R41, R51 and R66, Table 4).

Several cells, often consisting of a mixture of both singletons and loosely attached pairs of groups, were injected centrally against the blastocoelic surface of the ICM of each host blastocyst. The origin and approximate numbers of donor cells injected per blastocyst are indicated in Table 5, together with the corresponding rates of normal development and chimaerism, and also the frequencies of different patterns of colonization. Estimates of the extent to which donor cells contributed to the fractions that they colonized are presented in Table 6 for all 34 chimaeras obtained in these experiments. Three fractions, parietal endoderm, visceral yolk-sac endoderm and visceral yolk-sac mesoderm, were analysed in all but seven conceptuses. Additional fractions were analysed in a proportion of conceptuses developed from blastocysts which had received parietal or visceral endoderm cells from 7th day donor embryos (see Table 6 for details). In no case were donor cells detected in fractions other than the parietal and/or visceral endoderm.

A high frequency of chimaeras was obtained with 6th day visceral endoderm cells, regardless of whether they were isolated from the embryonic or extraembryonic region of the egg cylinder. In addition, donor cells showed as strong a bias towards parietal rather than visceral endodermal colonization as endoderm cells from 5th day donor embryos (cf. Tables 3 and 5, and Fig. 3). The rate of chimaerism fell dramatically when 7th day visceral endoderm cells were transplanted (6 % overall as opposed to 79 % for 6th day cell injections). Nevertheless, three of the four chimaeras showed exclusively parietal endodermal colonization, donor cells being detected in the visceral endoderm layer only in the fourth chimaera (V27– V30 in Table 6 and Fig. 3). The somewhat higher rate of chimaerism obtained in the 7th day parietal endoderm cell injections (13 %) is probably attributable to the fact that, in general, more cells were transplanted into each blastocyst than was the case in the corresponding visceral endoderm injections (Table 5). Colonization of their tissue of origin only was seen in all four unequivocal chimaeras produced in parietal endoderm cell injections (P1– P4 in Table 6).

Two further points are worthy of note. First, donor contributions are somewhat variable but, as is evident from comparison of Tables 4 and 6, generally lower than those obtained by injection of single cells or cell pairs from 5th day donors. Secondly, parietal endoderm fractions that had been colonized by postimplantation visceral endoderm cells appeared indistinguishable from those that had not when inspected at a magnification of × 50 during removal of contaminating trophoblasts from Reichert’s membrane.

Altogether, 128 chimaeras were obtained in these experiments. All except 5 could be assigned unequivocally to one of two classes on the basis of the distribution of chimaerism. Thus, donor cells were confined to the extraembryonic endoderm only in 80 chimaeras, and to the extraembryonic mesoderm of the visceral sac and/or foetal fraction in a further 43. Furthermore, the pattern of chimaerism corresponded with the morphological type of donor cells injected in all but 3 of these 123 chimaeras. In one of the three exceptions, putative primitive ectoderm cells colonized the extraembryonic endoderm (chimaera S29 in Table 2), while in the other two putative primitive endodermal cells colonized the foetal or visceral yolk-sac mesoderm fractions (chimaeras R2A and R15A in Table 4). These presumably represent instances in which donor cells were misclassified, possibly because, as found earlier, the morphological distinction between the two categories of ICM cell depends rather critically on their treatment during and after dissociation (Gardner & Rossant, 1979).

Four of the remaining five chimaeras that did not conform to either of the above patterns of colonization were obtained in experiments in which several cells were injected into each host blastocyst. Hence, they may represent cases in which a mixture of primitive ectoderm plus polar trophectoderm (S4, S39, and S41, Table 2) or primitive endoderm cells (S40, Table 2) were transplanted inadvertently. Contamination of donor ICMs or primitive ectoderms with polar trophectoderm cells might well occur if the latter tissue had already embarked on the process of thickening that leads to extraembryonic ectoderm formation (Copp, 1979). Extraembryonic ectoderm cells resemble those of the primitive ectoderm in their smooth appearance, and can clearly colonize the trophecto-dermal derivatives of host embryos following blastocyst injection (Rossant, et al. 1978). Likewise, the distinction between primitive ectoderm and endoderm cells is less marked in earlier 5th day blastocysts which were used as donors in the experiments in which S40 was obtained. The fifth anomalous chimaera, in which donor cells were detected in both endoderm and mesoderm layers of the visceral yolk sac, was produced in a series of experiments in which a single cell was injected into each host blastocyst (R31 in Table 4). This is the only instance in which a 5th day ICM cell clone appeared not to partition between the extraembryonic endoderm and other ICM derivatives. However, as noted earlier, there are grounds for suspecting that contamination of visceral yolk-sac mesoderm with corresponding endodermal cells may account for this particular result.

If the two extraembryonic endoderm layers had separate origins, as proposed by Dziadek (1979), individual early endoderm cells would be expected to colonize either the parietal or the visceral endoderm of host embryos (Fig. 1). In fact, at least 13% of clones clearly spanned both tissues. In addition, one might expect to find primitive ectodermal clones that contributed specifically to the visceral layer of the extraembryonic endoderm as well as to the corresponding mesodermal layer and/or foetus (Fig. 1). No unequivocal example of such clones was found. One explanation for the failure to detect them might be that they originate from morphologically intermediate ICM cells (Gardner & Rossant, 1979) whose transplantation was avoided in the present experiments. This is most improbable because, although intermediate cells are usually found in both isolated entire primitive ectoderms and corresponding endoderms (Gardner & Rossant, 1979), these two tissues consistently yield identical patterns of chimaerism to dissociated ‘smooth’ and ‘rough’ cells, respectively, following injection into blastocysts (Gardner, 1982). A further possibility, that the cells in question are no longer present at the stage at which donor embryos were recovered, is equally unlikely. Although individual earlier ICM cells can indeed contribute to both extraembryonic endoderm and other ICM derivatives of host embryos, they do so to the parietal as well as the visceral layer as often as to the latter alone (Rossant & Gardner, 1982).

One further point about Dziadek’s hypothesis concerns the in vitro experiments on which it is based. Regeneration of endoderm has been observed consistently only in ICMs isolated from giant blastocysts composed of three or more embryos (Pedersen et al. 1977; Dziadek, 1979). Similar investigations on ICMs from standard blastocysts have yielded completely negative results (Sherman, Strickland & Reich, 1976; Hogan & Tilly, 1977; D. Solter, personal communication of unpublished observations) or, at best, a very modest rate of regeneration (Atienza-Samols & Sherman, 1979). Even when regeneration is obtained the evidence attributing it to primitive ectoderm cells is not com-pelling (Gardner, 1981). Indeed, in very recent experiments it has been found that, whereas ‘primitive ectoderms’ isolated by immunosurgery from 5th day blastocysts almost invariably regenerated an endodermal layer, those isolated microsurgically did not do so even if their cell number was enhanced by aggregation (Gardner, 1982). The reason for this discrepancy seems to be that the endoderm is already multilayered by the stage at which it is clearly discernible, so that some of its cells are protected from exposure to the immunosurgical reagents. Results of both in vitro and in vivo experiments lead to the conclusion that it is from these residual endoderm cells rather than the primitive ectoderm that the entire regenerated layer is derived (Gardner, 1982).

The fact that single 5th day endodermal cell clones can contribute to both the parietal and visceral endoderm of host embryos is clearly consistent with the alternative hypothesis that the two types of extraembryonic endoderm cell share a common precursor (Gardner & Papaioannou, 1975; Gardner, 1978a).

Donor cells were detected exclusively in the extraembryonic endoderm in all chimaeras produced in these experiments, despite the screening of one or more additional fractions in each case. Sixth day visceral endoderm cells gave a relatively high rate of chimaerism, and displayed as strong a bias towards parietal rather than visceral colonization as 5th day endoderm cells. Nevertheless, the level of chimaerism was usually lower, suggesting that 6th day cells form smaller clones following transplantation than 5th day cells. Seventh day visceral endoderm cells yielded a very low rate of chimaerism compared with their 6th day counterparts, regardless of whether they came from the embryonic part of the egg cylinder, the extraembryonic part, or from the junctional zone between the two. However, since the donor contributions in the four chimaeras compared favourably with those produced by similar numbers of 6th day cells, the difference between the two stages may be attributable to a lower frequency of clonable cells in 7th day visceral endoderm rather than to a reduction in size of clones that such cells can form.

What is the significance of the unexpected finding that established visceral endoderm cells yielded exclusively parietal chimaerism following injection into blastocysts in the vast majority of cases? Do the transplanted cells actually undergo phenotypic change, or do they retain their visceral characteristics and simply accumulate in the parietal endoderm through failure to become integrated in their normal site? The latter seems most unlikely, especially since, even in chimaeras in which donor allozyme accounted for 20 % or more of the total glucose phosphate isomerase activity, no morphological peculiarities could be discerned in this tissue and post-implantation development appeared to have proceeded normally. Support for the former possibility can be found in other studies. Thus, Diwan & Stevens (1976) obtained histological evidence that visceral endoderm tissue isolated from 6th day mouse embryos can form parietal endoderm when transplanted to the testis. In addition, Hogan & Tilly (1981) have provided both ultrastructural and biochemical evidence that transformation of visceral into parietal endoderm cells can also take place in 7th day egg cylinders in vitro. However, it is not clear either in these studies or the present one, whether all visceral endoderm cells retain this option, or only a subpopulation of relatively immature stem cells. The dramatic drop in frequency of chimaeras when 7th as opposed to 6th day visceral cells were injected into blastocysts is certainly consistent with the latter possibility.

Four chimaeras were produced by transplanting 7th day parietal endoderm cells, but despite injection of up to 10 cells per blastocyst the donor contribution was very modest in each case. Given the extent to which parietal chimaerism is favoured even with visceral endoderm, a much larger series of chimaeras would have to be analysed in order to establish whether parietal cells retain the option of forming their visceral counterparts. Furthermore, the significance of any positive results would be difficult to evaluate unless higher donor cell contributions were achieved than in current injections using these cells.

The much higher rate of parietal than visceral chimaerism that is evident in the present endoderm cell injections affords an explanation for the relative paucity of chimaeras in those undertaken earlier in which only the visceral component of the extraembryonic endoderm was analysed (Rossant et al. 1978; Gardner & Rossant, 1979). This very marked bias towards parietal chimaerism does not seem to depend simply on preferential attachment of donor cells to the mural trophectoderm rather than the ICM of host blastocysts. Some 5th day endoderm cells that were placed on the ICM clearly shifted peripherally during post-operative culture. However, a majority of ‘parietal endoderm only’ chimaeras was obtained even in those cases in which the donor cells remained attached to the ICM (Table 4). In addition, the distribution of chimaerism did not seem to be influenced by the stage of host blastocysts at injection, relatively earlier blastocysts giving similar results to the more advanced ones which already showed signs of endodermal differentiation. A further possibility is that donor endoderm cells that colonize the visceral layer may be less likely to survive if, for example, the visceral embryonic endoderm undergoes degeneration rather than displacement during formation of the definitive embryonic endoderm (Gardner, 1978 a). However, this would not account for the finding that, even within the parietal endoderm, distal chimaerism appears to be favoured over proximal. It also fails to explain the fact that those single ICM cells from early 4th day blastocysts which yield extraembryonic endodermal chimaerism following blastocyst injection do not show preferential parietal colonization. The majority of clones formed by such cells contribute to both tissues of the extraembryonic endoderm, those confined to the parietal layer being no more common than those confined wholly to the visceral yolk-sac endoderm (Rossant & Gardner, 1982).

The work of Hogan & Tilly (1981) suggests that disruption of the epithelial organization or alteration in the cellular substratum of visceral endoderm cells may be responsible for their conversion to parietal endoderm cells. It raises the interesting possibility that such factors may also play a role in the normal differentiation of the extraembryonic endoderm. Thus, initially, all primitive endoderm cells are in contact with primitive ectoderm cells because they evidently differentiate as a monolayer on the blastocoelic surface of the ICM. However, their increase in number in the late blastocyst is not accompanied by a corresponding increase in surface area of the underlying ectoderm. Hence, mitoses in the primitive endoderm are likely to be so oriented that, in general, only one daughter cell remains adjacent to the primitive ectoderm, the other being forced into a new superficial layer (Fig. 4). In addition, peripheral endoderm cells may suffer direct lateral displacement on to the adjacent mural trophectoderm. Release from contact with the ectoderm by either means might be the critical event enabling the cells to migrate peripherally and differentiate into parietal endoderm. Once implantation has occurred, rapid expansion of the ectodermal surface during egg-cylinder formation would permit both daughters of dividing visceral endoderm cells to remain within this layer (Fig. 4). This scheme accounts for the multi-layering of endoderm cells seen in the blastocyst prior to parietal endoderm formation (Gardner, 1982). It also explains the paucity of visceral relative to parietal endoderm cells in the late blastocyst, and the rapid increase in number of the former following implantation (Enders, Given & Schlafke, 1978).

The blastocyst injection data may be explained as follows. Single early ICM cells that undergo endodermal differentiation following transplantation typically contribute to both parietal and visceral layers because they are efficiently incorporated into the primitive endoderm cell monolayer in host blastocysts (Fig. 4). Differentiated endoderm cells, by contrast, are less likely to achieve the necessary integration following isolation, and will therefore tend to form the first and hence most distally migrating parietal endoderm cells (Fig. 4).

Available data on early differentiation of the extraembryonic endodertn in the rodent embryo can thus be accommodated in the above scheme. However, many questions need to be answered before the underlying processes can be understood in detail. For example, is transformation to parietal cells a property of all visceral cells or of a specific subpopulation within this layer that retains a primitive endodermal character? Is it an irreversible or reversible change? How late in development can it occur, and does it play a role in normal growth of the parietal yolk sac? Recent studies on endodermal differentiation in embryonal carcinoma cells in culture in response to defined molecules offer a promising system for examining more closely the relationship between parietal and visceral cells (e.g. Hogan, Taylor & Adamson, 1981; Strickland & Mahdavi, 1978; Strickland, Smith & Marotti, 1980). Nevertheless, much more needs to be learned about the patterns of growth of parietal and visceral endoderm in the embryo and, in particular, the properties of cells in the region of continuity between the two layers of extraembryonic endoderm that is eventually incorporated into the placenta.

I wish to thank Dr R. Beddington, Mr S. Buckingham, Mrs W. Gardner, Dr C. F. Graham, Professor H. Harris, Mr K. Mabbatt and Mrs J. Williamson for help in preparation of the manuscript, and Mrs M. Carey and Mrs L. Ofer for technical assistance. The work was supported by the Medical Research Council and the Royal Society.