Two new roles for the Drosophila AP patterning system in early morphogenesis

During embryonic development, spatial patterns commonly arise in a sequence of discrete steps. In Drosophila, the anterior-posterior (AP) patterning system establishes fine-scale pattern through a cascade of transcriptional regulation, beginning with maternally active coordinate genes, through gap genes and pair-rule genes, finally achieving cell by cell resolution at the level of the segment polarity genes (Pankratz and Jäckle, 1993). These changes in transcriptional activity occur concomitantly with a series of morphological transitions. The embryo begins its development as a syncytium of dividing nuclei (Foe et al., 1993). By the tenth nuclear division, all nuclei, with the exception of the yolk nuclei, have migrated to the periphery of the embryo. It is also during this tenth nuclear cycle that zygotic transcription begins. In cycle 14, the initial pattern of pair-rule gene activities is established and the embryo begins a process of cellularization. Each cortical nucleus is packaged in its own individual plasma membrane, thus forming a monolayered epithelium that sits upon a yolk sack (Schejter and Wieschaus, 1993a). Cellularization is the first morphological process reported to require a major input from the genome of the embryo. Injection of transcriptional inhibitors like α-amanitin blocks cellularization and causes a defect in the clearing of lipid droplets from the periphery of the embryo (Arking and Parente, 1980; Schubiger and Edgar, 1994; Merrill et al., 1988). During cellularization, the invaginating membrane front forms a widened canal held together apically by a basal adherens junction (Fullilove and Jacobson, 1971; Hunter and Wieschaus, 2000). The width of this ‘furrow canal’ increases as cellularization proceeds, eventually separating the blastoderm cells from the underlying yolk sack. The connections between the forming cells and the yolk sack [‘yolk stalks’ (Rickoll, 1976)] narrow as the furrow canal widens; a myosin driven contraction of the stalk may drive the widening of the canal. As the process of cellularization is finishing, pair-rule genes are expressed with single cell resolution and the first movements of gastrulation begin. In the cephalic furrow region, the first sign of gastrulation is a rapid cell shape change in which a row of cells, termed the initiator cells, shorten dramatically along their apical-basal axis and widen laterally. These movements begin while connections between the yolk and the cytoplasm of the cell still exist.

These seemingly discrete stages of development use common cytoskeletal components that must be adapted to serve new functions as the morphology of the embryo changes. For example, during the syncytial stages, nuclei are associated with broad caps of actin, separated by regions enriched in myosin, which are transformed into pseudocleavage furrows during mitosis (Foe et al., 2000). The function of these caps is unknown, but they may be required for the proper spacing of nuclei in the syncytium (Postner et al., 1992). After the embryo enters cycle 14, the actin/myosin cytoskeleton is reorganized to allow cellularization. Myosin and actin now localize to the furrow canal where they contribute to the constriction of the stalks and may provide the force that drives membrane into the interior. The changing pattern in the cytoskeleton means that properties imparted by previous morphological events may set up constraints for subsequent stages of development. We report that, indeed, regulation in the spacing of syncytial nuclei affects properties of the actin/myosin array that forms during cellularization, and that regulation of the diameter of the actin/myosin array during cellularization may, in turn, affect the ability of gastrulation movements to occur.

We present two new morphogenetic processes that generate AP asymmetries long before those observed at gastrulation. We find that the first AP asymmetries driven by the genome of the embryo are visible by cycle 11, indicating that the complex patterns of gene expression during syncytial stages influence morphogenetic movements of these stages. We also argue that regulation by bicoid of the diameters of actin caps during syncytial cycles generates an anterior domain of lower nuclear densities. Second, we find that a new domain of faster cellularization occurs at the same position of the embryo as where the CF will eventually form. We suggest that this domain provides a link between cellularization and gastrulation in the requirement for the early expression of paired (prd), which appears to affect both the contractility of the cellularization apparatus and the initiation of the cephalic furrow at gastrulation.

OregonR stocks were used as the wild-type strain in these experiments. The following mutant alleles were used in our analyses: bcd^E1, nos^L7, tsl¹⁴⁶, prd⁴, prd⁷, eve^1.27, eve^R13. BB5+18/FM7 stock was used to collect females carrying six copies of bcd (Driever and Nüsslein-Volhard, 1988). The fly strain carrying the histon2A-GFP fusion protein is described elsewhere (Clarkson and Saint, 1999).

Rabbit anti-Myosin (a generous gift from C. Field) was used to visualize yolk stalks, and mouse anti-neurotactin (BP106 from Developmental Studies Hybridoma Bank) was used as a marker of the basolateral surfaces of cells. Rabbit and guinea-pig anti-Eve (Rabbit anti-Eve originally generated by M. Frasch) as well as guinea pig anti-Prd were kindly given by J. Reinitz. Rabbit anti-Prd was a generous gift from C. Desplan. Oli-green (Molecular Probes) and Hoechst (Molecular Probes) were used to visualize nuclei. All the above reagents were used on heat fixed embryos following the protocol of Peiffer et al. (Peiffer et al., 1994). Secondary antibodies against rabbit and guinea pig IgG and conjugated to Alexa-488, Alexa-546, Alexa-568 and biotin were from Molecular Probes. Streptavidin-Cy5 was used to visualize biotinylated secondaries. Actin was visualized with Alexa-568-conjugated phalloidin (Molecular Probes) on embryos that were hand de-vitellinized after fixing for 45 minutes in heptane/18.5% formaldehyde-phosphate-buffered saline (PBS). Injected embryos were also hand de-vitellinized though the same process. Embryos were mounted in Aqua-Polymount (Polysciences) and examined on a Zeiss LSM 510 confocal microscope. Images were analyzed, processed and assembled with Adobe Photoshop 5.0.

To block RNA transcription by RNA Pol II, a 100 μg/ml solution of α-amanitin (Sigma) was injected into dechorionated (Clorox bleach) and partially dessicated embryos. In the absence of lipid droplet clearing, which requires zygotic transcription, a golden haze of uncleared lipid droplet sits in the periphery of the embryo. Efficacy of injection was judged by this halo formation during lipid clearing.

Embryos were dechorionated, fixed in formaldehyde (heptane/15% formaldehyde PBS), and de-vittelinized with methanol. They were then post-fixed in 4% osmium tetraoxide, dehydrated in an ethanol series and washed into a 100% tetramethylsilane (Ted Pella) solution that was allowed to sublimate away overnight. Embryos were mounted on aluminum tape on an aluminum stud and sputter coated with a gold-palladium mixture for 80 seconds. Images were taken on a Philips XL FEG-SEM scope.

Means and standard deviations were calculated using Microsoft Excel. Probabilities for the role of prd in the pre-CF domain were calculated using the rank-ordering system of a Mann-Whitney test for non-normal distributions. Probabilities for actin cap data were calculated using the Student’s t-test. Measurements of actin caps were made with help of NIH Image v1.62, and then data pasted into Microsoft Excel and analyzed.

To assay the degree of uniformity along the AP axis during cellularization, we visualized the furrow canals of wild-type embryos by staining for Myosin. We found that by mid-cellularization, three domains of differing diameters of yolk stalks could be observed. The first domain is centered around the anterior pole of the embryo, and possesses the largest diameters (arrow in Fig. 1B,C). The second domain, immediately posterior to the first anterior domain, has the smallest diameters (arrowhead in Fig. 1B,C). This domain is centered around the location of where the cephalic furrow will eventually form. The third domain lies posterior to this pre-CF domain and has furrow canals of intermediate diameter (Fig. 1B,C).

These three domains also differ in the depth of the cellularization front. The anterior domain, which has large yolk stalk diameters, is the shallowest part of the embryo in its depth of cellularization (Fig. 1G,H). By the end of cellularization, the most anterior part of this domain can possess a depth of cellularization (15-20 μm) half that of the rest of the embryo (∼35 μm on average). The region posterior to the pre-CF domain has an intermediate depth (Fig. 1D-H), whereas the pre-CF domain, which has the smallest yolk stalk diameters, has the greatest depth of cellularization (arrow in Fig. 1D-F). The easiest way to visualize the greater depth of the pre-CF domain is by using the optical sectioning of the confocal microscope. By taking thin, 1 μm sections, we see a difference of 2-3 μm between the base of the pre-CF domain (Fig. 1D) and the base of cells in the rest of the embryo (Fig. 1F). Thus, the regions of differing yolk stalk diameters correspond to the regions of varying depth during cellularization.

To test whether these cellularization phenomena are due simply to the position or geometry of the embryo (e.g. the curvature of the anterior part of the egg), or if these phenomena are being specified and positioned by the AP patterning system, we looked for the three domains in embryos that lacked all positional information along this axis. In embryos derived from females triply mutant for bicoid nanos torso-like, yolk stalk diameters are uniform, even though other aspects of cellularization occur normally (Fig. 2B,C). The shallow cellularization front of the anterior domain (Fig. 2F,G), as well as the greater depth of the pre-CF, is also lost (Fig. 2B,C). Thus, these phenomena require the activity of the AP patterning system for their morphogenesis.

These cellularization phenomena respond to specific levels of the bicoid gradient. Embryos that carry six copies of bicoid produce more Bicoid protein, shifting any given concentration of Bicoid to a position posterior to where it would be in a wild-type embryo. In embryos carrying six copies of bicoid, the pre-CF domain of small yolk stalk diameters and greater depth is shifted posteriorly, while the anterior domain of large yolk stalk diameters expands to cover close to half of the embryo (Fig. 2D,E). bicoid appears to be the important factor specifying these phenomena, as embryos derived from females mutant for bicoid lack the anterior and pre-CF domains (data not shown).

Since the pre-CF domain is a relatively narrow domain, it seemed unlikely that the broad Bicoid gradient would be directly specifying the pre-CF domain. Several pair-rule genes are expressed in approximately the right time and place to be mediating the formation of the pre-CF domain. While Even-skipped (Eve) is expressed in the posterior half of the pre-CF domain (data not shown), Prd expression is directly centered on the pre-CF domain (Fig. 3C-E). Prd expression also has an unusual feature. When Prd expression is first detected during cycle 13, it is observed in a single, gap gene-like domain (Fig. 3A,B). Thus, Prd is expressed before the pre-CF domain is formed, and in the right location in the embryo. Moreover, the early expression of Prd in a single stripe, rather than in the stereotypical seven-stripe pattern, suggests that it could specify the pre-CF domain independent of other factors.

We then analyzed various patterning mutants for a disruption of the pre-CF domain. prd homozygotes show a normal anterior domain of larger yolks (ad in Fig. 3G), but there is very little difference in the yolk stalk diameters between the pre-CF domain and the posterior domain (compare pcf and pd diameters in Fig. 3G). The relatively subtle appearance of the pre-CF domain in different genetic backgrounds required the scoring of embryos in a blind test. On a classification scale of 0-5, with 0 indicating a complete absence of the pre-CF domain, prd homozygotes scored a 1.2, while their non-homozygous siblings scored a 3.3. A Mann-Whitney ranked sum test indicates that these classifications are significant (P<<0.001, n=61). Representative examples of the pre-CF domain in prd/+ (Fig. 3F) and prd^– (Fig. 3G) are shown in Fig. 3. Thus, prd is already active in the regulation of morphogenesis during the process of cellularization.

The pre-CF domain is first visible when the cellularization front has reached about 25% of its depth. However, the larger yolk stalks of the anterior domain are visible throughout cellularization. We find that the density of nuclei in the anterior is reduced by about 30% from the nuclear density that is found in the rest of the embryo. This anterior domain of lower nuclear density is centered around the anterior pole and extends to approximately two to three nuclei in front of the first stripe of Eve, or ∼70% EL (see Fig. 5, cycle 14 embryo; Fig. 7A). Judging from the staining patterns of Eve and Prd, this is the same area of the embryo in which the anterior domain of large yolk stalk diameters meets the pre-CF domain. During cellularization the anterior nuclear domain is consistently observed in OreR embryos with an average decrease in nuclear density of 27.6% (n=42, s.d.=4.4%). In this sample, the values for individual embryos ranged from 34% to 20%. This domain of lower nuclear densities is maintained and stays constant throughout the process of cellularization (Fig. 4A,B), and can still be observed in gastrulating embryos (Fig. 5). This greater spacing of nuclei in the anterior would necessarily lead to a cellularization network with larger diameters. Additionally, by mid-cellularization, when a pre-CF domain is already visible, a slight clustering of nuclei in this region occurs (arrow in Fig. 4B).

As the anterior domain of lower nuclear densities is present at the start of cycle 14, we asked when this AP asymmetry arises. While the sphere of nuclei that migrate to the surface of the embryo during cycle 9, and the cortical nuclei of cycle 10 are uniform along the AP axis, the first sign of increased nuclear spacing in the anterior can be observed in cycle 11 embryos. By cycle 12, a pronounced AP asymmetry is observed, which is maintained through the subsequent mitosis to the start of cellularization (Fig. 5, fixed 6 × bcd embryo (for reasons stated in figure legend) on left, GFP-histone time-lapse series on right). The anterior domain of lower nuclear densities is also observable in living embryos. Nuclear counts on GFP-histone embryos demonstrates a similar ∼30% reduction in nuclear density in vivo (Fig. 5, movies from three different embryos showed 30%, 29% and 26% differences in density at cycle 14). These observations on living embryos also reveals that, except for when the nuclei are undergoing mitosis and the associated yolk contractions, the anterior nuclear domain is present throughout the rest of the cell cycle.

Before cycle 14, cortical nuclei are associated with large actin caps. In embryos without actin caps, the regular spacing of nuclei is disrupted (Postner et al., 1992). One possible mechanism for the formation of the anterior domain of lower nuclear densities would use the regulation of the size of these actin caps. To see if asymmetries in actin cap size occur along the AP axis, we stained wild-type embryos with fluorescently labeled phalloidin to visualize F-actin. Measurements showed that anterior actin caps were larger than caps in the rest of the embryo in cycle 11 and cycle 12, but not during cycle 10 (Fig. 6, Table 1). This difference, although small, is highly reproducible from embryo to embryo. In cycle 11, the values for six different embryos were 8%, 7%, 8%, 6%, 7% and 9%.

The exceptionally early appearance of the AP asymmetries in nuclear distributions led us to examine whether the anterior domain of lower nuclear densities is specified by the AP patterning system. In embryos derived from bicoid nanos torso-like mutant females, nuclear spacing is uniform along the AP axis (Fig. 7B). In addition, the anterior nuclear domain can be expanded posteriorly in embryos carrying six copies of bicoid (Fig. 7C). Additionally, embryos from bicoid females are uniform in their nuclear densities (data not shown), arguing that the anterior domain of lower nuclear densities is set up by the Bicoid gradient. Finally, the larger diameters of the actin caps in the anterior do not occur in embryos from bicoid nanos torso-like females (Fig. 6, Table 1).

As zygotic transcription only starts around cycle 10, the early appearance of the anterior domain by cycle 11 raised the question of whether zygotic transcription is required for its formation. It is possible that bicoid might be regulating nuclear spacing through a post-transcriptional mechanism (Chan and Struhl, 1997). To assay the effect of transcription on nuclear spacing, we injected embryos with α-amanitin to block RNA polymerase II. Consistent with earlier studies, such embryos develop to cycle 14 with normal gross morphology. However, when we examine these embryos with nuclear stains, they show a total absence of asymmetric nuclear distributions. Embryos injected with α-amanitin fail to form the anterior domain of lower nuclear densities (less than 5% difference between nuclear density from the anterior to the posterior, n=18, also Fig. 7E,F), but all control embryos injected with water continued to show an anterior domain (greater than 20% difference in nuclear density, n=15, also Fig. 7D). We conclude that bicoid regulates the zygotic transcription to bring about the formation of the anterior domain of lower nuclear densities.

Our previous work has identified two genes necessary for cephalic furrow formation. Embryos mutant for either eve or buttonhead (btd) lacked cephalic furrows (Fig. 8E,F). Further work indicated that both eve and btd have a defect in the earliest phase of CF formation, which is initiator cell formation. At no point during development are the stereotypical cell shape changes of shortening along the apical-basal axis and widening of the cell observed in these mutants. Mutations in prd also cause a disruption in CF formation. At the onset of gastrulation, the cephalic furrow does not form, nor do initiator cells undergo their characteristic cell shape change (Fig. 8H). Because there is an absolute correlation between initiator cell behavior and the middle nucleus of the first stripe of Eve in wild-type embryos (Vincent et al., 1997), the location where initiator cells should form in prd embryos can still be identified. In prd embryos these Eve-marked cells are indistinguishable from their neighbors at stages during gastrulation when the ventral furrow has formed and the CF would normally be visible in wild-type embryos (Fig. 8G,H). However, by mid-germband extension (GBE), prd embryos have formed a regular, CF-like invagination (Fig. 8D). We believe that this late fold arises through the same mechanisms that govern normal CF formation, but that these processes are delayed relative to the development of wild-type embryos (in the stage 7 prd embryo (Fig. 8B), there are no cell shape changes in the region where the CF would form, while the stage 6 wild-type embryo (Fig. 8A) has an obvious CF). Initiator cells form in wild-type embryos at stage 6, at the onset of gastrulation (arrows in Fig. 8G). Imaging of prd embryos reveals initiator cell activities beginning at stage 7 (arrows in Fig. 8J). At this stage, wild-type embryos have already begun to deform the yolk sack with a basal bulge of the epithelium (Fig. 8I). At the beginning of stage 8, or germband extension, wild-type embryos have a furrow that is many cells deep (Fig. 8K), while prd embryos have just begun to deform the yolk sack (Fig. 8L). It is only by mid-GBE that most, but not all, prd embryos have a regular CF stretching around the entire circumference of the embryo.

The abnormalities in CF formation observed at the beginning of gastrulation in prd embryos are superficially similar to those observed in embryos mutant for eve or btd. In the absence of the activity of eve, btd or prd, the cell shape changes that occur in the row of cells that initiate CF formation at the beginning of gastrulation do not occur. In contrast to eve or btd, however, we have identified early activities of prd during cellularization that may account for the later differences observed in CF formation. Consistent with this view our analysis indicates that prd embryos often recover and form a fairly regular CF by mid-germband extension, unlike the severe disruption of the CF in eve and btd embryos (Fig. 8) (Vincent et al., 1997). Sander et al. (Sander et al., 1980), who discovered the original prd allele, noted that ‘gastrulation seemed delayed and that the cephalic furrow either fails to appear or forms some time after the onset of germband extension’. Our work refines and extends their observations. While the start of gastrulation occurs normally in prd embryos, CF formation and initiator cell behavior is delayed. We propose that this delay is due not to a function of prd during gastrulation in the specification of initiator cells, as we have proposed for eve and btd function. These latter mutants completely block initiator cell and CF formation. Instead, we show that prd function is necessary for the formation of the pre-CF domain, and suggest that it is only through this disruption of the pre-CF domain that prd functions in CF formation. The recovery of the CF observed in prd embryos may be a reflection that cellularization throughout the embryo has finally reached the point that the pre-CF domain reaches at the very beginning of gastrulation, and so CF formation, although delayed, may be correctly initiated. The advanced rate of cellularization in the pre-CF domain may reflect a required premature closing of the base of the initiator cells so that cell volumes may be maintained during the severe cell shape changes of gastrulation, or that cytoskeletal components involved in cellularization must be freed for initiator cell movements. Because little deformation, or loss of volume, occurs in the presumptive initiator cells of early prd embryos (arrows in Fig. 8H), we favor the latter of these two possibilities. Thus, the subtle regulation of one stage of development can have profound effects upon a later, seemingly discrete, process of development.

The cellularization front that arises during early cycle 14 is rich in actin and myosin, and is thought to provide a contractile force that orients and drives the process of cellularization (Warn et al., 1980; Warn and Magrath, 1983). This network, as well as cellularization itself, proceeds through a two-phase process. The first phase is a basal movement of the cellularization front towards the interior of the embryo. During this phase, the furrow canals stay constant in their small size, and the actin/myosin array has a hexagonal shape. The second phase is a lateral movement of the cellularization front, which creates a pinching off at the bases of cells in the newly forming cell sheet (Schejter and Wieschaus, 1993b). We have found that the morphology of the cellularization front is not uniform along the AP axis. The pre-CF domain is distinguished from other regions of the embryo by its small yolk stalk diameters and greater depth of cellularization.

Our results suggest the following model for the generation of AP asymmetries during cellularization. In the pre-CF domain, the Bicoid gradient directs the correct localization of the early gap gene-like expression domain of prd, which, in turn, directs a greater local contraction of the cellularization network. When cellularization is in its first phase of inward directed movement, the greater contraction of the pre-CF domain leads to a greater advance inwards of the cellularization front, thus creating the greater depth of the pre-CF domain. Then, when cellularization shifts to the second phase of a lateral, pinching-off movement, the greater contractility of the pre-CF domain leads to a greater widening of the furrow canals, which creates the smaller yolk stalk diameters observed in the pre-CF domain. The creation of this domain of advanced cellularization may be necessary for the initiator cell shape change required for cephalic furrow formation.

In certain respects, the anterior domain of large yolk stalk diameters and shallow cellularization appears to be the opposite of the pre-CF domain, and thus might be produced by a downregulation of the same contractile mechanisms that we propose operate in the pre-CF domain. However, we have found that the formation of the larger yolk stalk diameters in the anterior domain clearly involves a different mechanism. The anterior cellularization domain is pre-figured by an anterior domain of lower nuclear densities, while the pre-CF domain initiates in a region where nuclear densities are uniform. The greater spacing of nuclei in the anterior necessarily causes the formation of actin/myosin arrays of greater diameter during cellularization. We favor a model for the formation of the anterior domain in which the bicoid gradient, by cycle 11, has regulated the transcription of a set of zygotic genes, which in turn regulate the size of the actin caps overlaying the nuclei. This regulation of the actin caps results in larger caps in the anterior that necessitates a greater spacing of nuclei. By cycle 14, when cellularization is initiated, the greater spacing of nuclei dictates the generation of a cellularization network in the anterior in which the hexagonal components are larger in diameter. The initiation of a large hexagons would thus lead to larger yolk stalks. It may be that the larger hexagons of the cellularization network contract less efficiently, thus generating the shallower depth of cellularization that is observed for the anterior domain.

The concept of a mid-blastula transition (MBT) has generally referred to a time when the genome of an embryo begins to exert an influence on development, presumably through zygotic transcription. The best-defined MBT is for Xenopus (Newport and Kirschner, 1982a; Newport and Kirschner, 1982b), where the MBT was characterized by a cessation of synchronous mitotic cycles, the start of zygotic transcription, and a change in the morphology of blastula cells (i.e., the acquisition of cell motility). Various aspects of this characterization have since been called into question. There is low level zygotic transcription before the MBT, and it appears as though cell motility may not be a function of zygotic transcription, but of slower mitosis (Kimelman et al., 1987). Although these discrepancies call into question the usefulness of the MBT as a concept (Yasuda and Schubiger, 1992), the idea of an MBT remains an attractive model for explaining the changing morphology of the Drosophila embryo.

In flies, the MBT has traditionally been discussed in terms of cycle 14 development. It is at this point that the fly embryo ceases its synchronous syncytial divisions and several morphogenetic processes require zygotic transcription for their genesis. At a superficial level, the first observable defects in embryos deficient for zygotic transcription was at cycle 14, when α-amanitin-injected embryos showed defects in the processes of lipid droplet clearing and cellularization. However, as in Xenopus, defining a precise MBT has some difficulties. Although cycle 14 marks a major increase in transcriptional efficiency (Zalokar, 1976), the start of zygotic transcription occurs in different nuclear cycles for different genes. In general, for the early acting genes involved in patterning, transcription begins around cycle 10, although some genes are transcribed as early as cycle 8 (Pritchard and Schubiger, 1996), and mutations in specific segmentation genes show subtle disruption as early as cycle 10 (Karr et al., 1985). What is striking in the results we report for Bicoid and Prd is the correspondence between spatial patterns of expression and regional alterations in morphology. Our results show that the genome of the embryo is active in the spatial regulation of morphogenesis at a much earlier time than previously described. The genome directs reproducible asymmetries in morphology by cycle 11. The generation of an anterior domain of lower nuclear densities argues against a single discrete MBT at cycle 14 (Yasuda and Schubiger, 1992). Our results further suggest a more refined view of the Drosophila MBT in which there is a gradual shifting in the guidance of development from maternal to zygotic gene products, one that stretches from as early as cycle 8 until the cessation of mitosis and formation of a cellularized embryo at cycle 14.

We would like to thank members of the Wieschaus and Schupbach labs for critical comments and discussions. This work was supported by the Howard Hughes Medical Institute and grant 5R01HD22780 from the National Institute of Health.