Teleost epiboly: a reassessment of deep cell movement in the germ ring

In 1936, Pasteels suggested that it was impossible to describe a ‘gastrula’ in terms of its structure alone, and that conceptually gastrulation should be viewed as a dynamic series of cell and tissue rearrangements (Pasteels, 1936b). Although there is agreement that the end result of these morphogenetic processes is to elaborate and to juxtapose the primary germ layers of the embryo, comparative studies have led to controversy as to the unity of mechanisms deployed throughout the major vertebrate phyla (reviewed by Ballard, 1981). One such question surrounds the origin of the hypoblast in the teleost embryo.

The teleost blastodisc consists of three distinct cell layers surmounting a large fluid-filled yolk mass. Adjacent and overlying the yolk itself is the yolk syncytial layer [YSL], a multinucleate syncytium arising by collapse of marginal blastomeres into the yolk cell [YC] (Kimmel & Law, 1985). Covering the blastodisc, and firmly attached to the outer margin of the YSL by a complex of tight and close junctions, is a thin squamous epithelium - the enveloping layer [EVL]. This layer maintains a very high electrical resistance and provides an effective barrier to movement of even small ions (Betchaku & Trinkaus, 1978; Keller & Trinkaus, 1987). Between these two layers are the deep cell blastomeres [DCs] - a population of large, rounded and mitotically active cells. The embryo and yolk sac are entirely derived from this layer and thus a knowledge of the way in which these cells move during epiboly is central to a mechanistic understanding of teleost gastrulation.

In a number of teleosts, the blastodisc maintains excellent optical transparency throughout epiboly, permitting direct observation of DC behaviour using differential interference contrast (DIC) microscopy. This technique has been used extensively to characterize the morphology, movement and contact behaviour of individual DCs (Lesseps et al. 1979; Trinkaus & Erickson, 1981; Van Haarlem, 1979). However, direct observation of the coordinated movement of the DC population, as a coherent tissue, has been neglected. To date, this problem has only been approached using particle and dye displacement experiments (Pasteels, 1936a; Ballard, 1966a,b, 1973a).

However, these ‘fate-mapping’ techniques provide a static view of what is in essence a dynamic process and do not convey information about the actual mechanism(s) of translocation. Indeed, it is precisely this approach that has led to confusion. Pasteels (1936b) concluded, as a result of his own dye marking experiments on a number of vertebrate species, including Salmo, and interpretation of earlier studies by Vogt (1929) using amphibia, that all vertebrates gastrulate using essentially similar movements of epiboly, invagination and convergence. More recently, Ballard (1981) has criticized these early studies for not considering the ability of dyes to diffuse passively from their site of application and has categorically stated that no cell internalization occurs at the edge of the teleost blastodisc and that there is no involution of the DC layer (Ballard, 1981). Indeed, the current appreciation of teleost gastrulation rests largely on this interpretation and suggests that there is no global rearrangement of DCs to establish the hypoblast which typifies gastrulation in virtually all higher vertebrates (Trinkaus, 1984).

Our strategy has been to observe movement of the DC layer within the teleost germ ring directly, using the precise optical sectioning abilities of Nomarski DIC microscopy. Using the nuclei of these large cells as an intrinsic marker, we have observed rapid and extensive involution of DCs within the germ ring during the early stages of epiboly from the most superficial levels (i.e. adjacent to the EVL) to deep regions within the blastodisc.

Fertilized eggs of the Rosy Barb, Barbus conchonius, were collected from natural matings and incubated in singlestrength Steinberg’s medium (pH 7·4) at 25°C. Late cleavage stages were dechorionated using 0·25 % Pronase (ex. Streptomyces griseus) [Sigma Chemicals] for 3 min at 25°C and then immersed in three changes of fresh Steinberg’s medium. Subsequently, early epiboly stages were placed in 0·5 % low gelling temperature agarose (< 30°C) [Sigma Chemicals] in Steinberg’s medium that had been previously cooled to 28°C. Embryos were immediately removed in a drop of this agarose solution to a glass coverslip contained in a culture chamber and oriented using a hair loop, such that the germ ring could be viewed ‘en face’ or in profile (N.B. when viewed using Nomarski DIC microscopy the animal-vegetal axis was perpendicular to the objective lens) (see Fig. 1). After manipulation, the agarose drop was allowed to cool at 25°C until gelling was complete (approximately 10 min) and the chamber flooded with Steinberg’s medium. Finally, the chamber was sealed with a further coverglass, inverted and placed on the stage of an Olympus photomicroscope (BHS/PM-10AD) maintained at 25°C and viewed using Nomarski DIC microscopy and a ×20 planapo objective. Still pictures were taken on Ilford Pan F 35 mm negative film using an Olympus camera (C-35AD-2) and time-lapse sequences recorded at 15 s intervals on Ilford Pan F 16 mm negative film using a Bolex cine camera (H16J) connected to an Olympus intervalometer (PM 1VM). Single-frame analysis of cine sequences was achieved using a Steenbeck editing projector (ST1600) [W. Steenbeck, Hamburg, W. Germany] at an effective enlargement of ×1700.

Sectioned material was fixed in Bouin’s fixative for 12 h, transferred to 70% ethanol for 24 h, dehydrated in a graded series of ethanol dilutions, cleared in toluene and infiltrated with paraffin wax at 56°C in a vacuum embedding oven. Sections were cut at 4μm and stained with haematoxylin and eosin.

Figs 2–5 illustrate a series of Nomarski DIC micrographs taken en face (i.e. at an axis perpendicular to the embryo’s surface) between 4 and 8 h of development postfertilization. At approximately 4 h of development and just prior to formation of the germ ring, the YSL extends, vegetally, considerably in advance of the mass of DC blastomeres (up to 20 μm). Tension striae, presumably generated by contractile forces within the YSL, are visible within the YC and extend along lines of longitude at regular intervals to surround the entire yolk mass (see arrows in Fig. 2). Numerous lipid droplets can also be seen collecting on the undersurface of the blastoderm at the yolk/YSL interface. Fig. 3 illustrates an optical section beneath the EVL at the vegetal margin of the germ ring, taken during mid epiboly (approximately 5 h). This equatorial axis clearly demonstrates the multinucleate nature of the internal YSL. Above this layer, the large, rounded DCs remain closely associated with few intercellular spaces and can be identified by their prominent nuclei. As in the previous figure, the yolk has a pseudocellular appearance created by the packaging of lipid droplets within it. At 8h, epiboly is virtually complete and DCs within equatorial regions disassociate as extracellular space increases. Fig. 4 shows a group of such cells adjacent to the EVL. At this stage, net movement of these DCs is latitudinal and convergent (i.e. towards the future axis of bilateral symmetry). These cells can be seen extending typical bleb-like processes (see arrows). Net movement of the upper strata of DCs within the germ ring continues vegetally until epiboly is completed.

The 16mm time-lapse cine sequence shown in Fig. 5 demonstrates the vegetal progression of the germ ring/YSL interface during an 80 min period. Frame A was taken at the commencement of epiboly and shows the YSL/YC interface. Frames B–I illustrate a later sequence from the same embryo; the plane of focus is just behind this interface and beneath the EVL, such that the most vegetal DCs within the germ ring (see area between curved arrows) have always been maintained in sharp relief. The advance of the germ ring margin is rapid (240 μm h^-1) and is accompanied by the continuous disappearance of DCs at the vegetal margin. Close examination reveals that these peripheral cells sink out of view and become replaced by proximal neighbours. Six embryos were examined en face, using an analysing projector, which permitted tracking of individual DCs. The 7·5 min sequence illustrated in Fig. 6 demonstrates that at least half the most vegetal DCs within the optical section were lost from view, to be replaced at the periphery by proximal neighbours. Further analysis of all six embryos revealed that this was a consistent feature, suggesting a phase of DC internalization at the margin of the germ ring during early to mid epiboly.

It was possible to quantify these observations by monitoring all DCs within an optical section approximately one cell deep, during a 30min period in two early - mid epiboly stage embryos. The site for analysis was en face, chosen at random during the embedding procedure, and the plane of focus was similar to the optical section described for Fig. 5. The resulting field of view extended latitudinally for 500μm and up to 150 μm proximal to the germ ring margin. Of a total of 75 cells tracked in both embryos, 54 disappeared at the germ ring/YSL interface and the remaining 21 cells [all located proximally at time 0], although not internalized, were displaced to the vegetal margin of the germ ring. All cell tracks were highly directional [longitudinal and parallel] resulting in vegetal, meridional displacement of the entire DC population under observation. Furthermore, no DCs sank from view at any location other than at the edge of the germ ring.

In order to establish the exact nature of this movement, the entire dorsoventral thickness of the germ ring was viewed in profile (i.e. at a latitudinal axis passing directly through the germ ring itself) (see Fig. 1). Figs 7, 8 illustrate this aspect in histological section (vertical) at the commencement (Fig. 7) and mid (Fig. 8) stages of epiboly. Mitotic figures are always prevalent in such sections and can be found throughout the DC population. A fissure is detectable (see arrows in both figures) extending longitudinally from a central region within the germ ring and appears to subdivide this structure into an outer (i.e. towards the EVL) stratum of DCs, up to five cells across, and a less-closely associated inner layer, which is in direct contact with the internal-YSL. The interface is invariably characterized by cell flattening at areas of close apposition between DCs from inner and outer strata and correlates with the axis of shear between layers during epiboly, as observed in timelapse sequences (not illustrated).

As the ‘subgerminal’ cavity develops, the first cells to re-establish contact with the YSL appear at the margin of the germ ring. These cells are loosely associated and remain separated from overlying cells (see Fig. 8). Fig. 9 demonstrates this morphology on either side of a transversely bisected mid-epiboly-stage embryo. Indeed, serial sections of ten early to mid epiboly stages show that this event is qualitatively similar around the entire circumferential margin of the blastoderm during this period of development.

The mechanism by which this morphology is generated was revealed using time-lapse cinemicroscopy. The vegetal progression of the germ ring was filmed in profile using Nomarski DIC optics during early to mid epiboly. Ten embryos were filmed in total and the area of germ ring selected for observation was chosen randomly during agarose embedding. A period of DC internalization was recorded at the margin of the germ ring in all cases. Indeed, this process could be detected around an entire hemisphere if both en face and profile axes (on opposite sides of the blastoderm) were observed. Fig. 10 shows a typical sequence of 55 min duration at a stage approaching mid epiboly. A pair of DC nuclei have been traced from a position adjacent to the EVL (75 μm behind the EVL/YSL interface), around the margin of the germ ring to a more proximal position within the lower stratum of this structure and 50 μm from its vegetal margin.

Figs 11 and 12 have been made from 16 mm timelapse cine sequences by tracing the paths of nuclear displacement in two embryos using an analytical projector. In both diagrams, the tracks of six nuclei from different regions of the DC layer have been plotted. Each sequence lasts for approximately 30 min and is unambiguous in its depiction of involution within the DC population. Translocation of DCs from superficial locations towards the vegetal margin of the germ ring occurs at rates of up to 300 μm h^-1 but becomes reduced as they are displaced to deeper regions adjacent to the EVL/YSL interface. Subsequently, their movement is less directional and, although net translocation is usually longitudinal and away from the germ ring margin, considerable latitudinal displacement is also observed. This phenomenon is much less apparent in preinvoluting DCs, which are continually replacing those lost by involution at the margin of the germ ring. As a result of these movements, during early to mid epiboly, the DC population within the germ ring is constantly turning over, as the structure itself moves towards the vegetal pole. It should be noted that at no point was the EVL observed to invaginate.

This study does not attempt to define the locomotory mechanisms by which epibolic movement is accomplished and at this stage it is not possible to define contributions by active migration or passive displacements of DCs to the advancement of the germ ring. Indeed, this movement may take place entirely independently of DC involution since the rate of vegetal progression remains relatively constant throughout epiboly (250 μm h^-1), including later stages when involution is not detectable.

These results present the first direct evidence of DC involution during epiboly of the teleost blastodisc and contradict the current and widely held view, based on indirect evidence from vital dye and particle marking experiments, that no involution or global rearrangement of DCs occurs during gastrulation to establish the hypoblast (Ballard, 1966a,b, 1973a,b, 1981; Trinkaus, 1984).

Using the prominent cell nucleus as a marker, individual DCs have been followed during rotational displacement about an apparent axis of shear, located centrally within the germ ring. During the early stages of epiboly, this process is qualitatively similar at any location around the entire circumferential margin of the blastodisc. Involution is extensive, involving superficial DCs (adjacent to the EVL) as well as those located more centrally within the germ ring and is associated with continuous vegetal displacement of superficial DCs from points at least 200 μm from the edge of the blastodisc. From our data, it cannot be excluded that quantitative differences may occur, although the two sites examined quantitatively both illustrated at least 70% of superficial DCs were involved in involution. Furthermore, this figure may be artificially low, since those cells counted as not involuting were carried to the edge of the germ ring and may have been involuted after the period of observation. Postinvoluting DCs are found in association with the YSL or within deep regons of the DC layer and illustrate latitudinal and longitudinal displacement. In contrast, invagination of the EVL was never observed at any stage during epiboly.

The origin of the hypoblast in the teleost embryo has long been a subject of dispute. Goette (1873) first suggested that this tissue could arise by ‘wheeling movement’ of DCs at the margin of the blastodisc and Kowalevski (1886) accepted this idea on the basis of evidence accumulated from serial sectioning. More recently, Pasteels (1936a) and Devillers (1951) supported such a mechanism, but extended their hypothesis to include the EVL, since application of dyes or particulate markers to its surface frequently resulted in disappearance and penetration of label to deep regions within the blastodisc, which they interpreted as active internalization. However, this approach has been criticized by Ballard (1966a, 1981) who, rightly, suggested that these earlier studies did not pay enough attention to the accurate placement of particles and to the passive diffusion of vital dyes away from superficial regions, to stain adjacent underlying cells. In an attempt to resolve these problems, Ballard (1966a,b, 1973a, b) conducted an extensive series of controlled dye and particulate marking experiments using Salmo, Catostomus, Gobius and Perca. As a result of these studies, he suggested, first, that no involution occurs at the rim of the teleost blastodisc and, second, that the hypoblast arises by delamination and outward migration of centrally placed DCs, a possibility first put forward by Rieneck (1869). The results described in this brief résumé present a confusing picture. Apart from the inherent inability of the early marking techniques to represent accurately cell movements, the desire to draw phylogenetic comparisons has not adequately considered the highly specialized nature of the teleost blastodisc. In particular, the fate of the EVL is to differentiate to form a transient periderm, which makes no contribution to the embryo proper and is sloughed off during or shortly after hatching (Bouvet, 1976).

Therefore, this structure cannot be considered homologous with the involuting marginal zone of the amphibian gastrula - a structure with which it is often compared (see Ballard, 1981). Indeed, Keller & Trinkaus (1987) have provided cinematographical evidence that, although individual cells within the EVL are constantly rearranging themselves to accommodate circumferential distortions during epiboly over the yolk mass, the EVL remains firmly attached to the external margin of the YSL by a complex of close and tight junctions (Betchakau & Trinkaus, 1978) and this arrangement would not be expected to facilitate invagination at the blastodisc edge.

To address the second part of Ballard’s conclusions, although this study does not preclude contribution to the hypoblast by delamination of DCs from central regions within the blastodisc, in Barbus the first rudiments of the hypoblast are almost certainly elaborated from DC involution at the germ ring and not by delamination and outward migration from central regions within the blastodisc, as is thought to be the case in Salmo (Ballard, 1966b). Transverse, serial sections of early epiboly stages consistently reveal separation of the overlying cellular blastodisc from the YSL. During late cleavage, DCs maintain close contact with the YSL, but, once formed, this ‘subgerminal’ cavity persists until at least 60% epibolic overgrowth. Significantly, the first cells to invade this space are located precisely adjacent to the germ ring, and polar regions of the YSL are the last to regain contact with this layer. This phenomenon was documented by Pasteels (1936a) from examination of serially sectioned trout embryos and considered by him to substantiate involution at the germ ring. Such a static description does not, in itself, justify our conclusion but, coupled with the direct evidence for DC involution presented here, provides a compelling argument.

Within the germ ring, not only is the net displacement of pre- and postinvoluted DCs in opposing directions, relative to each other, but these two populations appear behaviourally distinct. The first DCs to be involuted are surrounded by considerable extracellular space, permitting latitudinal movement and spreading over the YSL. In contrast, overlying DCs are more rounded, remain contiguous with their neighbours and show little latitudinal displacement. As epiboly continues and more DCs become involuted a distinct bilayer is generated between pre- and postinvoluted tissue, with no obvious mixing between layers. Thus, the first appearance of a layered blastoderm occurs within the germ ring at this interface and is maintained and elongated during epiboly. The precise chemical and ultrastructural nature of this interface is at present under investigation. However, Boulekbache et al. (1984) have demonstrated the presence of a network of fibronectin fibrils underlying the subgerminal cavity and also within the germ ring in Salmo. Such an arrangement may suggest a morphogenetic role for this molecule during teleost gastrulation, particularly during elaboration of the primary tissue layers.

Our observations are not totally at variance with the marking experiments in Salmo, which illustrate vegetal displacement of dye and accumulation at the germ ring after implantation of a carrier centrally within the blastodisc, prior to formation of the subgerminal cavity (Ballard, 1966b). Such a result would be expected on the basis of the considerable vegetal flow of DCs during the early stages of epiboly. Ballard may have failed to detect DC involution in Salmo, either because his methods were inadequate or because the timing and duration of this event may differ in this teleost. However, we cannot agree with his interpretation that ‘the hypoblast constituent of the germ ring receives deep central cells that migrate outward’ and, to date, no direct evidence has been presented to support such a proposal. Although the precise extent of the DC contribution to the hypoblast by involution remains to be established, this report demonstrates that teleosts, in line with other vertebrate groups, do employ a phase of cell internalization during their early development.

The authors are grateful to G. H. R. Booms for careful preparation of the histological sections, to H. F. Wood for drawing the diagrams and to Dr P. Thorogood for critical reading of the manuscript. The study was partly supported by a research grant from the Agricultural University, Wageningen, and the Department of Biology, Southampton supported A.T.W.’s travel costs.