The pattern of cell death in fushi tarazu, a segmentation gene of Drosophila

The ‘pair-rule’ class of segmentation genes was discovered by Nüsslein-Volhard & Wieschaus (1980); larvae die before hatching and are diminutive with parts of alternate metameres missing. In the case of fushi tarazu (ftz, Wakimoto & Kaufman, 1981), the deleted pieces of pattern approximate to the even-numbered parasegments (Martinez-Arias & Lawrence, 1985), as they extend from near one Keilins organ to the next (Nüsslein-Volhard et al. 1982; Struhl, 1984, 1985; Ingham et al. 1985). It has been suggested that pair-rule deletions correspond with the zones where the gene is expressed in the wild type (Hafen et al. 1984) and, indeed, in the case of ftz, the anterior borders of the stripes of cells expressing the gene at blastoderm (Carroll & Scott, 1985) do demarcate the parasegments exactly (Lawrence et al. 1987; Carroll et al. 1988).

What happens in the mutant embryos to those cells that would normally express ftz⁺?There is local cell death in embryos mutant for some segmentation genes (Ingham et al. 1985; Jäckle et al. 1985; Martinez-Arias, 1985; White & Lehmann, 1986), so one possibility is that all these cells die and leave no descendants. Here we map cell death in the epidermis with respect to the parasegment boundaries in ftz⁻ embryos. We find that cell death begins around completion of germ band extension, and is found inside and outside the territory equivalent to the even-numbered parasegments. However, cells continue to divide in areas where other cells are dying and it is possible that some of these cells survive and contribute to the cuticle pattern of the mutant larvae.

We used Ki ftz^w20/TM1 red (see Lindsley & Grell, 1968) to provide ftz embryos as ftz^w20 is a null allele (Weiner et al. 1984) with a strong phenotype (Wakimoto et al. 1984). This stock also carried a hybrid gene in which βgalactosidase is expressed under the 5’ control of even-skipped (eve-βgal) (Lawrence et al. 1987). We also used a ftz-β galactosidase (ftz-βgal) hybrid gene kindly provided by Hiromi & Gehring (see Hiromi et al. 1985).

The embryos were fixed and devitellinized according to Mitchison & Sedat (1983), treated with a monoclonal antibody against βgalactosidase (kindly provided by Chris Doe), and stained with horseradish peroxidase (Macdonald & Struhl, 1986). After antibody staining, the embryos were washed in distilled water for 15 min, and refixed for 20min in a solution containing 95% ethanol, 2 ml; glacial acetic acid, 1 ml; formalin 0·2 ml; and distilled water 2·8 ml. After several washes in 70 % ethanol the embryos were stained in fuchsin following Zalokar & Erk (1977).

The eggs were fixed for 7 min according to the phase partition method of Zalokar & Erk (1977). The vitelline membranes were removed by hand, and the embryos fixed again in a solution containing 2% glutaraldehyde, 2% paraformaldehyde in 0·1 M-cacodylate buffer, pH7·0. The specimens were embedded in Araldite and semithin (1–2 μm) and ultrathin sections cut and studied.

Embryos were staged according to the descriptions of Campos-Ortega & Hartenstein, 1985 and Wieschaus et al. 1986.

Embryos were mounted in glass tubes (Zackson, 1982), or between coverslips, so that they could be rotated under observation. Dying cells were mapped only when the cell was positively seen to be within the epidermis; this was achieved by rotating the embryo until that particular cell could be seen in clear optical section (Fig. 4). Care was needed to distinguish between a dying cell and one in late prophase or early telophase: dying cells were more intensely chromatic, chromosomes could not be seen by focussing up and down, and usually there was a characteristic unevenness of staining. This unevenness of staining probably derives from the sphere-within-a-sphere appearance visible in the electron micrographs (Fig. 1). The position of the cell in the anteroposterior axis was then estimated by eye with respect to the sharp lines of βgalactosidase staining and marked on a standard diagram. Later, this diagram was divided up into 10 equal zones and the number of marks added up in each zone to give the histogram in Fig. 5. More sophisticated methods to estimate position could have been used, but as there are only about 12 cells in the anteroposterior axis of a parasegment and these are unevenly shaped and packed, we believe such methods would give a false impression of accuracy.

The position of the border between the two presumptive parasegments was confirmed by the use of embryos carrying a ftz - β gal hybrid gene, which in ftz⁻embryos gave useful expression of βgal only in the mesoderm (Hiromi & Gehring, 1987). The anterior border of /3gal staining in the mesoderm is sharp and lined up with the epidermis halfway between the surviving parasegmental grooves. As we know that the mesoderm invaginates in register and does not slip relative to the epidermis (Martinez-Arias & Lawrence, 1985), this mesodermal boundary is a valid landmark for the ectodermal cells which, in the wild type, would constitute the anterior margin of the even-numbered parasegments.

Our initial task was to determine whether the intensely stained droplets we could see in the epidermis of ftz⁻ embryos were indeed dying cells. There are many chromatic droplets (Wigglesworth, 1942) in the ftz mutant embryos, and most of these are loose in the body cavity where they form masses in each segment (cf. Martinez-Arias, 1985, for fused;Ingham et al. 1985, for hairy;Perrimon & Mahowald, 1987, for 1(1) dishevelled). Electron microscopy confirmed that they are dying cells and showed that some become engulfed in large macrophage-like cells (Figs 1,2). In the epidermis, a smaller number of similar droplets could be seen and study of sections in both the light and electron microscopes confirmed that they are dying cells (Figs 1, 2). It would appear that they are expelled from the epidermis, and become clumped underneath or, sometimes, outside the embryo between the epidermis and the vitelline membrane. Comparison of the sections with the whole mounts confirmed that the chromatic droplets that stain intensely with fuchsin and the dying cells seen in the electron microscope are one and the same.

In stage-10 wild-type embryos, dying cells are very rare indeed. Using the criteria described in the methods, only one was seen within the epidermis of seven embryos (45 parasegments were screened). In these wild-type embryos, there are a few dead cells below the epidermis but we do not know their origin.

In mutant embryos, the dying cells in the epidermis can be first seen at the end of germ-band extension at about stage 9, and by stage 11 the subepidermal clumps are visible (Fig. 2). The location of these clumps may not depend on where, precisely, the cells die, instead the dead cells may be forced into the spaces midway between the parasegmental grooves, which are particularly deep in ftz mutant embryos (Fig. 3). In order to find out exactly which cells are dying, we map only the dead cells which are located within the epidermis – we presume these cells have died in situ and will shortly be thrown out. We have mapped cell death only early because that will be most informative about the effects of ftz; later on the pattern of cell death could be influenced by more indirect consequences of the mutant.

The embryos used also carried a hybrid gene in which βgalactosidase (βgal) is synthesized in the same pattern as the stripes of even-skipped (eve). The anterior boundary of these stripes demarcates exactly the anterior boundaries of the odd-numbered parasegments (Lawrence et al. 1987) and we were therefore able to map the dying cells with respect to these borders (Figs 3, 4). In ftz mutant embryos at the extended germ band stage, alternate grooves are missing (Wakimoto et al. 1984) so between the borders defined by βgal staining there are cells which would, in the wild type, make two parasegments (Martinez-Arias & Lawrence, 1985). Those cells that would normally make the even-numbered parasegments constitute the posterior halves of these double parasegmental units and it was here that most of the dying cells were found. Fig. 5 shows a peak of cell death near the middle of the double parasegment. Cell death is rare in the anterior halves of the presumptive odd-numbered parasegments, but affects all other regions. In stage-10 embryos about five dying cells were found in the epidermis per double parasegment unit (Fig. 5) – there being approximately 1200 epidermal cells in each such unit at that stage. As the time taken for a dying cell to be expelled from the epidermis is unknown, we cannot estimate the overall proportion of cells that die.

As development proceeds, the number of chromatic droplets within the epidermis appears to increase somewhat and by stage 11 large numbers of dying cells begin to accumulate below the epidermis (Figs 1, 2) – some of these may originate from the neuroectoderm and mesoderm. Cell death continues until much later (stage 14 and beyond), it does not occur in blocks of epidermal cells, being sparse and with no groups of more than a few dying cells. Finally the embryos become shrunken and deformed (Fig. 3).

We have mapped cell death in embryos lacking the fushi-tarazu (ftz) gene. These embryos form larvae in which zones approximating to the even-numbered parasegments are deleted from the cuticle pattern (Nüsslein-Volhard et al. 1982) and we find that the dying cells are distributed amongst all the cells that would, in the wild type, have formed the even-numbered parasegments as well as affecting part of the regions that form the odd-numbered parasegments.

The type of cell death that occurs, when studied in both the light and electron microscopes, has been called apoptosis. This type of cell death is thought to be caused by regulation, and differs from necrosis which results from direct insult to cells (Wyllie et al. 1980).

The first dying cells were detected immediately after germ band elongation (at stage 9) at a time when the ftz protein is rapidly declining in the wild type (Carroll & Scott, 1985). It might be that the cells that die are the descendents of those that would have expressed ftz⁺ in the wild type (Hafen et al. 1984), but we do not think this is a significant correlation. While the stripes of ftz -expressing cells are at one time as broad as the interstripes, this happens only transiently at midblastoderm and from then on the stripes narrow as cells at the posterior edge cease expressing the gene (Carroll & Scott, 1985; Lawrence et al. 1987). In contrast, cell death is scattered all over the regions corresponding to the even-numbered parasegments and also occurs in the presumptive odd-numbered parasegments and therefore affects cells, which in the wild-type blastoderm, only transiently or never express ftz⁺. Further, the deletion frame of ftz mutants is not precisely coincident with parasegments and takes out a variable region anterior to the Keilin’s organ (Ingham et al. 1985; Struhl, 1985). Also, cells do not die in a block, but rather are interspersed amongst healthy, dividing cells. For these reasons we believe that the cells die, not because they cannot express the ftz gene, but because of some more global and indirect effect (cf. Howard & Ingham, 1986) – for example, the lack of parasegmental boundaries might flatten the segmental gradient locally and this could cause an increase in the proportion of dying cells in that region (Lawrence, 1987). If so, the results suggest that the areas immediately behind the anterior boundaries of the odd-numbered parasegments are somewhat protected in the mutant embryos, perhaps by the normal functioning of the evenskipped gene which is primarily responsible for those boundaries (Lawrence et al. 1987).

In conclusion, we imagine a lack of ftz product causes a specific defect in the supracellular pattern, such as a failure in the anterior borders of even-numbered parasegments. This could lead to abnormal landscapes of positional information (for review, see Lawrence, 1973). Although many cells that normally express ftz⁺ are eliminated, the remainder could generate descendents that contribute to the cuticle pattern of the mutant larva. Cells that never express ftz⁺ are also affected and some of these die. The ancestors of cells that make the surviving metameres in the ftz mutant embryos may therefore not be fixed and localized subsets of the blastoderm cells and this principle may also be true of other segmentation mutations.

We thank Ruth Lehmann for her perspicaceous criticism of the manuscript. Two anonymous reviewers were a credit to their calling.