Development 101, 627-652 (1987) 627 Printed in Great Britain (G) The Company of Biologists Lunited 1987
Cell fate, morphogenetic movement and population kinetics of embryonic endoderm at the time of germ layer formation in the mouse
KIRSTIE A. LAWSON1 and ROGER A. PEDERSEN23 with an appendix by SARA VAN DE GEER4 lHubrecht Laboratory, Netherlands Institute for Developmental Biology, Uppsalalaan 8, 3584 CT Utrecht, The Netherlands, 2Laboratory of Radiobtology and Environmental Health and ^Department of Anatomy, University of California, San Francisco, CA 94143, USA ^Centre for Mathematics and Computer Science, Knuslaan 413, 1098 SJ Amsterdam, The Netherlands
Summary
The fate of the embryonic endoderm (generally called and early-streak-stage embryos is heterogeneous in its visceral embryonic endoderm) of prestreak and early germ layer fate. Whereas the germ layer location of primitive streak stages of the mouse embryo was descendants from anterior sites did not differ after 1 studied in vitro by microinjecting horseradish peroxi- day from that expected from the initial controls dase into single axial endoderm cells of 6 7-day-old (approx. 90 % exclusively in endoderm), only 62 % of embryos and tracing the labelled descendants either the successfully injected posterior sites resulted in through gastrulation (1 day of culture) or to early labelled cells exclusively in endoderm; the remainder somite stages (2 days of culture). contributed partially or entirely to ectoderm and Descendants of endoderm cells from the anterior mesoderm. This loss from the endoderm layer was half of the axis were found at the extreme cranial end compensated by posterior-derived cells that remained of the embryo after 1 day and in the visceral yolk sac in endoderm having more surviving descendants (8-4 h endoderm after 2 days, i.e. they were displaced population doubling time) than did anterior-derived anteriorly and anterolaterally. Descendants of cells cells (10-5 h population doubling time). There was no originating over and near the anterior end of the early indication of cell death at the prestreak and early primitive streak, i.e. posterior to the distal tip of the streak stages; at least 93 % of the cells were prolifer- egg cylinder, were found after 1 day over the entire ating and more than half of the total axial population embryonic axis and after 2 days in the embryonic were in, or had completed, a third cell cycle after 22 h endoderm at the anterior intestinal portal, in the culture. foregut, along the trunk and postnodally, as well as We suggest that the visceral embryonic endoderm, anteriorly and posteriorly in the visceral yolk sac. derived from the primitive endoderm of the late Endoderm covering the posterior half of the early blastocyst, is displaced onto the yolk sac by a new primitive streak contributed to postnodal endoderm population of endoderm inserted from epiblast at the after 1 day (at the late streak stage) and mainly to anterior end of the early primitive streak. This cell posterior visceral yolk sac endoderm after 2 days. population has colonized the axial endoderm by the Clonal descendants of axial endoderm were located neural plate stage and contributes to the embryonic after 2 days either over the embryo or in the yolk sac; endoderm of the early somite embryo. the few exceptions spanned the caudal end of the embryo and the posterior yolk sac. Key words: cell fate, morphogenetic movement, cell proliferation, embryonic endoderm, mouse embryo, The clonal analysis also showed that the endoderm postimplantation development, cell lineage, horseradish layer along the posterior half of the axis of prestreak- peroxidase, microinjection.
Introduction layer of primitive endoderm that faces the blastocyst cavity and a core of primitive ectoderm (Nadijcka & During late preimplantation development of the Hillman, 1974; Enders, Given & Schlafke, 1978). The mammal, the inner cell mass differentiates into a fate of the primitive endoderm has been studied in 628 K. A. Lawson, R. A. Pedersen and S. van de Geer rodent embryos using preimplantation injection streak from embryos on the 9th day of gestation also chimaeras and by analysis of tissue potency at early contributed to gut endoderm (Tarn & Beddington, postimplantation stages. Primitive endoderm cells 1987). No gut derivatives were formed by distal isolated from 4-5 day of gestation mouse embryos and donors transferred to other host locations or from injected into 3-5-day host blastocysts formed only other donor sites transferred to a distal host location visceral and parietal yolk sac endoderm (Gardner (Beddington, 1982). These apparent constraints in & Papaioannou, 1975; Gardner & Rossant, 1979). contribution to gut tissues in the postimplantation Single primitive endoderm cells from 5-5-day donors chimaeras imply that formation of the definitive and groups of visceral embryonic endoderm cells endoderm in the intact mouse embryo is a precisely from 6-5- or 7-5-day embryos formed both parietal organized morphogenetic process. and visceral endoderm descendants, leading to the We have recently described the fate of axial embry- concept that the primitive endoderm lineage con- onic endoderm cells in midstreak and late streak/ tains stem cells capable of forming both cell types neural plate stage (7-5 day) mouse embryos using (Gardner, 1982, 1984). Cells isolated from the primi- microinjected horseradish peroxidase (HRP) as a tive ectoderm layer at 4-5 days and injected into host short-term lineage tracer (Lawson, Meneses & Peder- blastocysts were capable of forming all the fetal sen, 1986). This lineage tracer (Weisblat, Sawyer & tissues, including the embryonic gut, as well as Stent, 1978; Balakier & Pedersen, 1982), while lack- extraembryonic mesoderm and amniotic ectoderm. ing some of the advantages of a cell-autonomous There is no evidence from the blastocyst injection genetic marker (Rossant, Vijh, Siracusa & Chapman, studies (using glucose phosphate isomerase as a 1983; Gardner, 1984), nonetheless facilitates labelling lineage tracer) that visceral embryonic endoderm of single cells in situ without disrupting the native cell contributes any descendants to the fetal gut at mid- relationships of the intact embryo. Our analysis gestation, although a minor contribution might not revealed that the visceral embryonic endoderm (here- have been detected with this approach (reviewed by after called embryonic endoderm) is a mixed popu- Rossant, 1986). lation at midstreak and late streak stages, consisting When visceral embryonic and extraembryonic en- of progenitors of visceral extraembryonic (yolk sac) doderm are isolated from primitive-streak-stage rat endoderm and progenitors of the embryonic gut and mouse embryos, they have extremely limited endoderm. Based on the relative locations of these tissue potency, as determined by ectopic grafting. distinct progenitor cell populations, we proposed that Visceral embryonic or extraembryonic endoderm the progenitors of embryonic foregut endoderm either yielded only parietal endoderm-like cells emerge from the epiblast earlier in gastrulation, (Solter & Damjanov, 1973; Diwan & Stevens, 1976) replacing the primitive endoderm cells, which con- or did not survive in ectopic grafts (Levak-Svajger & tribute to yolk sac endoderm. This proposal em- Svajger, 1971, 1974). By contrast, embryonic ecto- phasizes the similarity between rodent and avian derm isolated from mouse and rat embryos before embryos, as previously noted (Levak-Svajger & and during gastrulation and grafted to ectopic sites Svajger, 1974; Rossant & Papaioannou, 1977; has broad tissue potency (Grobstein, 1952; Levak- Gardner, 1978; Beddington, 19836). Svajger & Svajger, 1971; Diwan & Stevens, 1976). In the current study, we extend this approach to Both distal and anterior regions of the mouse embry- prestreak- and early-streak-stage mouse embryos (6-7 onic ectoderm isolated at the primitive streak stage days) to trace the origin of the axial endoderm of and grafted to ectopic sites formed midgut and midstreak- and late-streak-stage embryos and to foregut derivatives (Beddington, 1983a). Taken examine the extent of the contribution of embryonic together with the blastocyst injection studies, these endoderm to the yolk sac. This was accomplished by patterns of postimplantation tissue potency imply marking axial endoderm cells with HRP at prestreak that the primitive ectoderm or its derivatives, rather and early streak stages, culturing embryos for 1 or 2 than the primitive endoderm, gives rise to the defini- days, then determining the location of labelled de- tive fetal gut tissues. scendants. This approach also provides information Little is known about when the definitive endo- about the morphogenetic movements and population derm forms (reviewed by Beddington, 1983b, 1986). dynamics of the embryonic endoderm during this When postimplantation mouse chimaeras were period. formed by transferring small groups of [3H]thymi- dine-labelled donor ectoderm cells from the distal Materials and methods region of the egg cylinder on the 8th day of gestation to a distal host site, labelled descendants were Embryos detected in the embryonic gut (specifically, midgut) Noninbred Swiss mice of the Dub:(ICR) strain were used. (Beddington, 1981); orthotopic grafts of the primitive Gestation was considered to have begun at midnight before Cell fate in mouse endoderm 629 the morning on which a copulation plug was found. Females Cells containing HRP were detected by treating the intact were killed by cervical dislocation between 15.00 and embryos with 0-1% Hanker-Yates reagent (Polysciences) 16.00 h on the 7th day of pregnancy (6-7-day embryos). Egg in OlM-phosphate buffer, pH5-5 (Streit & Reubi, 1977) cylinders, including most of the ectoplacental cone, were plus 5 % (w/v) sucrose and 002 % H2O2 for 15-100 min in dissected from the decidua in Dulbecco's phosphate- the dark. The position of visible, labelled cells and their buffered saline. Reichert's membrane was removed with approximate number were recorded on freehand drawings glass needles, and this and all further manipulations were of the embryos. Embryos were fixed in glutaraldehyde and made in flushing medium II (Spindle, 1980) containing 10 % embedded in glycol methacrylate as previously described fetal calf serum. There was considerable variation in (Lawson et al. 1986), and exact localization and number of developmental stage both between and within litters of labelled cells were determined on 10^m serial sections, nominal 6-7-day embryos. Embryos were classified as 'early followed by photographic reconstruction (Lawson et al. streak' or 'prestreak'. In early-streak-stage embryos meso- 1986). derm formation had begun caudally: in the most advanced embryos of this group the posterior amniotic fold had been initiated and the thin end of the mesodermal wedge Results indicating the anterior end of the primitive streak was located about one third of the length of the embryonic (A) Embryos cultured for 1 day ectoderm posterior to the distal tip of the egg cylinder (1) Embryo development (fig. 64/1 in Theiler, 1972). The bilateral symmetry of the prestreak embryos was recognized by the slightly thicker After 1 day of culture, the majority of prestreak-stage posterior embryonic ectoderm and slight asymmetry of the embryos had developed to the midstreak stage endodermal outline (fig. 52 in Theiler, 1972). (Table 1). Early-streak-stage embryos had mostly reached late streak and neural plate stages; the Embryo culture initially most advanced were forming a head fold and Embryos that had been microinjected with HRP were foregut invagination. The embryos appeared normal cultured in Dulbecco's modified minimal essential medium except that, after amnion closure, the amniotic cavity containing 50% rat serum, as previously described (Law- and the exocoelom overexpanded slightly. son et al. 1986). Four to six embryos were cultured for 22h±l-7hinlml medium. Embryos maintained for 2 days (2) Validity of the labelling technique in culture were transferred to fresh medium (two In control embryos stained directly after injection, embryos/l ml) after the first day and incubated for a total of 72 % (58/81) of the sites were successfully injected 44h. (Table 2). Of these, most of the labelled cells were in endoderm and the remainder in ectoderm/meso- Cell labelling with HRP derm. (Ectoderm/mesoderm indicates ectoderm The conditions for intracellular injection by iontophoresis and/or mesoderm.) In more than half of the success- have been described previously (Balakier & Pedersen, fully injected sites a single cell was stained (Table 3, 1982; Lawson etal. 1986). One axial or near axial endoderm cell per 6-7-day-old embryo was injected with 4% HRP series 1; Fig. 1A); almost all of the remaining sites (Sigma Type VI) in 0-05 M-KCI for 15 S with 5 nA continuous had two adjacent cells labelled. After 22h culture, positive current. The site of injection was recorded on a 73 % (117/160) of injected embryos had labelled cells freehand drawing of the embryo. Control (unincubated) (Table 2), indicating that there was no significant loss embryos were injected at one to three sites along the axis. of injected progenitors during culture.
Table 1. Developmental stages reached by 6-7-day-old embryos after culture Somites
Initial stage time (h) n MS* LSt NPt HF§ 1-3 4-6 7-10
Prestreak 22 67 43 17 7 0 0 0 0 Early streak 22 93 12 37 31 13 0 0 0 Prestreak 44 69 0 0 0 26 14 28 1 Early streak 44 93 0 0 0 2 2 44 45
* Midstreak: mesoderm wings extend about two thirds around egg cylinder; posterior amniotic fold with developing exocoelom; head process not exposed at distal tip of egg cylinder. (The head process is the cranial extension of the primitive streak and is connected to the streak by the node, visible as a protuberance at the distal tip of the egg cylinder at late streak and later stages (Fig. 1C). For detailed description see Snell & Stevens (1966), Poelmann, (1981b).) t Late streak: mesoderm present in midline anteriorly; amnion closing; head process exposed at distal tip of egg cylinder but not more anteriorly. t Neural plate' amnion complete; allantois developing; edges of neural plate defined; head process exposed to surface over one- quarter length of the anterior half of the egg cylinder. § Head-fold: early neural folds and foregut invagination. No somites. 630 K. A. Lawson, R. A. Pedersen and S. van de Geer Table 2. Incidence of embryos with HRP-labelled cells in various germ layers after injection into a single axial endoderm cell and 22 h culture
No. with HRP-labelled cells
Endoderm plus Ectoderm/ Stage at Culture No. ectoderm/ mesoderm injection time (h) injected Total Endoderm mesoderm only
Prestreak + early streak 0 81 58 51 (88 %) 0 7(12%)
Prestreak 22 67 51 30 14 7 Early streak 22 93 66 58 5 3
Total 22 160 117 88(75%) 19(16%) 10(9%)
Table 3. Cells labelled/injection site(controls) HRP-labelled sites Sites injected Total One cell Two cells Three cells
Series 1 81 58 34 (58 %) 23 (40 %) 1 (2 %) Series 2 Anterior 47 34 16 16 2 Posterior 45 32 18 13 1 Total 92 66 34 (52 %) 29 (44 %) 3 (4 %)
Series 3 66 43 27 (63 %) 15 (35 %) 1(2%)
(3) Localization of labelled cells a horseshoe or crescent in the region of the presump- Of the 117 cultured embryos with HRP-labelled cells, tive foregut invagination (Fig. 3, zone IVB; Fig. 5). 75% had labelled cells only in endoderm, 16% in That endoderm cells originating in zone IV~ are endoderm plus ectoderm/mesoderm and 9 % in ecto- involved in foregut invagination is demonstrated by derm/mesoderm only (Table 2). For the purpose of the two initially most advanced embryos in the series analysis, the position of injection was classified on an that had been injected in zone IV, and in which the arbitrary scale: the length of the anterior-posterior foregut invagination was evident after 22h culture. axis of the embryo from its anterior limit at the Labelled cells were present in the floor of the invagi- nating foregut of one embryo (Fig. 6A); in the other, junction of embryonic and extraembryonic ectoderm slightly more advanced, embryo, which also had three was divided into five imaginary zones (Fig. 2); the labelled cells in the node, labelled cells were present relative position of the zones was the same in all at the rostral tip of the roof of the invaginating embryos, but the absolute size of the zones depended foregut (Fig. 6B). Cells over the main part of the on the size of the embryo. primitive streak (zone V) contributed endoderm descendants to the axis of the posterior half of the (a) Embryos with labelled endoderm only. Descen- embryo and to the posterior visceral yolk sac (Fig. 4, dants of cells situated on the axis of the anterior half zone VB). (zones I and II), including the distal tip (zone III), of The pattern of labelled cells seen in cultured early-streak-stage embryos were located over the embryos after labelling at the prestreak stage in zones most anterior and anterolateral part of the embryonic I, II and III was broadly similar to that obtained from ectoderm and neighbouring visceral yolk sac at late early streak stages (Fig. IB; Fig. 3, zones IA, IIA, streak and neural plate stages (Fig. 3, zones IB, IIB, and IIIA). However, descendants of zone IV cells and IIIB). They tended to be aligned at right angles to were found in the anterior half only, with consider- the embryonic axis. Endoderm descendants of sur- able anterolateral spread and overlap with those from face cells from, and near, the anterior end of the early zone III (Fig. 3, zone IVA), and zone V contributed primitive streak (zone IV) were found along the to the entire axis, resembling the behaviour of the entire axis of the embryo (Fig. 3, zone IVB), but cells of zone IV of the early-streak-stage embryo mainly in the anterior half (Fig. 1C), including the (Fig. 4, zone VA). The primitive streak is initiated in distal tip. These groups tended to be aligned along this most posterior region before spreading anteriorly the axis, skirting the head process and extending into into zone IV. Cell fate in mouse endoderm 631
These results indicate that there is a relative (b) Embryos with labelled cells in both endoderm and anterior shift of endoderm during gastrulation that ectoderm/mesoderm. The incidence of embryos with involves all axial endoderm back to, and including, labelled cells in both endoderm and ectoderm/meso- that covering the anterior portion of the primitive derm was significantly greater in the cultured em- streak, bryos than in the controls (Table 2) (jr2= 10-94,
v - y a
-'• i
\ •
1A 1
/7P V G
Fig. 1. Sections of injected embryos. (A) Longitudinal section of control embryo injected at three sites, showing one labelled endoderm cell (single arrow), two labelled adjacent endoderm cells (two arrows) and one labelled ectoderm cell (arrow head), a small part of which extrudes into the endoderm layer. (B) Near-sagittal section of midstreak-stage embryo that had been injected in zone III 22 h earlier at the prestreak stage, showing 5 of the 23 labelled endoderm descendants near the anterior limit of the embryo. (C) Sagittal section of neural-plate-stage embryo that had been injected in zone IV at the early-streak stage: two of the four labelled squamous endoderm cells (large arrows) lie over and immediately anterior to the head process, the anterior limit of which is indicated by a small arrow, a, anterior limit of embryo; am, amnion; ec, embryonic ectoderm; hp, head process; m, mesoderm; n, node at anterior end of primitive streak; ps, primitive streak. Bar, 100^m. 632 K. A. Lawson, R. A. Pedersen and S. van de Geer
Labelled mesoderm cells were found either at the anterior edge of the mesoderm (midstreak stage) or nearby (late streak/neural plate stage), or near the anterior end of the primitive streak. Labelled ecto- derm was found in the distal tip of the egg cylinder at the anterior end of the primitive streak. The results from embryos with label.in ectoderm/mesoderm only are shown in Fig. 8: after injection into early-streak- stage embryos, the labelled mesoderm cells tended to be more posteriorly situated than in embryos in which IV the endoderm was also labelled (Fig. 7); this differ- IV ence was not seen in the embryos injected at the prestreak stage. The positions of labelled cells in embryos with label in endoderm, both exclusively (Fig. 9) and in combination with ectoderm/meso- derm (Fig. 7), supplement the data from the earlier Fig. 2. Diagrammatic sagittal sections of egg cylinders experiments (Figs 3, 4) and emphasize the anterior without Reichert's membrane at the prestreak stage (A) and anterolateral shift of cells from the anterior zone and the early-streak stage (B) to show the position of the injection zones (I-V). The future cranial-caudal axis compared with the axial, mainly anterior, spread of runs from anterior (ant.) to posterior (post.) via the distal the posterior cells. tip of the egg cylinder, dots, embryonic endoderm; We conclude from these results that the posterior hatching, epiblast; broken line, primitive streak; eec, axial endoderm is heterogeneous during the time extraembryonic ectoderm; vee, visceral extraembryonic when bilateral symmetry is becoming visible: the endoderm. majority of the cells will contribute to the endoderm of the late gastrula, but a substantial minority will df=2, P<0-005), with a relatively larger contri- contribute to ectoderm/mesoderm. Cell pairs were bution by prestreak-stage embryos. However, pre- streak-stage embryos had been injected dispro- labelled in 40% of the injections (Table 3); although portionately often in posterior zones. To determine many of these pairs must have been sisters connected whether the presence of descendants in ectoderm/ by a cytoplasmic bridge, occasional labelling of two mesoderm was zone-dependent or stage-dependent, nonsister cells by inaccurate injection cannot be one axial endoderm cell in either the anterior half excluded. In addition, the labelled cells in a' few (zone II or junction I—II) or posterior half (zone IV or embryos will have descended from inadvertently junction IV-V) of the embryo was injected. Initial injected epiblast. Because of these uncertainties, the developmental stage in each litter was equalized over proportions of presumptive endoderm and ectoderm/ the two groups and the controls were taken at random mesoderm cells in the posterior axial endoderm from each group after injection. The number of cells cannot be estimated from the germ layer localization labelled per injection site in the controls (Table 3, of labelled descendants in the cultured embryos, nor series 2) and the germ layer distribution of these cells can the question be answered whether endoderm and (Table 4) were similar to those in the previous series. ectoderm/mesoderm can be derived from one cell in The incidence of cultured embryos with labelled cells the surface of the streak or only from adjacent, not not found exclusively in endoderm (i.e. endoderm necessarily sister, cells. plus ectoderm/mesoderm or ectoderm/mesoderm only) was increased after injection into posterior (c) Embryos with anomalous distribution of labelled endoderm compared with anterior endoderm cells. The positions of labelled cells in nine embryos (Table 4, c and d) and with controls; the frequency in the first series of experiments and three in the distribution for the anterior endoderm, however, did second series were anomalous and are not shown in not differ from the controls (Table 4, c and a). There the figures. Whereas most of these results could be was no significant difference in the behaviour of attributed to mistaken orientation of prestreak-stage posteriorly labelled cells between prestreak- and embryos or inadvertent injection into epiblast, two early-streak-stage embryos (Table 4, d] and d2). specimens could not be easily explained. Zone II of The localization of labelled cells in the embryos these embryos was injected at the early streak stage, with label in endoderm plus ectoderm/mesoderm and mistaken orientation was therefore unlikely; after injection posteriorly are summarized in Fig. 7: however, labelled mesoderm as well as labelled the results of embryos in this category from the first endoderm was found in zone I after culture. If only series of experiments (Table 2) have been included. endoderm had been injected, the labelled mesoderm Cell fate in mouse endoderm 633 cells must have either descended from anterior endo- (c) Cell death. The number of cultured embryos with derm or acquired evenly distributed cytoplasmic labelled cells was not less than that expected from the HRP by nonlineage transfer: both possibilities are controls (Tables 2, 4), which implies that, not only without precedent. If both epiblast and endoderm was there no direct toxic effect of HRP, but there was were labelled at injection, the mesoderm could have also no significant cell death in the axial endoderm descended from epiblast directly, without passing the population at prestreak and early streak stages. primitive streak. Such direct delamination of meso- derm from ectoderm occurs in utero and indicates the (d) Cell cycle distribution. The frequency distribution beginning of mesectoderm formation, which has of labelled cells/embryo indicates heterogeneity in already started by the neural plate stage in rodents the division rate of succeeding cell generations (Vermeij-Keers & Poelmann, 1980; Smits van (Fig. 10). In addition, interpretation of the frequency Prooije, 1986). distribution is complicated by the expectation from the controls that 42 % of the embryos initially had a (4) Cell numbers and population dynamics pair of labelled cells and 3 % had a triplet of labelled In theory, estimates of population doubling time, cell cells (Table 3): this will shift the distribution towards death and cell cycle distribution can be obtained from a greater number of descendants compared with a the quantitative data on labelled cells (Fig. 10). population derived from single labelled cells. So, when the data are classified into groups representing (a) Population growth. The total population increase number of cell generations (1, 2, 3-4, 5-8, etc., (IT) is obtained from the following relation (Lawson cells/embryo) (Table 5), the groups obtained are etal. 1986): 'mixed generation' classes. However, assuming that _ Total labelled cells (22 h) 'Total labelled cells (Oh) single and paired labelled cells will have the same cell r Embryos injected (22 h) Sites injected (Oh) cycle kinetics, the proportions of initial single and paired labelled cells in controls can be used to 1605 /68 + 104+12 calculate the frequency distribution of the number of 261 173 descendants from single labelled cells and, in the = 5-78. absence of cell death, the distribution of cell gener- ations (Table 6). By this estimate, only 7-6% of the (Data from Fig. 10; Tables 2, 3 and 4 [series 1 and 2].) total population failed to divide; the majority of cells Similarly, the population increase due to descend- were in, or had completed, a third cell cycle (Table 6) ants in the endoderm layer is given by after 22 h culture. Total labelled , Total labelled endoderm cells (22 h) 'endoderm cells (Oh) (e) Position-dependent variations. Stage- and pos- IF = Embryos injected (22 h)/ Sites injected (Oh) ition-dependent variations in cell cycle characteristics 1466/60+104+11 could also contribute to the spread in the frequency distribution of labelled cells (Fig. 10). There was no 261/ 173 significant difference between prestreak- and early- = 5-55. streak-stage embryos (data not shown), but the (b) Population doubling time. The population doub- number of labelled cells after culture did depend on ling time can be calculated from the number of the site of injection: posterior cells had more surviv- labelled cells. Assuming N = A-ebt, where N = ing descendants than did anterior cells (Fig. 11). number of labelled cells at t h, and N = 1 when t = 0. When the contribution of the total initial population Then the population doubling time (T) is given by to endoderm was examined, there was no difference in the number of descendants from anterior and ^ In2xt posterior sites (Fig. 12), indicating that the surface 1 lnN layer of endoderm expands uniformly. However, For the total population of labelled cells: when those embryos with descendants only in ecto- derm/mesoderm were omitted from the analysis and In2x22 Ti the contribution of the remaining initial sites to In 5-78 endoderm was examined, posterior sites contributed = 8-7h. significantly more descendants to the endoderm layer For the population in endoderm: than did anterior sites (Fig. 13). Therefore, the loss of surface cells to ectoderm7mesoderm in the streak In 2 x 22 T is balanced by a relatively large contribution from the In 5-55 remaining posterior axial endoderm cells to the = 8-9h. endoderm of midstreak-neural plate stages. 634 K. A. Lawson, R. A. Pedersen and S. van de Geer
A A post. ant post. ant post. post.
Ill
post. anL I \ \\\ p°sl
01 mm Cell fate in mouse endoderm 635
Standard deviations of the means of labelled cells labelled cells. When the descendants of cells con- were large (Figs 11-13) because the population ana- tributing only to endoderm were classified in 'mixed lysed was derived from a mixture of single and paired generation' classes (Table 5), the frequency distri- bution of the numbers of descendants of anterior cells (zones I—III) was different from that of descendants of posterior cells (zones IV and V) (^ = 12-02, df= 5, P<0-05). The population kinetics were therefore post. calculated for these two subpopulations, yielding population doubling times of 10-5 and 8-4 h for anterior and posterior regions, respectively (Table 6). Simplified estimators of population doubling times of the anterior and posterior regions (10-49 and 8-97 h, respectively) were used for statistical comparison and found to be significantly different (P< 0-001) (see Appendix). The remaining group of 48 embryos with labelled cells in ectoderm/mesoderm, both exclus- ively and with endoderm, had a frequency distri- bution of labelled cells in 'mixed generation' classes (Table 5) that was different from both anterior- derived endoderm 0^ = 27-16, df=6, P< 0-001) and posterior-derived endoderm (j2 = 15-45, df=6, P<0-02) and suggested that this part of the popu- lation was dividing more rapidly than the rest of the posterior endoderm. However, the data did not fit the
Fig. 4. Position of endoderm descendants after injection into single axial endoderm cells in zone V of (A) prestreak-stage and (B) early-streak-stage embryos. For further explanation, see legend to Fig. 2. Fig. 3. Position of endoderm descendants after injection into single axial endoderm cells in zones I-IV of (A) prestreak-stage and (B) early-streak-stage embryos. Injection position for individual embryos (one dot/labelled embryo) is indicated on the upper figure for each group. The location of descendants is projected onto a sagittal section (middle figure) and on the ventral surface of the flattened, embryonic part of the egg cylinder (lower figure). The anterior limit of the embryonic axis is indicated with a single arrow and the posterior limit with two arrows in the middle and lower figures. The anterior boundary of the neural plate and the wedge-shaped extension of the exposed head process are also represented in the lowest figure of B. The median position of the labelled cells of any one embryo is marked by a dot, the linear spread by a continuous line. Widely Fig. 5. Neural-plate-stage embryo that had been injected separated clumps of labelled cells in an embryo are in zone IV at the early-streak stage: seven of the eight marked by dots and connected by a broken line. labelled descendants (large arrow) are visible in a Overlapping positions on and near the midline of sagittal crescent shape in the region of the prospective foregut sections have been displaced outside the section for invagination. The small arrow indicates the anterior clarity. •' (II) indicates an embryo that was presumably boundary of the neural plate, am, amnion; n, node; misoriented at injection. ps, primitive streak. Bar, 100^m. 636 K. A. Lawson, R. A. Pedersen and S. van de Geer iterative model, possibly because the sample was too (a) Embryos with labelled endoderm only. Endoderm small and the population kinetics could not be calcu- of the anterior half of the axis (zones I—III) contri- lated. buted to the visceral yolk sac (Fig. 14, zones I—III; In conclusion, the axial endoderm displays pos- Figs 16A, 17A). The majority of zone IV cells of ition-associated heterogeneity in cell cycle kinetics prestreak embryos (6/8) also contributed to the during the first day of gastrulation. visceral yolk sac; the remainder had descendants in trunk endoderm (Fig. 15, zone IVA). The pro- (B) Embryos cultured for 2 days portions were reversed after injection at the early (1) Embryo development streak stage (*2 = 6-98, df=2, P<0-05): only a Prestreak-stage embryos reached head-fold-6-somite minority of zone IV cells (6/25) had descendants stages (Table 1) and 19 % had beating hearts: the exclusively in the visceral yolk sac; the majority majority of early-streak-stage embryos developed (17/25) contributed only to axial endoderm (Fig. 15, 5-8 somites; 80 % had beating hearts. zone IVB), with descendants at the anterior intestinal portal, in the ventral and dorsal foregut (Fig. 17B), (2) Localization of labelled cells and along the trunk and postnodal endoderm The controls did not differ significantly from those (Fig. 16B). Zone V contributed mainly to the pos- for 22h culture, either in number of labelled cells/ terior visceral yolk sac, but also to postnodal endo- injection site (Table 3, series 3) or in germ layer derm, and to axial endoderm after injection at the localization (Table 7). Of the 165 injected cultured prestreak stage (Fig. 15, zone V). Therefore, while embryos, 71 (43 %) had labelled cells, 85 % of these descendants of most of the axial endoderm cells of in endoderm only (Table 7). prestreak- and early-streak-stage embryos colonize
to.
Fig. 6. Foregut initiation. (A) Near-sagittal section of head- 6A fold-stage embryo that had been injected in zone IV 22 h earlier, showing two of the three labelled descendants (arrows) in the ventral endoderm of the very early foregut; 15 additional endoderm descendants were located posterior to the node. (B) Sagittal section of head-fold-stage embryo that had am been injected in zone IV 22 h earlier, showing two labelled cells (arrow) in the rostrodorsal endoderm of the invaginating foregut. Two additional labelled cells in adjacent sections were located in the base of the head process and five in postnodal endoderm. am, amnion; h, heart; hf, head fold; hp, head process; ps, primitive streak. Bar, 100^m. B Cell fate in mouse endoderm 637
Table 4. Incidence of embryos with HRP-labelled cells in various germ layers after injection into single anterior or posterior axial endoderm cells and 22 h culture
No. with HRP-labelled cells
Endoderm + Ectoderm/ Stage at Culture No. Endoderm ectoderm/ mesoderm Region injection time (h) injected Total only mesoderm only
Anterior Prestreak + 0 47 34 33 (97 %) 1(3%) 0 a early streak Posterior Prestreak + 0 45 32 31 (97 %) 0 1(3%) b early streak Anterior Prestreak 22 15 11 10 (91 %) 1 (9 %) 0 Early streak 22 36 24 21 (88 %) 2 (8 %) 1(4%) 51 35 31(89%) 3 (9 %) 1(2%) c
Posterior Prestreak 22 14 11 8 (73 %) 2(18%) 1 (9 %) d, Early streak 22 36 29 17 (59 %) 5 (17 %) 7 (24 %) d, Total 22 50 40 25 (62 %) 7 (18 %) 8 (20 %) d
cvd :tf l = 7-39, df=2, P<0-01. cvd .yC! = 2-04, df=2, P>0-l. z d,vd2:^ = 0-66,df=2,P>0-25.
A v
ant ant.
f
01 mm
Fig. 7. Position of labelled cells in embryos with label in both endoderm and ectoderm/mesoderm, after injection into single posterior axial endoderm cells of (A) prestreak-stage and (B) early-streak-stage embryos. The position of zone IV is indicated by a black strip. •, endoderm; D, ectoderm; •, mesoderm. For further explanation, see legend to Fig. 2. 638 K. A. Lawson, R. A. Pedersen and S. van de Geer (3) Cell numbers and population dynamics The frequency distribution is shown in Fig. 18. In contrast to embryos after 22h culture, a significant ant. proportion had no labelled cells. From the controls, the expected number of successfully injected embryos would be: 43x165/66=107, (Table 7) but only 71 (66%) were found with labelled cells after 44h. Therefore, either all the descendants of 34% of the successfully injected cells died during the second half of the culture period, or the HRP concentration was below the detection level in these descendants and, possibly, in some cells of other embryos. Many axial endoderm cells die between midstreak and head-fold stages (Poelmann, 1980; Lawson et al. 1986), but it is not known whether cell death is lineage determined. The equation for population increase (see section A4a) gives an unbiased estimate even in the presence of cell death, provided all surviving descendants are detected. According to the data in Fig. 17, Table 3 (series 3), and Table 7, the popu- lation increase in the endoderm layer during 44 h was 1041 /16 + 30 + 3 = 7-06. 165, 66 Since the population increase due to descendants remaining in endoderm during the first 22 h was 5-55, this implies that the increase during the following 22 h 01 mm was only 27 %. The population increase of endoderm Fig. 8. Position of labelled ectoderm and mesoderm cells of 7-5-day-old embryos (midstreak to neural plate in embryos with no label in endoderm, after injection stages) was 101 % during 24 h culture to early somite into single posterior axial endoderm cells of (A) stages (Lawson etal. 1986); although 6-7 day embryos prestreak-stage and (B) early-streak-stage embryos. The may be growing more slowly during their second day position of zone IV is indicated by a black strip. of culture, the discrepancy suggests that at least some D, ectoderm; •, mesoderm. For further explanation see endoderm descendants of labelled cells were not legend to Fig. 2. detected after 44h culture.
Discussion the visceral yolk sac 44 h later, a relatively short stretch of endoderm in the region of the anterior end HRP has a distinct but limited usefulness as a lineage of the early primitive streak contributes to embryonic marker in the postimplantation mouse embryo. HRP was injected intracellularly to analyse endoderm fate endoderm at early somite stages. in situ, on the assumption that the technique would Labelled descendants in the visceral yolk sac were not interfere with the behaviour of the system being aligned parallel to the equator of the conceptus studied. We found no evidence of interference: suc- (Figs 14, 15) or as a loosely coherent clump cessfully injected cells did not die within the first cell (Fig. 16A); those in the embryo tended to be aligned cycle, no effect was found on the number of cells in along the embryonic axis (Fig. 15) and did not form a either S phase or mitosis 10 h after injection (Lawson coherent patch (Fig. 16B). et al. 1986), and the majority of injected cells went through three cell cycles during the first 22 h. The (b) Embryos with labelled ectoderm/mesoderm. Of results from embryos cultured for 44 h, however, the eleven embryos with labelled cells in ectoderm/ indicate that while the qualitative data from these mesoderm, seven had been injected in zone IV at the embryos are reliable, the quantitative data may be early streak stage; four prestreak-stage embryos had biased because labelling in some samples was not been injected in zones II, IV (two embryos) and V. resolvable. Up to 64-fold dilutions of intracellularly The sample was too small to draw conclusions about injected HRP can be detected reliably (Lawson et al. the distribution of labelled descendants. 1986): it should therefore be possible to follow Cell fate in mouse endoderm 639
Anterior Posterior
ant.
0-1 mm
Fig. 9. Position of endoderm descendants after injection into single axial endoderm cells of anterior and posterior regions of (A) prestreak-stage and (B) early-streak-stage embryos. The positions of zone II (anterior) and zone IV (posterior) are indicated by a black strip. For further explanation see legend to Fig. 2.
10 12 14 16 18 20 22 24 26 28 30 32 34 36 HRP-labelled cells/embryo
Fig. 10. Frequency distribution of the number of HRP-labelled cells/embryo after 22 h culture {n = 191). Open blocks, embryos with labelled endoderm only; hatched blocks, embryos with labelled endoderm and ectoderm/mesoderm; cross- hatched blocks, embryos with labelled ectoderm/mesoderm only. 0' represents the number of injected embryos without labelled cells corrected for controls. 640 K. A. Lawson, R. A. Pedersen and S. van de Geer
Table 5. Frequency distribution of the number of labelled cells/embryo after 22 h culture according to site of injection and germ layer position of descendants
Germ Total HRP-labelled cells/embryo* Injection layer no. of zones descendants embryos 1 2 3-4 5-8 9-16 17-32 33-64
All All 192 8 (4%) 18 (9%) 27 (14%) 69 (36 %) 52 (27 %) 17 (9%) 1 (1%) I,II,III Endoderm only 66 1 (2%) 11 (17%) 16 (24%) 24 (36 %) 12(18%) 2 (3%) 0 IV,V Endoderm only 78 5 (6%) 5 (6%) 11 (14%) 25 (32 %) 27 (35 %) 5 (6%) 0 IV,V Ectoderm/mesoderm 48 2 (4%) 2 (4%) 0 20 (42 %) 13(27%) 10 (21 %) 1 (2%) + Endoderm
*The data have been grouped in classes according to cell generation ('mixed generation' classes). Each class, except 1, contains a mixture of descendants of initially single cells and cell pairs; e.g. class 5-8 contains third generation descendants of labelled single cells and second generation descendants of labelled cell pairs.
Table 6. Population kinetics derived from the calculated number of descendants of single labelled cells during 22 h culture
Germ Total Percentage distribution over 0-5 cell generations * Population Injection layer no. of - Population doubling zones descendants embryos 0 1 2 3 4 5 increase time (h)
All All 192 7-6 11-3 16-6 52-2 8-5 3-7 5-98 8-5 I,II,III Endoderm only 66 2-8 28-3 22-4 46-5 0 0 4-30 10-5 IV,V Endoderm only 78 11-7 2-8 22-9 40-8 21-9 0 610 8-4
*The fraction with descendants traversing each generation was calculated as follows, assuming that control and incubated populations had the same initial frequencies of single, double and triple labelled cells, and that the probability of division was the same regardless of the initial number of labelled cells. The number of undivided single cells was obtained from the cultured embryos with only one labelled cell. This figure was applied to the proportion of cell pairs to singletons in the initial population to estimate the number of cultured embryos with two labelled cells that was due to initial cell pairs that had not divided and, hence, by subtraction, to estimate the number of single labelled cells that had divided once. The process was iterated on the succeeding 'mixed generation" classes to obtain the entire distribution for single labelled cells and, hence, the population increase, population doubling time and (assuming no cell death) percentage distribution of cell generations. descendants through six generations if only dilution is detected for a relatively brief period (1-2 days) involved, but additional metabolic degradation or during postimplantation development in the mouse. intracellular segregation of the enzyme would reduce It is, however, reliable for short-term lineage studies this sensitivity. The generation time of embryonic and for analysing morphogenetic movement and endoderm at the head-fold stage in cultured embryos population kinetics over three to five cell generations. is maximally 11-5 h (Lawson et al. 1986), so that at least five cell cycles will have been completed by the Morphogenetic movement dividing population of 6-7-day-old embryos after 2 The position and alignment of endoderm descendants days. Even a small proportion of faster dividing cells, relative to the site of injection along the embryonic or an uneven distribution of cytoplasmic HRP be- axis, together with a comparison of embryos injected tween sister cells at this dilution, could result in at prestreak and early streak stages and the hetero- failure to detect descendants and cause a bias in the geneity of cell fate in the posterior endoderm, led us frequency distribution of labelled cells. The detection to the following interpretation. At the time the level for injected cells is therefore probably higher posterior ectoderm thickens to produce the primitive than that expected from dilution alone. This means streak and overt bilateral symmetry, a subpopulation that, compared with amphibians (Jacobson & Hirose, of cells appears in the axial endoderm of the posterior 1978; Hirose & Jacobson, 1979; Heasman, Wylie, half of the embryo. Some of these surface cells, or Hausen & Smith, 1984; Masho & Kubota, 1986), fish their progeny, move internally, contributing to ecto- (Kimmel & Law, 1985; Kimmel & Warga, 1986), a derm and mesoderm. Others, concentrated near the variety of invertebrates (Weisblat et al. 1978; Komi- anterior end of the early streak, spread anteriorly, nami, 1983; Nishida & Satoh, 1983; Taghert, Doe & either displacing or replacing the axial endoderm Goodman, 1984) and murine preimplantation stages from the anterior half of the embryo. The displaced (Balakier & Pedersen, 1982; Cruz & Pedersen, 1985; anterior endoderm itself shifts anteriorly and antero- Pedersen, Wu & Balakier, 1986), HRP can be laterally towards, and partly onto, the visceral yolk Cell fate in mouse endoderm 641
20-
o &• 10- 1 9-
15 f = 3H
2-
12 3 4 5 2 3 4 5 a-p axis (relative units) a-p axis (relative units)
Fig. 11. Total HRP-labelled cells/embryo after 22 h Fig. 12. HRP-labelled endoderm cells/embryo after 22 h plotted against position of the injection site on the culture, plotted against position of the injection site on anterior-posterior axis. Bars, mean ± S.D. Statistical the anterior-posterior axis. To make the log scale significance of linear regression: F= 9-84, df= 1,188, analysis possible, sites contributing exclusively to P< 0-005. ectoderm/mesoderm were assumed to have one descendant in endoderm. Bars, mean ± S.D. NO sac. The posterior-derived endoderm cells, spreading regression. forward over the distal tip of the egg cylinder, avoid or are themselves displaced by the head process, position of labelled cells 2 ddys later (Fig. 19B): there which begins to insert into the endoderm layer at the is a border in the endoderm between cells that will node at the late-streak stage and becomes progress- populate the anterior and posterior regions of the ively more exposed anteriorly. Endoderm descend- yolk sac; the border area contains cells that contrib- ants from the posterior half of the streak spread along ute to the entire axial endoderm of the early somite the axis but remain posterior to the node; some move embryo. posteriorly onto the visceral yolk sac. Thus, before Although this picture is incomplete, it is strikingly foregut invagination has begun, most of the embry- similar to the detailed map of morphogenetic move- onic axis is occupied by endoderm of posterior origin, ment of endoderm in the avian embryo (Spratt & partially bisected by the head process. Haas, I960; Vakaet, 1962, 1970; Nicolet, 1971; During the following day, the anterior-derived Rosenquist, 1972), where endoderm expands, in- endoderm cells are displaced further anteriorly and itially anteriorly, from the anterior end of the early anterolaterally onto the yolk sac. The anterior shift primitive streak, displacing the hypoblast towards the along the axis continues between the midstreak and margin of the area pellucida and then invaginates to neural plate stages (Lawson et al. 1986) and, by the form the foregut of the early somite embryo (Rosen- time somites are forming, the posterior-derived cells quist, 1966, 1972). This cell behaviour may be a are located at the anterior intestinal portal and have morphogenetic phenomenon common to both birds been incorporated in the foregut, trunk and postno- and mammals, for which the difference between the dal endoderm. In addition, midstreak- and late- flat blastoderm of the chick and the cup-shaped streak-stage postnodal endoderm shifts posteriorly embryo peculiar to some rodents is a geometric towards and onto the visceral yolk sac. The expansion irrelevance. of endoderm from the anterior end of the primitive Since endoderm expands from the anterior end of streak after the early-streak stage is clearly illustrated the primitive streak, local differences in cell behav- when the site of injection is classified according to the iour could be involved in the expansion. Posterior 642 K. A. Lawson, R. A. Pedersen and 5. van de Geer
endoderm cells have more surviving descendants Since the regression disappeared when the contri- after 24 h than do anterior cells, but a proportion of bution of the total initial population to endoderm these contribute to ectoderm/mesodenn and so are only was considered, the results represent the behav- lost to the surface layer. However, the linear re- iour of a surface population with a uniform expansion gression of endoderm descendants on the position of in the face of a drain posteriorly to ectoderm/meso- ancestors along the axis remained when the part of derm, i.e. they do not account for the anterior the population contributing only to ectoderm/meso- displacement of endoderm. Even if the exit of cells denn was excluded, indicating that posterior endo- from the surface to deeper layers is counteracted by derm cells contribute more descendants to late- entry of new cells to the surface, this would allow streak-stage endoderm than do anterior cells: the expansion from the point of entry but would not calculated population doubling time in zones IV and impose directionality on the expanding layer. V was 8-4h, compared with 10-5h in zones I—III. The end of the first 22 h culture period of early- streak-stage embryos coincides with a stage when there is substantial cell death in the axial endoderm (Poelmann, 19816) and the number of cells dying is 20' greater anteriorly than postnodally (Lawson et al. 1986). The head process inserts into the endoderm layer at this time. Although cell death may create additional space anteriorly that other posterior- 10. derived cells can occupy, it is unlikely to be the main 9 source of displacement, since endoderm displace- 8- ment occurs before the midstreak stage and there is T 6' no cell death at the early-streak stage.
5- Endoderm displacement could be passive, reflect- ing morphogenetic movement of the underlying 4- ectoderm. If so, the position of labelled endoderm descendants implies anterior displacement of axial =?. 3- ectoderm in the anterior half of the egg cylinder and 0. posterolateral shift at the anterior end of the embryo. OS X 2- In addition, cells contributing to the primitive streak would elongate the streak part of the axis, initially anteriorly and, later, both anteriorly and posteriorly. In the rabbit epiblast, lateral and anterolateral cells move towards the primitive streak along the rim of 12 3 4 5 the embryonic shield (Daniel & Olson, 1966) and a-p axis (relative units) similar movements occur in the avian epiblast (Spratt & Haas, 1965; Vakaet, 1984); the anterolateral align- Fig. 13. HRP-labelled endoderm cells/embryo after 22 h ment of descendants of anterior endoderm in the culture, plotted against position of the injection site on mouse is consistent with such a movement. Forward the anterior-posterior axis, excluding embryos with sites contributing only to ectoderm/mesodenn. Bars, expansion of axial cells in the epiblast between the mean ± S.D. Statistical significance of linear regression: anterior end of the streak and the anterior end of the F= 4-55, df= 1,172, P<0-05. embryo could be produced by high proliferative
Table 7. Incidence of embryos with HRP-labelled cells in various germ layers after injection into a single axial endoderm cell and 44 h culture
No. with HRP-labelled cells
Endoderm + Ectoderm/ Stage at Culture No. Endoderm ectoderm/ mesoderm injection "time (h) injected Total only mesoderm only
Prestreak + early streak 0 66 43 42(98%) 0 1(2%)
Prestreak 44 65 27 23 2 2 Early streak 44 100 44 37 3 4 Total 44 165 71 60(85%) 5(7%) 6(8%) Cell fate in mouse endoderm 643
II
ant
III
ant. ant.
Fig. 14. Position of endoderm descendants 44 h after injection into single axial endoderm cells of zones I—III of (A) prestreak-stage and (B) early-streak-stage embryos. The location of descendants is projected onto a sagittal section (middle figure) and onto the ventral surface of a flattened, early somite embryo. O, notochord. For further explanation 0-1 mm see legend to Fig. 2. 644 K. A. Lawson, R. A. Pedersen and S. van de Geer activity at or near the cranial end of the primitive streak (Daniel & Olson, 1966; Snow, 1977, 1978). The cellular correspondence between the two layers is, however, less close than would be expected if they were behaving as a unit. First, injections of HRP into the distal tip (zone III) of prestreak-stage- egg cylinders, made in such a way that both ectoderm and endoderm were labelled, showed a slight anterior shift of ectoderm descendants, but this was clearly less than the anterior shift of endoderm (J. J. Meneses, personal communication). Second, an indi- cation of the behaviour of ectoderm cells in the streak compared with the overlying endoderm was given in the present experiments by embryos with labelled cells in both ectoderm and endoderm. Of nine such embryos, injected in zones IV and V (Fig. 7), labelled ectoderm cells were later found at the distal tip of the embryo, at the anterior end of the streak, i.e. the labelled cells had moved forward with the extending streak but retained their original position relative to the anterior end of the streak. In contrast, the labelled endoderm cells of eight of these embryos were situated far anterior, and those of one embryo just anterior, to the labelled ectoderm. Third, although the most proximal endoderm will be freed from association with embryonic ectoderm when the latter contributes to forming the anterior and pos- terior amniotic folds (Snell & Stevens, 1966), the eventual colonization of the yolk sac by even the most distal endoderm makes a close cellular corre- spondence of endoderm and ectoderm movements ant. ant. unlikely. Therefore, while the direction of endoderm expan- sion and the alignment of endoderm descendants may reflect morphogenetic movement in the underlying ectoderm in the anterior half of the embryo and growth of the primitive streak in the posterior half, the expansion per se appears to be a property of the endoderm layer.
Cell fate and cell lineage Except for a limited stretch of posterior endoderm, descendants of cells from all zones along the axis of prestreak and early-streak embryos were found exclusively in the yolk sac of early-somite-stage em- bryos 2 days later. Since only descendants of primi- tive endoderm have been found in yolk sac endoderm at midgestation (Gardner & Papaioannou, 1975; Gardner & Rossant, 1979) and visceral embryonic
Fig. 15. Position of endoderm descendants 44 h after injection into single axial endoderm cells of zones IV and V of (A) prestreak-stage and (B) early-streak-stage embryos. The surfaces of the ventral (v) and dorsal (d) foregut and dorsal hindgut (d1) are shown separately. A, blood island. For further explanation, see legends to 01 mm Figs 2 and 9. Cell fate in mouse endoderm 645
\
16A B
Fig. 16. Embryos cultured for 44 h. (A) 6-somite embryo injected in zone I at the early streak stage has 12 labelled cells (arrow) in the yolk sac endoderm. (B) 6-somite embryo injected in zone IV at the early streak stage. The distribution of the 64 labelled cells is indicated by arrows. A fairly coherent strip of 43 cells is located caudally over the embryo and spreads onto the yolk sac. The remaining 21 cells are scattered along the axis of the trunk, the right side of the anterior intestinal portal, and in the foregut; aip, anterior intestinal portal; a/, allantois; h, heart; mb, mesencephalon; s, somite; ys, visceral yolk sac. Bar, 200 jim. endoderm shows the same behaviour as visceral part of the primitive endoderm lineage. If, on the extraembryonic endoderm both as donor in blasto- other hand, this new population is inserted into the cyst chimaeras (Rossant, Gardner & Alexandre, endoderm layer from the epiblast, it could contain 1978; Gardner, 1982) and in vitro (Hogan & Tilly, the ancestors of the fetal endoderm: the descendants 1981), it seems likely that most of the axial endoderm in the early somite embryo would contribute perma- at the onset of gastrulation is visceral embryonic nently, and not transitorily, to the developing gut. endoderm derived from primitive endoderm. The arguments against an epiblast origin of the The presence of descendants of early-streak stage, posterior-derived endoderm originate in the results of but rarely prestreak stage, posterior endoderm in ectopic transplants. In such experiments (Grobstein, embryonic endoderm 2 days later indicates that a new 1952; Levak-Svajger & Svajger, 1971, 1974; Diwan & population emerges in the axial endoderm at the time Stevens, 1976; Beddington, 1983a), derivatives of of primitive streak formation. The descendants of this definitive endoderm were formed by epiblast from population, which occupies a relatively short stretch prestreak to late-streak stages, while endoderm alone of endoderm over the anterior part of the early failed to develop or formed parietal endoderm. Only primitive streak, were spread along the entire embry- at the head-fold stage was the capacity to form onic axis by the neural plate stage and contributed to endoderm derivatives no longer present in ectoderm, embryonic endoderm, including the foregut, at early but it was present in transplants of mesoderm plus somite stages. The two possible sources for this endoderm (Svajger & Levak-Svajger, 1974). While population are nonaxial endoderm and epiblast. If these experiments illustrate the potency of the epi- nonaxial visceral embryonic endoderm moved into blast, they do not necessarily pinpoint the normal fate the region of the primitive streak at the beginning of of the transplanted cells (Beddington, 1981, 1982, gastrulation and then spread along the axis, the 1983a,fe) or that of the isolated endoderm that failed descendants of these cells would make no permanent to develop in an ectopic site. A second line of contribution to the developing gut, since they form argument comes from histological studies, the 646 K. A. Lawson, R. A. Pedersen and S. van de Geer
-*»zs: ?
a/77 • .-> •'V.
v ,#.» ys
••*
17A B
Fig. 17. Sections of embryos cultured for 44h. (A) 9-somite embryo injected in zone II at the early primitive streak stage, showing 7 of the 19 labelled cells in the visceral yolk sac endoderm. The honey-combed appearance of the cell apices (arrow) is due to large endocytotic vacuoles that do not contain HRP. (B) Transverse section of the 6-somite embryo shown in Fig. 15B, which had been injected in zone IV, showing labelled cells in dorsolateral and ventrolateral foregut (arrows), am, amnion;/g, foregut; h, heart; ne, neurectoderm;>\s, visceral yolk sac. Bar, 100jim.
o 10-
o
0 2 4 6 8 10 12 14 16 18 20 22 24 26 28 30 32 34 38 42 45 51 55 64 HRP-labelled cells/embryo
Fig. 18. Frequency distribution of the number of HRP-labelled cells/embryo after 44h culture, n = 71. For further explanation, see legend to Fig. 9. authors of which have concluded that the head endoderm layer at midstreak to late-streak stages process is the main or sole source of the fetal (Lawson et al. 1986). endoderm (Jolly & Fe"rester-Tadie\ 1936; Snell & The arguments for an epiblast origin of the pos- Stevens, 1966; Poelmann, 1981). However, such mor- terior-derived endoderm are as follows. First, the phological studies do not take into account the presence of mesoderm and ectoderm descendants heterogeneity of origin and fate of cells in the from posterior endoderm indicates that cells can Cell fate in mouse endoderm 647
also forming the mesoderm layer and (2) the flat- tened, squamous endoderm will cover a greater surface area than the columnar epiblast (e.g. 2 am. 1000 fim of basal lamina would be occupied by 10-14 epiblast cells (more if the mitotic cells at the lumen Yolk sac are included), but only by 2-3 squamous endoderm cells). Fourth, the behaviour of endoderm at mid- streak and late-streak stages strikingly resembles that in the chick embryo (Lawson et al. 1986); earlier in avian gastrulation the epiblast-derived definitive en- doderm inserts into the hypoblast through the primi- tive streak (for reviews see Nicolet, 1971; Bellairs, 1982, 1986) and its subsequent behaviour (Rosen- quist, 1966, 1971, 1972) is similar, if not identical, to post. ant. that of the posterior-derived endoderm in the mouse described here. Finally, an analysis of the prestreak epiblast has revealed that the epiblast associated with zone IV, and a slightly larger region in early-streak- stage embryos, does indeed have descendants in endoderm at midstreak to neural plate stages (K. A. Lawson, unpublished data). Fig. 19. Site of injection at (A) prestreak stage and (B) On the basis of the information available, it seems early-streak stage, classified according to position of most likely that epiblast derivatives are first inserted endoderm descendants 44 h later. •, anterior and into the endoderm layer very early in gastrulation at anterolateral yolk sac; O, posterior and posterolateral the anterior end of the primitive streak. Descendants yolk sac; T, anterior intestinal portal; V, ventral foregut; •, dorsal foregut; D, trunk; A, postnodal of these cells would be incorporated into the ventral endoderm; brackets, descendants in different embryonic foregut and anterior intestinal portal, whereas the regions in the same embryo; ', additional descendants in descendants of cells that emerged slightly later would postnodal endoderm or yolk sac endoderm. maintain a more axial position and colonize the dorsal foregut. The head process, the cranial extension of exchange between epiblast and endoderm as the the anterior end of the primitive streak, has begun to primitive streak is forming. Although we have not form by the midstreak stage but becomes progress- shown conclusively that labelled descendants in endo- ively incorporated into the endoderm layer from the derm and ectoderm/mesoderm were derived from late-streak stage onwards: the axial part forms noto- one cell or a pair of sister cells, it is unlikely that the chord and endoderm of the trunk (midgut). Descend- 18 % of embryos in this category were the result of ants from lateral cells in the head process may be injection into two neighbouring but unrelated cells. added later to foregut endoderm, just before the head Second, there is virtually no overlap of clonal de- folds develop and the foregut invaginates (Jolly & scendants colonizing yolk sac and embryonic endo- Ferester-Tadie\ 1936; Snell & Stevens, 1966; Poel- derm; the only exceptions were 4 of the 63 embryos mann, 1981). cultured for 44 h, and these had labelled descendants spreading from the most caudal or laterocaudal part Heterogeneity of the posterior endoderm of the embryo onto the posterior yolk sac: more The presence of cells in the posterior axial endoderm prolonged culture would be necessary to resolve the that have descendants in ectoderm/mesoderm indi- fate of these descendants. Third, the expansion of cates that the separation of the epiblast and embry- endoderm from the anterior end of the primitive onic endoderm is no longer complete when the streak is greater than can be accounted for by the primitive streak begins to form. Formation of the morphogenetic movement of the ectoderm alone, primitive streak in mammals is associated with (a) an although the generation time in the endoderm is increase in the frequency of mitotic spindles oriented longer than in the epiblast: 8-4 h in the posterior- perpendicular to the cell sheet in the streak region, derived endoderm compared with 7-6h (Poelmann, thus producing several cell layers (Snow & Bennett, 1980), 6-25 h (Solter, Skreb & Damjanov, 1971), 1978), (b) extensive disruption of the basal lamina in 4-4-6-7 h (Snow, 1977) and 3-7-4-2h (Lewis & Ros- the streak region, which is much greater than the sant, 1982) in vivo and 7-5 h in vitro (K. A. Lawson, local discontinuities observed before streak forma- unpublished data) in the epiblast. This argument, tion and in lateral regions thereafter (Poelmann, however, is not compelling since (1) the epiblast is 1981; Takeuchi & Takeuchi, 1981; Franke, Grund, 648 K. A. Lawson, R. A. Pedersen and S. van de Geer
Jackson & Illmansee, 1983), (c) increase in mem- variation, as well as of the size of the dividing brane specialization (adhesive plaques, gap junctions population. and nuclear pores) in the epiblast cells (Batten & Population doubling time and generation time are Haar, 1979) and (d) acquisition of vimentin and loss only equivalent when all cells are dividing and there is of cytokeratins by emerging primary mesenchyme no cell death. The population doubling time of 8-7 h is cells (Franke, Grund, Kuhn, Jackson & Illmansee, therefore a maximum estimate of generation time. It 1982). In addition, it must be supposed that cell is markedly shorter than that of 16-6 h based on cell contacts within the endoderm layer (apical tight counts between 6-5 and 7-5 days in vivo (Snow, 1977). junctions) (Batten & Haar, 1979) and desmosomes The discrepancy can be explained by the fact that (Solter, Damjanov & Skreb, 1970; Franke etal. 1983) Snow counted cells contained within borders set by are sufficiently unstable at this stage to allow depar- the embryonic ectoderm, while we counted the de- ture of cells. Our interpretation of the phenomenon is scendants of a sample of cells only initially within that a flux of epiblast- derived cells into the endoderm these borders. Since descendants of zone I and some layer coincides with streak formation and that some zone V cells will have moved onto the yolk sac within cells, or their descendants, are released back into the 24 h, and some from zones IV and V will have passed mesoderm or ectoderm via the streak. Alternatively, into the interior of the embryo, the population disruption of the basal lamina could be a sufficient doubling time within the boundaries set by the condition for visceral embryonic endoderm cells to embryonic ectoderm will be longer than the gener- leave the surface layer and temporarily contribute to ation time of cells initially within these boundaries, mesoderm and ectoderm. This seems to be unlikely, even if loss to the interior is compensated by insertion since caudal endoderm cells grafted caudally into the of new cells from the epiblast. On the other hand, primitive streak of 7-5-day embryos do not incorpor- HRP injections were limited to the axial endoderm ate into embryonic structures (Copp, Roberts & and may not be representative of the rest of the Polani, 1986). endoderm. The calculated population doubling time for cells that had exclusively endoderm descendants Population kinetics varied even along the axis, being 10-5 h in the anterior The frequency distribution of labelled cells after 22 h cells and 8-4 h in the posterior cells. The value for (Fig. 10) indicates lack of synchrony in the cell cycles anterior endoderm agrees with the cycle time of of succeeding generations, but could also reflect 10-7h in visceral extraembryonic endoderm during attrition due to cell death. There was no indication the same period in vivo, using the labelled mitoses that labelled progenitors died at the prestreak or method (Solter & Skreb, 1968) but is longer than that early streak stage or of loss of all labelled descendants of combined visceral embryonic and extraembryonic in some embryos: both phenomena would have led endoderm calculated from colcemid-blocked mitoses to an increase in the number of embryos without (6-6h) (Lewis & Rossant, 1982). labelled cells. It is less easy to assess the loss of The population kinetics of axial endoderm in vitro subpopulations of labelled descendants, but import- during the first day of gastrulation contrast with those ant in the present context, since there is significant in the following 24h, when the population doubling cell death in axial endoderm injected at midstreak time increases to 23-9h (Lawson et al. 1986): about and later stages (Lawson et al. 1986). Cell death of half the axial endoderm cells die during the time that some descendants would not bias the calculated the head process inserts into the surface layer; the frequency distribution of descendants of single cells, generation time of the surviving cells is 11-5 h or less. assuming that the chance of cell death was the same Thus, intracellular injection of HRP provides con- for the descendants of singletons and pairs: this siderable information about cell population kinetics calculation showed that more than 90% of the that would not be obtained from other, non-clonal population were dividing. Further classification into approaches. cell generations, however, assumes there is no cell death and the classification will be distorted if half or more of a generation dies and the effect of cell death Appendix by Sara van de Geer is compounded by heterogeneity in cell cycle length. The data so classified (Table 6) are compatible with Statistical comparison of population doubling times some cell death in anterior endoderm during the The simplified estimator of the population doubling second cycle after injection, at the earliest, and in the time is based on the following assumptions: posterior endoderm during the third cycle. Complete (i) the distribution of labelled cells is the same for all analysis of population growth requires independent embryos within a cultured group or within the control estimation of cell death and cell cycle length and group, Cell fate in mouse endoderm 649
(ii) the distribution of the initial number of labelled Let population 1 and 2 be the embryos with cells is the same for the control group and the labelled endoderm descendants after injection into cultured groups, zones I, II, III and IV, V, respectively, Let ^ denote (iii) at time t (0=£t=£22h) the expected number of the expected number of labelled cells in an embryo of labelled cells per embryo given the initial number is population i at time t = 22h, and let T, denote the equal to population doubling time, i = 1,2. Then Ai = 6-348 A2 = 8-115 where b depends on the culture under consideration var (£,) = 0-312 var (fh) = 0-378 and j = 1, 2, 3,... is the initial number. T, = 10-491 = 8-974 t2 Let p and y. denote the expected number of labelled var (f,) = 0-463 var (T,) = 0-192 cells at time t = 0 and t = 22h, respectively. Under assumption (iii), The control values for series 1 and 2 (Table 3) were pooled to give one control group with Thus the doubling time is p = 1-484, var (p) = 0-003. 22 In 2 T = We use formula (*) to estimate the covariance In between Tj and T2:
Let COV total number of labelled cells in the control group P = number of labelled sites in the control group and The estimated correlation between T^ and T2 is total number of labelled cells in the cultured group ^ cover, ,f2) =niA7 £ = number of labelled embryos in the cultured group Vvar (f J var (t2)
Under assumption (i) p estimates the expected If T, = T2, then initial number of labelled cells in a site of the control group. Assumption (ii) ensures that p is also an T estimator of p, the number of labelled cells in the cultured group at time t = 0. We estimate T by is approximately normally distributed with expec- 22 In 2 tation zero and unit variance. (If two independent T = control groups are used, there is no covariance and V2(l-/&) reduces to V2) We have The estimated variances of p and fi can be calcu- lated in the usual way: -1-96 V2JF [average of (number of labelled cells in control site)2]-f>2 var (p) = number of labelled sites in the control group For t we find the value average of (number of labelled cells in cultured embryo)2]—fi2 var(/j) = " t = -3-86< -1-96. number of labelled embryos in the cultured group A Taylor expansion of T around (p, JX) gives Thus the hypothesis Tj = T2 can be rejected at the 5% level. t 221n2_ fp-p A~^ P A* I We thank Jenny Narraway for histology, Leen Boom and Carmen Kroon-Lobo for photographic and art work, and + smaller order terms. Mary McKinney for editorial assistance. We are indebted to Richard Gill of the Centre of Mathematics and Computer Hence, the variance of T can be approximated by Science, Amsterdam, for advice on the statistical compari- son of population doubling times. 22 In 2 \" I van var (f) = This work was partially supported by the Office of Health and Environmental Research, US Department of Energy, contract no. DE-AC03-76-SF01012 and by the Hubrecht Fund. 650 K. A. Lawson, R. A. Pedersen and S. van de Geer
References elements during mouse embryogenesis. III. Primary mesenchymal cells and the first appearance of vimentin BALAKIER, H. & PEDERSEN, R. A. (1982). Allocation of filaments. Differentiation 23, 43-59. cells to inner cell mass and trophectoderm lineages in GARDNER, R. L. (1978). The relation between cell lineage preimplantation mouse embryos. Devi Biol. 90, and differentiation in the early mouse embryo. In 352-362. Genetic Mosaics and Cell Differentiation (ed. W. J. BATTEN, B. E. & HAAR, J. L. (1979). Fine structural Gehring), pp. 205-241. Heidelberg: Springer-Verlag. differentiation of germ layers in the mouse at the time GARDNER, R. L. (1982). Investigation of cell lineage and of mesoderm formation. Anat. Rec. 194, 125-142. differentiation in the extraembryonic endoderm of the BEDDINGTON, R. S. P. (1981). An autoradiographic mouse embryo. J. Embryol. exp. Morph. 68, 175-198. analysis of the potency of embryonic ectoderm in the GARDNER, R. L. (1984). An in situ cell marker for clonal 8th day postimplantation mouse embryo. J. Embryol. analysis of development of the extraembryonic exp. Morph. 64, 87-104. endoderm in the mouse. J. Embryol. exp. Morph. 80, BEDDINGTON, R. S. P. (1982). An autoradiographic 251-288. analysis of tissue potency in different regions of the GARDNER, R. L. & PAPAIOANNOU, V. E. (1975). embryonic ectoderm during gastrulation in the mouse. Differentiation in the trophectoderm and inner cell J. Embryol. exp. Morph. 69, 265-285. mass. In The Early Development of Mammals (ed. M. BEDDINGTON, R. S. P. (1983a). Histogenetic and Balls & A. E. Wild), pp. 107-132. Cambridge: neoplastic potential of different regions of the mouse Cambridge University Press. embryonic egg cylinder. J. Embryol. exp. Morph. 75, GARDNER, R. L. & ROSSANT, J. (1979). Investigation of 189-204. the fate of 4-5 day post-coitum mouse inner cell mass BEDDINGTON, R. (19836). The origin of the foetal tissues cells by blastocyst injection. J. Embryol. exp. Morph. during gastrulation in the rodent. In Development in 52, 141-152. Mammals, vol. 5 (ed. M. H. Johnson), pp. 1-32. GROBSTEIN, C. (1952). Intraocular growth and Amsterdam: Elsevier. differentiation of clusters of mouse embryonic shields BEDDINGTON, R. (1986). Analysis of tissue fate and cultured with and without primitive endoderm and in prospective potency in the egg cylinder. In the presence of possible inductors. J. exp. Zool. 119, Experimental Approaches to Mammalian Embryonic 355-380. Development (ed. J. Rossant & R. A. Pedersen), HEASMAN, J., WYLIE, C. C, HAUSEN, P. & SMITH, J. C. pp. 121-147. New York: Cambridge University Press. (1984). Fates and states of determination of single BELLAIRS, R. (1982). Gastrulation processes in the chick vegetal pole blastomeres of A", laevis. Cell 37, 185-194. embryo. In Cell Behaviour (ed. R. Bellairs, A. Curtis & HIROSE, G. & JACOBSON, M. (1979). Clonal organization G. Dunn), pp. 395-427. Cambridge: Cambridge of the central nervous system of the frog. I. Clones University Press. stemming from individual blastomeres of the 16-cell BELLAIRS, R. (1986). The primitive streak. Anat. Embryol. and earlier stages. Devi Biol. 71, 191-202. 174, 1-14. HOGAN, B. L. M. & TILLY, R. (1981). Cell interactions COPP, A. J., ROBERTS, H. M. & POLANI, P. E. (1986). Chimaerism of primordial germ cells in the early and endoderm differentiation in cultured mouse postimplantation mouse embryo following embryos. J. Embryol. exp. Morph. 62, 379-394. microsurgical grafting of posterior primitive streak cells JACOBSON, M. & HIROSE, G. (1978). Origin of the retina in vitro. J. Embryol. exp. Morph. 95, 95-115. from both sides of the embryonic brain: A contribution CRUZ, Y. P. & PEDERSEN, R. A. (1985). Cell fate in the to the problem of crossing at the optic chiasma. Science ' polar trophectoderm of mouse blastocysts as studied by VSl, 637-639. microinjection of cell lineage tracers. Devi Biol. 112, JOLLY, J. & FERESTER-TADIE, M. (1936). Recherches sur 73-83. l'oeuf du rat et de la souris. Archs Anat. Microsc. 32, DANIEL, J. C, JR & OLSON, J. D. (1966). Cell movement, 323-390. proliferation and death in the formation of the KIMMEL, C. B. & LAW, R. D. (1985). Cell lineage of embryonic axis of the rabbit. Anat. Rec. 156, 123-127. zebrafish blastomeres. III. Clonal analyses of the DIWAN, S. B. & STEVENS, L. C. (1976). Development of blastula and gastrula stages. Devi Biol. 108, 94-101. teratomas from the ectoderm of mouse egg cylinders. J. KIMMEL, C. B. & WARGA, R. M. (1986). Tissue-specific natn. Cancer Inst. 57, 937-942. cell lineages originate in the gastrula of the zebra fish. ENDERS, A. C, GIVEN, R. L. & SCHLAFKE, S. (1978). Science 231, 365-368. Differentiation and migration of endoderm in the rat KOMINAMI, T. (1983). Establishment of embryonic axes in and mouse at implantation. Anat. Rec. 190, 65-78. larvae of the starfish, Asterina pectinifera. J. Embryol. FRANKE, W. W., GRUND, C, JACKSON, B. W. & exp. Morph. 75, 87-100. ILLMANSEE, K. (1983). Formation of cytoskeletal LAWSON, K. A., MENESES, J. J. & PEDERSEN, R. A. elements during mouse embryogenesis. IV. (1986). Cell fate and cell lineage in the endoderm of Ultrastructure of primary mesenchymal cells and their the presomite mouse embryo, studied with an cell-cell interactions. Differentiation 25, 121-141. intracellular tracer. Devi Biol. 115, 325-339. FRANKE, W. W., GRUND, C, KUHN, C, JACKSON. B. W. LEVAK-SVAJGER, B. & SVAJGER, A. (1971). Differentiation & ILLMANSEE, K. (1982). Formation of cytoskeletal of endodermal tissues in homografts of primitive Cell fate in mouse endoderm 651
ectoderm from two-layered rat embryonic shields. histological sections of M. miisculus
gastrulation and early organogenesis. Development 99, VAKAET, L. (1970). Cinephotomicrographic investigations .109-126. of gastrulation in the chick blastoderm. Archs Biol. THEILER, K. (1972). The House Mouse. Development and Liege Si, 387-426. Normal Stages from Fertilization to 4 weeks of Age. VAKAET, L. (1984). Early development of birds. In Berlin: Springer-Verlag. Chimaeras in Developmental Biology (ed. N. LeDouann VERMEIJ-KEERS, C. & POELMANN, R. (1980). The neural & A. McLaren), pp. 71-88. New York: Academic crest: a study on cell degeneration and the Press. improbability of cell migration in mouse embryos. WEISBLAT, D. A., SAWYER, R. T. & STENT, G. S. (1978). Neth. J. Zool. 30, 74-81. Cell lineage analysis by intracellular injection of a VAKAET, L. (1962). Some new data concerning the tracer enzyme. Science 202, 1295-1298. formation of the definitive endoblast in the chick embryo. J. Embryol. exp. Morph. 10, 38-57. {Accepted 16 July 1987)