Drosophila embryonic pattern repair: how embryos respond to bicoid dosage alteration

Establishment and maintenance of proper proportions and pattern are of paramount importance in animal development. A great deal has been discovered about the molecular biology and genetics of pattern formation in developing organisms. Meanwhile, our understanding of pattern maintenance and repair is quite limited. Pattern repair is defined as the process of correcting a tissue that has been altered by either genetic or physical means to generate a normalized final pattern or structure. This process may be carried out by any combination of cell behaviors, such as cell division, fate alteration and cell death. Pattern repair, for the most part, has been studied by observing regeneration in surgically altered insect and amphibian limbs and insect imaginal discs (Bryant et al., 1981; French et al., 1976). Irradiation with X-rays, ultraviolet light and ligation have been used to examine pattern repair in developing insect embryos (Bownes, 1975; King and Bryant, 1982; Schubiger, 1976). In the sea urchin, embryonic regulation has been extensively studied by cell and tissue recombination experiments (reviewed in Davidson, 1989; Ettensohn et al., 1996). These studies demonstrated that these organisms have an impressive capacity for pattern repair, but shed little light on its molecular/genetic basis. In an effort to gain a molecular/genetic understanding of pattern repair, one needs an efficient and reproducible means to mispattern the organism as an alternative to surgically induced mispatterning.

Nüsslein-Volhard and co-workers recognized that embryos laid by females with varying doses of the anterior morphogen, bicoid (bcd), showed early mispatterning of anterior morphological markers. Females possessing one to six doses of bcd lay eggs that show a number of early developmental alterations; the cephalic furrow (CF), a transient cleft that demarcates anterior midgut and anterior head structures from the rest of the head, thorax and abdomen, is shifted anteriorly in reduced bcd embryos or posteriorly in increased bcd embryos (Frohnhöffer and Nüsslein-Volhard, 1986; Berleth et al., 1988; Driever and Nüsslein-Volhard, 1988). Other groups showed that the spacing and domains of expression of a variety of gap, segmentation and segment polarity genes is compressed in embryos laid by females with extra copies of bcd (Cohen and Jürgens, 1990; Eldon and Pirrotta, 1991; Struhl et al., 1989; Kraut and Levine, 1991). Mitotic domains, stereotypic clusters of cells that divide synchronously, of embryos derived from females with differing doses of bcd are expanded or compressed resulting in altered numbers of cells (Foe and Odell, 1989). Surprisingly, these embryos developed into normal, healthy adults (Frohnhöffer and Nüsslein-Volhard, 1986; Berleth et al., 1988). These observations indicate that the Drosophila embryo is capable of pattern repair and that embryogenesis is plastic. In order to gain a better understanding of these repair mechanisms, we have undertaken an analysis of how bcd-induced pattern defects are repaired as well as an in depth analysis of what defects may be occurring.

For brevity, we will refer to embryos laid by females with n doses of bcd as nbcd embryos. This nomenclature refers to the maternal genotype not the zygotic genotype of the animal.

To date, researchers of bcd-induced pattern changes have examined embryos from the time of egg deposition to gastrulation, as well as cuticle pattern at the end of embryogenesis (Busturia and Lawrence, 1994). Little is known about the events between gastrulation and cuticle secretion – the time when patterns are established and probably when pattern defects are repaired. Here we attempt to fill this gap.

Oregon-R was used as a wild-type stock. 1bcd embryos were collected from st ri bcd^E1roe p^p/TM3 mothers. Females carrying extra copies of the bcd gene were mostly derived from BB9+16/CyO (kindly provided by G. Struhl; Struhl, 1989). The BB9+16 chromosome is a second chromosome derivative that carries two P elements which contain bcd genomic DNA inserts. A heterozygous balanced stock was used for the collection of 4bcd embryos. Homozygous BB9+16 females were used to generate 6bcd embryos. BB5+8/FM7 (stock T301 from the Tübingen Stock collection), which carries two P[bcd] inserts on the X chromosome, was used as an alternate source of extra copy bcd embryos.

Dechorionated, precellularized embryos were lined up on a no.1 coverslip and covered with halocarbon oil. Transmitted light images of these embryos were taken with an inverted microscope (Olympus) fitted with a scientific-grade, cooled CCD camera (Photometrics) with a 20× objective lens every 30 minutes for 48 hours or until they hatched. In order to record the development of up to 20 embryos per session, the microscope was equipped with a programmable stage (Melles Griot) that repeatedly cycled between embryos. The electronic images were stored and processed on a UNIX workstation (Silicon Graphics) using Deltavision (Applied Precision) and in-house written software.

Dechorionated, precellularized embryos were positioned on a coverslip dorsal side down. Transmitted light images of embryos within 10 minutes of CF formation were taken with either the afore-mentioned inverted microscope or a confocal microscope (MRC600, BioRad) using a 20× objective lens focused midway through the embryo. The position of the CF, expressed as percent egg length from the posterior, was calculated by determining the intersection of two lines connecting the anterior and posterior tips and the CF indentations.

Overnight collections of embryos were dechorionated and allowed to develop for 22-24 hours to ensure the completion of cuticle secretion. Because of their slower development, 6bcd embryos were allowed to develop for 36-42 hours before their cuticles were isolated. All unhatched embryos and larvae were subjected to vigorous shaking in 1:1 heptane:methanol to remove their vitelline membranes. Devitellinized embryos were cleared according to the protocol of Lamka et al. (1992). The cuticles of at least 200 animals of each embryo-type were analyzed using phase-contrast or dark-field microscopy.

Salivary glands were dissected from third instar larvae according to Rykowski et al. (1991). Fluorescence images of DAPI-stained salivary gland lobes using a 20× objective lens were recorded with a CCD camera.

To monitor cell divisions at the embryo surface in vivo, embryos were injected with either a solution of 0.25 mg/ml calcein or 1.5 mM RGPEG (Minden, 1996) into the intervitelline space according to the protocol reported by Minden et al. (1989). To visualize cell death, a solution of 0.25 mg/ml acridine orange in PBS was injected into the cytoplasm of syncytial embryos. Time-lapse recordings of laterally oriented embryos were initiated either at gastrulation for mitotic domain analysis or after the embryos developed to the germ-band- retraction stage, the time when cell death first is detectable. Each recording was composed of stacks of 6-8 optical sections, spaced 5 μm apart, taken at 1-5 minute intervals over a period of 8-12 hours. In order to analyze the cell death patterns over time in three dimensions, each image stack was projected into a single plane and the two- dimensional projections were viewed as a time-lapse series of images representing a 30-40 μm volume of the embryo.

The number of dying cells was determined by counting acridine- orange-positive nuclei from projected image stacks taken 50 minutes apart to ensure that dying cells were not counted twice. Macrophages that had engulfed dying cells appeared as large cells containing multiple fluorescent bodies were not counted. Anterior cell death counting was confined to the region anterior to the CF in stage 11-12 embryos and, in post-stage 12 embryos, to the region anterior to the yolk including the gnathal segments. Epidermal cell death counting was restricted to the germband and excluded signal from the yolk and amnioserosa.

Embryos were fixed and processed as described in Bomze and López (1994) using the following mouse monoclonal antibody dilutions; anti-INV undiluted to visualize ENGRAILED and INVECTED expression (Steve DiNardo), anti-AC 1:3 dilution (Sean Carroll) and anti-ELAV 1:1 dilution (Kalpana White). These antibodies were visualized with peroxidase-conjugated goat anti-mouse antibody (Vector Labs). A rat monoclonal anti-HB antibody (Paul McDonald) was used at a 1:500 dilution and detected with a goat anti-rat secondary antibody (Vector Labs). A rabbit polyclonal anti-EVE (Manfred Frasch) was used at a 1:4000 dilution and detected with a goat anti- rabbit secondary.

To determine the brain volume, fixed embryos of each class were incubated for 10 minutes in propidium iodide (20 μg/ml) and washed 3× 5 minutes in PBS. The fixation protocol used for antibody staining gave a propidium-iodide-staining pattern that highlighted a ring of the cytoplasm and weakly stained the nucleus. Image stacks of propidium-iodide-stained embryos were collected using a 60× objective lens on a confocal microscope (MRC 600, BioRad). Serial sections 1 μm apart were obtained from the embryo surface through the supraoesophageal ganglia for one side of the embryo. The areas of the brain in each section was determined and summed to obtain a volume. The cell density in random sections from each class of embryos was determined by counting the number of cells in an area of 750 μm² of brain.

A cDNA clone of rpr was isolated by RT PCR of third instar larval RNA with the following primers: CGCGAATTCATTGACTTT- TTGTTTGAC and CCGGAATTCTGAATAAGAGAGACACC (White et al., 1994).

The rpr and S59 cDNAs (a gift of Manfred Frasch; Dohrmann et al., 1990) were gel isolated and labeled with digoxigenin-dUTP (Boehringer-Mannheim) by extension of random hexamers with DNA polymerase I Klenow fragment. Embryos were processed as whole mounts as described (Tautz and Pfeifle, 1989). Complexed probe and RNA was detected using an alkaline-phosphatase-conjugated antibody against the digoxigenin dUTP (Boehringer-Mannheim).

Three general classes of phenotype were defined to indicate the severity of defects caused by altered bcd dosage: (1) wild-type, where there were no obvious changes seen at the stated level of resolution of the experiment; (2) mild, where only a partial loss or fusion of cells was seen in a few segments and (3) severe, where more than three segments were malformed. The severe class often included embryos with defects throughout the embryo.

Embryos endowed with one to four doses of bcd mRNA survived to adulthood with a high frequency (80%, see Table 1). In contrast, only 50% of the embryos generated by 6bcd females hatched into larvae and 60% of the surviving larvae eclosed as adults – giving an overall survival rate of 30%. Time-lapse microscopy of 6bcd embryos showed that 53% failed to hatch (n=57). The CF of the 6bcd embryos was displaced posteriorly compared to that of the 4bcd embryos and the clypeolabrum was significantly larger. After stage 12, clusters of free floating cells were often seen in the intervitelline space around the clypeolabrum in 6bcd embryos, but were rarely seen in wild-type embryos (Fig. 1A). These clusters of cells are believed to be dead cells and macrophages.

The embryos that failed to hatch were divided into five phenotypes (Fig. 1B-E). Class 1 embryos, 37% of unhatched embryos, first appeared abnormal during head involution, stage 14. The clypeolabrum started to retract but the dorsal ridge was not able to envelop the tip of the enlarged clypeolabrum to complete the head involution process. This often created a large mass anterior to the dorsal ridge constriction. In those embryos that managed to reach stage 17, the two mouth hooks were separated from one another by the excess head tissue pushing between them (compare the class 3 embryo in Fig. 1C to class 1 6bcd embryo in Fig. 1B). Most embryos, however, did not reach stage 17 due to secondary abnormalities that affected posterior development. Class 2 embryos, 27% of unhatched embryos, had yolk leakage outside the gut into the head cavity (Fig. 1D) or intervitelline space. The yolk breaches mostly occurred at, or after, the completion of germband retraction, stage 12–13, and were occasionally seen as late as gut constriction in stage 15. The cause of the ‘yolk displacement’ phenotype could possibly be compression of the abdominal fate map leading to failure of dorsal closure and/or gut development. This phenotype was observed in many class 1 (head involution defective) embryos after the failure of clypeolabrum retraction. Except for the site where the yolk leakage occurred, the development of the rest of the body appeared to proceed on schedule but with obvious abnormalities. Class 3, 17% of unhatched embryos, developed normally up to stage 17 (Fig. 1C). Class 4 and 5 each occupied 10% of the unhatched embryos. The development of class 4 embryos ceased before completing germband retraction in stage 12. These embryos developed a large void in the anterior intervitelline space that contained a massive amount of dead cells (Fig. 1E). Class 5 embryos showed unsynchronized cellularization (data not shown).

In order to determine the correlation between viability and CF position, the CF position was determined for live 6bcd embryos, which were then allowed to develop for a total of 48 hours and then scored for hatching. The average CF position for 6bcd embryos was 53.9±1.5% egg length (EL) relative to the posterior tip (Table 1). Embryos that formed the CF more posterior to the mean position died at almost twice the frequency (18/29) of the population of embryos that formed the CF anterior to the mean position (11/33). The mean CF position of the embryos that failed to hatch was 53.4% EL, compared to 54.1% EL for the survivors. These results demonstrate that embryonic lethality correlates to the extent of mispatterning, the more posteriorly shifted the CF, the lower the probability of survival.

To eliminate the possiblility that the 6bcd defects were the result of genetic background effects resulting from homozygous BB9+16 mothers, we assayed the viability and CF position of 6bcd embryos laid by BB5+8;BB9+16 transheterozygous females mated to wild-type males. The CF position of the transheterozygous 6bcd embryos was 55.5% EL. This was intermediate between 59.0% EL for BB9+16/CyO 4bcd embryos and 53.9% EL for BB9+16/BB9+16 6bcd embryos. The difference in CF position in 6bcd embryos laid by the different females is most likely due to variation of bcd expression depending on the chromosomal location of the various P [bcd] elements. The hatching rate of transheterozygous 6bcd embryos was 81%, compared to 50% for BB9+16/BB9+16 6bcd embryos and 90% for BB9+16/CyO 4bcd embryos. The mean CF position of the transheterozygous 6bcd embryos that failed to hatch was 53.1% EL as compared to 55.8 % EL for the survivors. This is very similar to the mean CF position of the failed BB9+16 homozygous embryos. The majority of unhatched transheterozygous 6bcd embryos displayed class 1 head involution defects. These data support the observation that the more posteriorly shifted the CF, the lower the embryonic survival. In addition, the two sources of 6bcd embryos produced similar phenotypes of embryonic lethality indicating that the observed phenotypes were dependent on bcd dosage and not on the genetic background of the stocks.

Wild-type and 1bcd animals developed normal cuticles, while 15% of the 4bcd animals had one missing or a pair of fused denticle belts (Fig. 2A). The most severely affected were the 6bcd embryos and larvae. 30% of 6bcd embryos and larvae had normal cuticular structures (Fig. 2B), while the rest were abnormal. Denticle abnormalities ranged from embryos with mild defects that had missing or fused abdominal denticle belts (25%, Fig. 2C), to severe defects that either failed to develop any visible cuticle segmentation (24%) or had grossly malformed denticles affecting more than two segments (21%, Fig. 2D). The denticle fusions and deletions generally occurred between the third and fourth abdominal segments for 4bcd and 6bcd embryos and larvae. Analysis of third instar 6bcd larvae showed that animals with fused or missing denticles had reduced viability relative to normal animals, as the proportion of defective animals decreased with each molt (data not shown).

There appeared to be a dramatic increase in mouth part defects between 4bcd embryos and 6bcd embryos. The mouth parts of 55% 6bcd animals formed aberrantly. A minority (13%) of the abnormalities had well-formed mouth parts that were positioned improperly (Figs 1B, 2F). The rest had abnormal or missing mouth structures (Fig. 2G). Some embryos completely lacked mouth structures leaving a large hole in the head cuticle, which is consistent with the head involution defects seen in the time-lapse recordings.

In an effort to determine if altered bcd doses perturbed internal- anterior structures, we measured brain volume and cell density and cells per salivary gland (Table 1). In all cases, the size of the brain was constant at approximately 40×10³ μm³. The number of brain cells per unit area was also independent of bcd dosage; 35.5 cells/750 μm² area. This indicated that either the size of the brain anlage was not affected by bcd-induced mispatterning or the altered brain was compensated for during subsequent brain development.

The number of cells per salivary gland was determined for third instar larvae since the cell number is fixed at the end of embryogenesis and the glands are much easier to dissect at this stage (Table 1). 2bcd embryos had an average of 141 cells per salivary gland lobe. There was less than one standard deviation increase in cell number for 4bcd animals, while 4bcd and 6bcd counts were nearly identical. Larvae from 1bcd mothers had salivary glands that were significantly reduced in cell number from wild type. These data indicate that there is an upper limit to the salivary gland cell number and that expanded regions can be regulated toward normal cell counts for those embryos that survive to third instar larvae. Conversely, compressed salivary gland anlagen do not appear to be regulated and cannot increase cell number to match wild type.

To determine changes in the mitotic patterns of nbcd embryos, fluorescent dye was injected into the intervitelline space of embryos during cellularization. The fluorescent dye occupies the spaces in between cells showing a bas relief of the embryo surface (Warn and McGrath, 1986; Kam et al., 1991). When a cell at the surface of the embryo enters mitosis, it rises out of the surface of the epithelium and can be seen as a round void surrounded by a fluorescent ring. The overall shape and order of appearance of the mitotic domains was unchanged in the 6bcd embryos that eventually hatched and those of classes 1-3 (Fig. 3). The size of the anterior mitotic domains were significantly enlarged, but there were no signs of mitotic silencing in the anterior or extra mitoses in the posterior of these embryos (data not shown).

The observation of the intervitelline cell clusters and the normalization of salivary gland size led to the suspicion that cell death played a role in repairing expanded regions of the fate map. Two approaches were taken to investigate changes in cell death patterns in nbcd embryos. First, time-lapse recordings of embryos injected with the vital dye acridine orange (AO) were used to determine the dynamics of cell death. Second, cells fated to be eliminated by apoptosis were visualized by in situ hybridization of reaper (rpr) mRNA (White et al., 1994). Three-dimensional, time-lapse recordings were made of 1bcd, 2bcd, 4bcd and 6bcd embryos. The fluorescent signal from a dying cell persisted for 30-40 minutes. The engulfment of dead cells by macrophages confused the time-lapse interpretation because the AO signal from the ingested dead cells persisted presumably because the macrophages were continually engulfing dead cells. In order to assess the cell death patterns, careful analysis of the location and time of a cell’s demise had to be discerned from a background of fast moving, highly fluorescent macrophages. Fig. 4 shows images taken from a time-lapse recording of AO injected embryos. At least four time-lapse recordings were made of 1bcd, 2bcd and 6bcd embryos. The 6bcd embryos used for this analysis did not exhibit either head involution or yolk displacement defects and all hatched. The average number of dying cells seen at 50 minute intervals was determined for the head and abdomen of nbcd embryos (Fig. 4D). This ensured that all non-marcophage signal originated from newly dying cells that were not counted in the previous time point. 6bcd embryos laid by BB9+16 or BB5+8 homozygous females exhibited the same changes in the cell death pattern. These results show a direct relationship between the amount of cell death and the expanded region of the embryo – there is an increase in cell death in the expanded head domains of 6bcd embryos and expanded trunks of 1bcd embryos. Conversely, compressed regions of the embryo appeared to have less apoptosis.

The time-lapse recordings of 6bcd embryos showed that the embryonic brain (which is visible due to the background fluorescence) first appeared enlarged relative to wild-type embryos but, over time, one could see macrophages hovering over the surface of the brain as it gradually decreased to normal size. There was also increased cell death in the clypeolabrum, which is believed to be the source of the perivitelline cell clusters. The cell clusters are a mixture of dead cells and macrophages as evidenced by their AO staining and, in some cases, motile behavior.

To corroborate the in vivo AO observations, embryos with different bcd doses were fixed and stained for rpr expression. 6bcd embryos at late stage 11 had clear increases in rpr expression in the expanded head domains and a corresponding decrease in rpr expression in their abdominal domains (Fig. 5). Similarly, in 1bcd embryos rpr expression was increased in the abdomen and decreased in the head. These results indicate that many of the excess cells in expanded regions are destined to be eliminated by cell death.

In order to determine if fate map compression led to segmental defects in all tissues, a number of molecular probes were used to assess the formation of epidermal, mesodermal, central nervous system (CNS), and peripheral nervous system (PNS) cell types.

ENGRAILED (EN) is expressed both in the early epidermis and CNS (DiNardo et al., 1985). Anti-EN antibody staining revealed that approximately 10% of 4bcd embryos between stages 12 and 15 had defects in the epidermal segments, but none in the CNS (Fig. 6). Defects included fusions or deletions of EN stripes. The remaining embryos appeared normal in both the CNS and epidermal stripes.

6bcd embryos between stages 12 and 15 were divided into three classes. 38% of the embryos had severe defects in EN pattern where the CNS clusters and epidermal stripes were either fused, missing or displaced. 30% of the embryos had defects only in the epidermis without a corresponding defect in the neural clusters of the ventral nerve cord. 32% of the embryos appeared normal in both the CNS and epidermal stripes (Fig. 6 and Table 2). There were no instances of embryos with defects in the CNS and not in the epidermis.

Embryos were probed with anti-ELAV antibody, which recognizes postmitotic neurons (Robinow and White, 1991). Approximately 94% of 4bcd and 1bcd embryos had wild-type numbers of peripheral neurons arranged in a normal pattern. No fusions were seen between clusters of peripheral neurons nor were there deletions of neuronal clusters. Approximately 50% of the embryos laid by 6bcd mothers had defects in their peripheral neurons, which could be seen as early as stage 12 when the antibody first recognizes some neurons. In many embryos, the neurons were dramatically displaced (Fig. 6). The remaining 50% of the 6bcd embryos appeared normal. The number of neurons in each segment was equivalent to wild-type and there were no fusions between adjacent clusters. Normal embryos were found at all stages of development. Thus, as seen with EN expression, the PNS and CNS are less affected by increased bcd dosage than the epidermis.

The CNS could prevent or repair fate map compression defects by two general schemes; either ensure the proper formation of neuroblasts during neuroectoderm fate selection or compensate for the initial reduction in neuroblast formation by allowing the neuroblasts to produce more progeny neurons. To discriminate between these two alternatives, embryos from each genotype were stained with an antibody against ACHAETE (AC), which transiently labels all S1 CNS neuroblasts in stage 8-11 embryos (Skeath and Carroll, 1992). Embryos were also stained with a HB antibody which transiently labels all neuroblasts (Tautz et al., 1987). Despite the obvious anterior expansion and posterior compression, the relative numbers of AC-staining S1 neuro- blasts appeared to be equivalent in 76% of 6bcd when compared to 2bcd embryos (Table 2). The remaining 6bcd embryos exhibited defects where several segments had missing or disorganized neuroblasts. Examination of CNS neuroblasts stained with HB gave similar results. HB-stained embryos were examined for peripheral neuroblast defects in stage 11-13 embryos. In spite of the size variation in the gnathal buds of nbcd embryos, approximately 4-6 neuroblasts were labeled in the gnathal segments in embryos from each class. The number of peripheral abdominal neuroblasts were equivalent in 1bcd, 2bcd, 4bcd and approximately 50% of 6bcd embryos at stage 11. The remaining 50% of the 6bcd embryos were split between those with mild and severe defects. Those with severe defects were missing numerous abdominal neuroblasts and often had obvious head expansion defects (Table 2). These results indicate that the number of cells allocated to become neuroblasts are equivalent among each of the 1, 2 and 4bcd genotypes and 50% of the 6bcd embryos. These 6bcd embryos with normal neuro- blast formation probably represent a majority of those that continue on to hatch.

Somatic mesodermal precursors during stages 12-15 were probed for S59 mRNA expression (Dohrmann et al., 1990). More than 80% of 4bcd were completely normal with respect to cluster size and position within each hemisegment. Increasing the bcd dosage to six reduced the fraction of embryos with a normal S59 pattern to 44%. The remainder were split between mild defects (25%) with one or two abnormal segments, and severe defects (31%), which often affected throughout the embryo (Fig. 7, Table 2).

The dorsal vessel, which gives rise to the larval heart, is derived from dorsally migrating mesodermal cells. A subset of these cells express eve (Frasch et al., 1987). Stage 15 embryos derived from nbcd females were immunostained for EVE. 1bcd embryos showed wild-type numbers of EVE-expressing dorsal mesoderm cells. 31% of 6bcd embryos had normally patterned dorsal mesoderm cells. 44% of the 6bcd embryos had mild defects where only a few cells were deleted or displaced. The remaining 25% had severe defects often missing the anterior eve-expressing cells and a severe compression of posterior cells (Fig. 7; Table 2). As seen with the nervous system markers, the mesoderm was less perturbed by increased bcd dosage than the epidermis. This result also indicates that the eve-expressing mesoderm cells are more affected than the somatic muscles.

Two major phenotypes were observed in the non-hatching 6bcd embryos: failure of head involution and a yolk leakage defect. The head involution failure appeared to be due to the excess of anterior tissue that could not be properly internalized. The excess of anterior tissue demonstrates an upper threshold for the number of cells that can be eliminated by programmed cell death in expanded regions. The yolk displacement phenotype may be the result of either posterior compression or anterior expansion. In some yolk-displaced embryos, the yolk leaked directly into the intervitelline space – indicating a breach in the epidermis, while in others the displaced yolk was contained within the epidermal sheath – indicating failure of gut formation. Some embryos displayed both head involution and yolk displacement defects. Our data indicate that the 6bcd- induced lethality is directly related to the extent of mispatterning. The absence of these phenotypes in 1bcd or 4bcd embryos demonstrates that this level of mispatterning did not exceed the limits of pattern repair for either compressed or expanded regions of the fate map.

There is a graded increase of abdominal denticle fusions and deletions with increasing doses of maternal bcd. These defects mostly affected segments A3 and A4. A possible explanation for the localized defects is that these embryos have a mostly normal posterior domain, while the anterior domain is expanded. The anterior expansion is graded, where the extent of expansion is greatest at the anterior tip and decreases toward the middle of the embryo (Foe and Odell, 1989). This causes the compression to be most severe near the middle of the embryo, where segments A3 and A4 arise. Nearly one-third of all 6bcd embryos have fused or deleted denticle belts. The defects are not limited to gross denticle mispatterning. Nearly 80% of the 6bcd embryos that hatch as first instar larvae have mild denticle defects that include denticle hair deletions and polarity errors (data not shown). The cuticle and mitotic domain results, and those of Bustaria and Lawrence (1994), show that the epidermis cannot increase cell number in response to a mispatterning event that reduces the cell number. The observation that cell number cannot be increased in the epidermis can be extended to include the salivary gland in 1bcd animals. The embryonic brain, however, does not show a reduction in size or cell density in 1bcd embryos. This indicates that different tissues may have different minimum cell number thresholds.

The abdomen of 6bcd embryos is compressed. We see that the majority of these embryos develop abnormally. When and where do these defects occur? 90% of 6bcd embryos develop to gastrulation where one observes the normal number of eve and en stripes, albeit the stripes are closer together. Postgastrulation mitoses appear to be normal as revealed by inter- vitelline injection of fluorescent dyes and anti-tubulin staining (Foe and Odell, 1989). Proneural patterning at stages 8 and 9 was 75% normal. By stages 11-13, the number of peripheral neuronal defects increased to 50%. The CNS at stage 12-15 was somewhat less affected than PNS at 62% normal. At the same stage, only 32% of 6bcd embryos had normal epidermal patterns. There were also reductions in cell numbers in the mesoderm at stage 15 where more dorsal mesoderm structures were slightly more affected. These data are in agreement with those of Maggert et al. (1995) who have shown that reducing the width of the mesodermal stripe leads to the loss of visceral mesoderm and cardiac precursors.

These results show that compression defects are not corrected and that reducing the number of cells below a minimum threshold of cells fated to particular structure leads to structural defects. The susceptibility to compression defects can be explained by the order of cell type specification. The mesoderm is the first group of cells to segregate from the epithelial monolayer. The number of mesodermal cells per segment is controlled by the width of the stripe of invaginating mesoderm and segmental width. Next, the CNS neuroblasts delaminate; followed by the PNS neuroblasts, or mother cells. Each of these segregation events reduce the number of potential epidermal cells. Thus, if the initial number of cells per segment is limited by fate map compression, it seems reasonable that the epidermis would be most affected, followed by PNS, and then CNS.

6bcd embryos that had head involution or yolk displacement defects generally had abdominal defects affecting all three germ layers in three or more segments. We believe these massive defects indicate a system-wide failure where too many segments were compressed below a certain threshold and cells that normally do not signal each other begin to interact, causing a breakdown in the overall signaling network. The severity of defects seen in 6bcd embryos probably depends on the combined effects of embryo length, cell density and the absolute amount of BCD protein. These factors result in the observed distribution of phenotypes. A precise correlation of these factors with developmental fate of 6bcd embryos has yet to be determined.

Cell death is a likely mechanism for repair of expanded domains. 6bcd embryos displayed increased levels of cell death in the expanded presumptive head region, while 1bcd embryos had increased cell death in the abdominal segments. Wild-type embryos require cell death for normal development. A deletion mutant that eliminates reaper, dramatically reduces the amount of cell death throughout the embryo and shows defects in head involution and over population of the CNS and PNS (White et al., 1994). The process of head involution appears to be dependent on cell death as evidenced by the mutation head involution defective, which also shows reduced levels of apoptosis (Grether et al., 1995). Indeed, the expanded head regions of 4bcd and 6bcd embryos show increased reaper expression. Therefore excess cells in the expanded head can be eliminated and proper head involution can take place. However, there is a threshold in the number of cells that can be eliminated. Almost half of all 6bcd embryos had some abnormality in the head cuticular structures indicating defects in head involution. Transmitted light movies showed that one third of 6bcd embryos recorded died with the inability to involute head tissue or because the excess tissue prevented proper gut formation. These embryos initiate retraction of the clypeolabrum, but fail to complete head involution. In both situations, embryos were unable to remove sufficient amounts of tissue by apoptosis. If the fate map expansion is too great, errors in morphogenesis will occur. We have shown that both the salivary gland and the brain cell number is normalized in 4bcd and 6bcd embryos as well as the epidermis in 1bcd embryos. This analysis shows that apoptosis plays a major role in tailoring tissue to its wild-type cell number.

Time-lapse recordings of cell death in wild-type embryos revealed that most embryonic tissues display some level of apoptosis during development. This, we believe, indicates that most embryonic tissues are initially established with an excess of cells. Once the correct arrangement of cells is established, the excess cells die by apoptosis. Varying the maternal bcd dosage leads to two types of mispatterning: fate map expansion and compression. Repair of expanded regions occurs by apoptosis of the excess cells. The additional cell deaths occur on the same schedule as seen in wild-type embryos. There is an upper limit to the extent of fate map expansion that an embryo can tolerate, which was seen as either tumorous structures or epithelial breaches, which were caused by too many cells dying in a concentrated area. There are numerous examples in mammalian development of cell death eliminating excess cells after a structure is formed. The best example is in neuronal development where a large number of neurons send axons to form a particular contact; once one of the axons makes contact, the unsuccessful neurons die by apoptosis (Raff et al., 1993; Truman et al., 1992). In Drosophila, the final step in eye development is the removal of excess cells after ommatidia formation. This cell death is required in order to generate the near crystalline array of ommatidia (Cagan and Ready, 1989; Wolff and Ready, 1991).

Compressed regions of the embryo showed a reduction in cell death. The compressed abdominal epidermis of 6bcd embryos showed less apoptosis than wild-type embryos indicating that the cell number had dropped below the lower limit of cells required to properly form an epidermal segment. Similarly, the salivary gland anlage in 1bcd embryos may have fallen below the threshold and formed a reduced-sized salivary gland. The developing brains of 1bcd embryos display a reduced, but still substantial, level of cell death indicating that the embryonic brain is supplied with a greater relative excess of cells for brain development than the epidermis and salivary glands. Under- standing the mechanisms that control embryonic cell death will provide information about embryonic pattern repair.

The first two authors contributed equally to this work and should be considered as first authors. We would like to thank Chuck Ettensohn and Javier López for critical reading of this manuscript. We would also like to thank the members of the Minden and Lopez labs for helpful discussions. J. M. would like to thank Bruce Alberts in whose laboratory this work was initiated and Christiane Nüsslein-Volhard for suggesting the project. T. P. was supported in part by an NSF graduate training grant to the Science Technology Center for Light Microscope Imaging and Biotechnology, BIR 9256343. J. M. is a Lucille P. Markey Scholar, and this work was supported in part by grants from the Lucille P. Markey Charitable Trust and NIH grant HD31642.