Abnormal chromatin (abc), a maternal-effect locus in Drosophila melanogaster

Early embryonic development in many metazoans requires rapid DNA synthesis and nuclear divisions and is accompanied by cytoskeletal changes. It therefore provides an avenue for examining such basic cell-biology processes. Analysis of mutants defective in some aspect of a process can illuminate steps involved in it. Isolation of many cell-division-cycle mutants in budding yeast (e.g. see Hartwell et al. 1974; Hartwell and Weinert, 1989) has enabled intensive work to be done on mitosis and the cell cycle: at least five parallel pathways of dependent interactions and over 50 genes appear to be involved. Less is known about the cell cycle in more complex eukaryotes. Molluscs, echinoderms, amphibians and arthropods have been subjects of cell-cycle studies, many of which focussed on embryos (for reviews, see McIntosh and Koonce, 1989; Murray and Kirschner, 1989). In Drosophila, the search throughout the genome for female-sterile (fs) mutations that halt development has been extensive (e.g. X chromosome: Gans et al. 1975; Komitopoulou et al. 1983; Mohler, 1977; Orr et al. 1989; chromosome 2: Schüpbach and Wieschaus, 1989; Szabad et al. 1989; chromosome 3: Erdelyi and Szabad, 1989). It has produced a rapidly growing understanding of genes and pathways involved in oogenesis and the storehouse of maternal molecules that direct such aspects of embryogenesis as cleavage, polarity and nuclear migration. Many of the female-sterile mutations, however, are known only as single alleles and have not yet been examined in detail. Here we report on three allelic, recessive, maternal-effect lethal mutations in Drosophila melanogaster that affect early nuclear behavior and halt development, usually before the cellular blastoderm stage.

In normal Drosophila development (see Campos-Ortega and Hartenstein, 1985, for general review), fertilization is followed by nine nuclear divisions in the center of the embryo at approximately 9-min intervals. During the last two divisions the nuclei start migrating to the surface, where the future germline nuclei are set aside in the posterior end and the rest make a single layer and divide four more times (Foe and Alberts, 1983; Zalokar and Erk, 1976). There is no further division in the next hour, which is when cell membranes form; by 3.5 h at 25°C the embryo has reached the cellular blastoderm stage and gastrulation begins.

A review by Glover (1989) includes more than 70 loci implicated by mutations to be involved in the cell cycle in Drosophila, but again most of these have only one allele. In a recent study Gatti and Baker (1989) examined a group of mutations that were lethal at the larval-pupal transition (late lethals), expecting some to be involved in mitosis because this is a period when imaginai cells are dividing to produce the adult insect: 30/59 proved to be cell-cycle mutants. Mutations in a few loci that affect nuclear division in the embryo have been studied in detail in the hope of understanding mitotic events in general as well as the extremely fast cleavages unique to the early embryo. These include polo (Sunkel and Glover, 1988), which has spindle abnormalities that are suspected to be caused by centrosome defects, and asp (abnormal spindle:Gonzalez et al. 1990; Ripoil et al. 1985), in which a structural element in the spindle is changed.

The abc locus is a previously unstudied gene on the second chromosome in region 51F-52A. Recent work on this unexplored area has revealed several maternal-effect genes (Underwood et al. 1990; Underwood and Lengyel, 1988). The embryos laid by females hemizygous for the abc mutations (ABC embryos) exhibit development that is slower and is arrested long before hatching, often between the third and sixth nuclear divisions. We have studied them with a variety of cytological tools: DNA stains to observe the nuclei; antibodies to actin and tubulin (cellular components of the cytoskeleton also involved in division) and to centrosomes (nucleation sites for mitotic spindles); neuroblast squashes for chromosome structure and mitotic data; and ultrastructural analysis of the chorion. Results from these studies lead us to suspect that the gene may be involved in DNA replication and/or nuclear division in the early embryo.

All visible mutants are as described by Lindsley and Grell (1968) or Lindsley and Zimm (1985,1987). The three alleles of abc (fsPL, fs27, fsA5) will be referred to by their allelic designations. Mutants fs27 and fsA5 were derived from a screen aimed at saturating the region uncovered by the deficiency Df(2R)XTE11 (51E;52A6-10), using γ-irradiation of cn bw sp (fs27) or cn bw (fsA5) males (Underwood et al. 1990). mat(2)synPL63 (abbreviated here as fsPL) was induced with EMS on a cn bw chromosome (Schüpbach and Wieschaus, 1989). The original fsPL chromosome also contained an allele of the female-sterile mutation fruh (T. Schüpbach, personal communication). We have eliminated this second mutation from the chromosome through recombination and find the ABC phenotype unaffected. All were maintained at 25°C by standard lab culture over CyO (=In(2LR)O,Cy dp^ivlpr cn²) or GlaCyO (=In(2LR)Gla, In(2LR)O, Cy dp^ivl) balancers, as was the deficiency, Df(2R)XTE11. “Wild types’ used for phenotypic comparisons were Oregon R and the parental chromosome, either cn bw or cn bw sp.

(1) Collections: virgin mutant/CyO females were crossed to XTE11/CyO males, mutant/XTE11 F₁ females were crossed to Oregon R males, and F₂ embryos were collected at 25 °C on molasses-agar plates plus yeast paste at timed intervals. All subsequent operations were at room temperature, ∼23°C. (2) Dechorionation: embryos were put into water, commercial bleach was added to a concentration of about 50 %, and after 2 min they were rinsed several times with water. (3) Vitelline membrane removal and fixation: the method of Mitchison and Sedat (1983) as modified by Karr and Alberts (1986) was scaled-down (1/10) for use in Eppendorf tubes, (a) For antibody studies, the embryos were pipetted into a tube containing 0.5 ml PEM (see Mitchison and Sedat, 1983) + 0.5 ml n-heptane + 10–100 μl 37% formaldehyde (with anti-tubulin, 1 pl 0.5 DIM taxol in dimethylsulfoxide was added before the formaldehyde, the tube shaken for <10 s, and formaldehyde added), shaken for 10 min, the aqueous portion was removed by Pasteur pipette, 0.75 ml methanol added and shaken for 10 min; two rinses followed, first in methanol, then in 100% ethanol, and the embryos were stored at 4 °C in 70% ethanol, (b) For staining with Hoechst 33258 alone, embryos were fixed in 0.9 ml n-heptane+90 μl Hoechst buffer B + 10 μl 37% formaldehyde (Mitchison and Sedat, 1983) and the same procedure was used.

The technique of Karr and Alberts (1986) was used except that some primary antibody incubations occurred overnight at 4 °C and staining with Hoechst was for 10 min. The primary antibodies were diluted 1/500, the secondaries 1/1000.

The anti-centrosome Rbl88 was a kind gift from D. Glover and W. Whitfield, as was taxol from M. Suffness (NCI). The primary monoclonal antibodies, anti-β-tubulin and anti-actin, were purchased from Amersham. Secondary antibodies goat anti-rabbit IgG and goat anti-mouse IgG conjugated with rhodamine came from Organon Teknika and goat anti-mouse IgG conjugated with fluorescein from US Biochemical Corp.

Embryos were viewed with epifluorescence using a Zeiss Axiophot photomicroscope. Film used was Kodak Tech Pan 2415 (ASA 100).

Preparations were by the methods of Matthews and Kaufman (1987), mounting in polyvinyl lactophenol (Bio/medical Specialties, Inc.).

Fixation for transmission electron microscopy of chorion fine structure was a modification (lengthening embedding times) of the Kalt and Tandler (1971) method. Embryos were embedded in Spurr’s (Electron Microscopy Sciences), sectioned on an LKB Microtome HI and viewed on a Zeiss EM10.

Third-instar larvae were obtained from cultures transferred daily and supplemented with liquid yeast. Squashes were as described by Karess and Glover (1989). Fields were defined by a 63 x Zeiss Plan-apochromat phase-contrast objective and 10 x eyepieces on a Zeiss Axiophot microscope.

We examined the nuclei, mitotic apparatus, elements of the cytoskeleton, and chorion of eggs and embryos produced by mutant females, as well as larval neuroblast chromosomes, to help define the role of this gene in early development. Complementation analysis indicated the three mutations are alleles (Underwood et al. 1990). They provide a range in severity of defects: fsA5 is the strongest (has the most severe effect) and fsPL has the weakest effect, permitting embryos to develop furthest.

Employing a chromosome with a deficiency in the area, Df(2R)XTEll, flies were made hemizygous for one of these mutations. Hemizygosity was used because both fs27 and fsA5 are homozygous semi-lethal: only two-thirds (fs27) or one-fifth (fsA5) of the expected number reach the adult stage; the rest die as pupae. This semi-lethality may be due to unrelated lethal(s) (outside the deficiency) induced on the chromosomes during mutagenesis (Underwood et al. 1990). Adult hemizygous females showed reduced viability: about 15% died within 4 days of eclosion, compared to no deaths in the hemizygous cn bw (sp) control. In addition the egg-laying rate was reduced compared to the control (hemizygous parental) rate: fsPL was 2/3 and fs27 and fsA5 1/10 the control rate. Embryos from homozygous fsPL mothers rarely progressed beyond 64 nuclei. Here we define FSPL, FS27 and FSA5 embryos as those produced by hemizygous females; none of these mutant embryos survived its first 24 h or hatched. The mutant phenotype was almost always evident in such embryos from the earliest nuclear divisions, regardless of their genotype, i.e. there was no paternal rescue to adulthood of these maternal-effect mutants.

All three alleles have in common: large nuclei, some of which stain intensely and are located along the anterior-posterior (A-P) axis; erratic distribution of nuclei; and mitotic asynchrony. In many embryos there are some bright and some faint nuclei of similar size.

Fig. 1 compares embryonic DNA stained with Hoechst 33258 during the first three nuclear divisions in controls and FSPL. Even at the 2-nuclei stage, the presence of bridges distinguishes FSPL (Fig. IB) from parental cn bw (Fig. 1A). Similarly, the mutant nuclei in the 4-nuclei stage (Fig. ID) may be joined by bridges, whereas the parental (Fig. 1C) nuclei are not. At the 8-nuclei stage the control (Fig. IE) nuclei are arranged in a sphere and mitotically synchronized (the fuzzier nuclei are in a different focal plane), while the mutant (Fig. IF) nuclei already show irregular distribution and mitotic asynchrony.

Many embryos from all three mutants show distribution of nuclei that differs from wild-type by the third cleavage. In wild-type the nuclei are arranged as if on the surface of a sphere that spreads into an ellipse with further cleavage. In the mutants these early nuclei are often not in a sphere and may not form an ellipse, but are rather aligned rowlike along the A-P axis (Fig. 2E). In wild-type the nuclei become smaller with successive divisions, so that by pole cell formation the nuclei are small and of uniform size within each embryo. The nuclei in the mutants frequently appear to remain large and there appear sometimes (on the basis of size and/or staining intensity) to be subgroups of nuclei (Fig. 2E); in addition irregularly shaped nuclei that resemble yolk nuclei may be present (Fig. 2C).

Yolk nuclei in wild-type embryos (Fig. 2B) become evident in the interior during migration of nuclei toward the embryo’s surface, starting with cleavage 7. After three mitoses (stages 8–10) the yolk nuclei stop dividing but continue replicating DNA and become polyploid, as seen by increased staining intensity (Foe and Alberts, 1983). They are initially spread throughout the embryo (with relatively few in the pole cell region) and appear gradually to withdraw into the central portion along the A-P axis with the remaining yolk. They maintain their integrity as individual nuclei, but sometimes clump together in small bunches.

ABC embryos (Fig. 2D), on the other hand, commonly have less uniform distribution of yolk nuclei, some of which may be near the pole cells (seen also in engrailed embryos (Karr et al. 1985)). Abnormal chromatin clumps that look like yolk nuclei may show up early (Fig. 2C), i.e. before the other nuclei have surfaced. There is a strong tendency to clump, so that an embryo may appear at low magnification to have a small number (e.g. 7) of huge, brightly staining yolk nuclei that are often strongly channelled along the A-P axis (Fig. 2D).

An opaque Hoechst-staining ring occurs in the posterior (Fig. 2F) of some embryos of all three mutants, least often in FSPL (5 %) and in about 20 % of the other two. The wildtype control shows such a ring in <1 % of its embryos. The ring is generally in the presumptive pole cell position.

In wild-type embryos, nuclei that reach the surface are uniformly distributed and mitotically synchronized prior to gastrulation (Foe and Alberts, 1983; Karr et al. 1985) (Fig. 2A,B). The phenotype in the small number of mutant ABC embryos that develop that far (Fig. 2C) is variable but characteristically includes abnormal chromatin clumps, unusual distribution of nuclei, and some chromatin extending between nuclei that are otherwise well-separated (i.e. nuclear bridges). Pole cells often form in those embryos with nuclei in the posterior part. In the 30 % of FSPL embryos where nuclei reach the surface (Fig. 2D), abnormal blastoderm forms in most (4/5), while the rest (1/5) have a normal surface pattern; about a third of those with surface nuclei attempt to gastrulate, and a fifth of those gastrulating survive long enough to make some aberrant cuticle (2 % of total embryos laid).

In embryos that reach blastoderm, ABC surface nuclei vary in location, extent and mitotic stage (Fig. 2D). Often there is a single row of nuclei in a different mitotic stage bordering regions with and without surface nuclei (Fig. 5B,D,F). Distribution of nuclei within one area may be normal, but in over half these embryos there are regions of different mitotic stages that are distinct from the ends-to-middle mitotic waves seen in wild type (parasynchrony: Zalokar and Erk, 1976); a similar phenotype has been described by Karr and his coworkers (1985) in engrailed embryos. In addition, the region containing surface nuclei may be compressed relative to the rest of the embryo, as if girdled.

To see if the early embryonic phenotype reflects weak expression of a general mitotic defect, we assessed the frequency of cell division and examined chromosomal structure during mitosis in neuroblasts (Gatti et al. 1974) from F₁ third-instar larvae. Homozygotes are cn bw and thus distinguishable by the color of their Malpighian tubules (white) from their heterozygous (yellow) sibs. For each genotype, mitotic figures in at least 10 fields containing actively dividing cells were counted from each of 5–11 individuals and averaged to give the mitotic index (mitotic figures/field). The proportion of mitotic figures in anaphase was similar across genotypes. No differences in mitotic index more extreme than 0.5–1.5 times the control values were observed (Table 1), a range that Gatti and Baker (1989) defined as equivalent to the control. Neither were chromosomal aberrations (in ploidy, condensation or integrity) seen, with or without colchicine and hypotonic shock.

The normal chorion comprises two layers of approximately equal depth: the exochorion is fairly homogeneous, while the endochorion is divisible into roof and floor connected by pillars (Fig. 3A). Abutting the floor are the waxy layer and vitelline membrane. All three of the abc mutants contained chorion defects in mature stage-14 oocytes and laid eggs (Fig. 3), most severe in FSA5 and least in FSPL. The structure of the exochorion was much thinner in FSA5 (Fig. 3D). In the endochorion, both FSA5 and FS27 had few pillars, the roof was thin and debris was often seen in the spaces between the pillars (Fig. 3C,D); FSPL had fewer than normal pillars (Fig. 3B). The fact that 10% of FSA5 and FS27 embryos did not survive dechorionation suggests problems as well with the innermost eggshell layer, the vitelline membrane.

Chorion defects were visible in whole mounts in 30 % of the eggs from mothers hemizygous for one of the stronger alleles and 3 % of FSPL eggs, compared with 0.3 % of eggs from the hemizygous control. We describe two overlapping classes of defects, one involving the appendages and one the overall thickness of the chorion. The dorsal chorionic appendages are respiratory elaborations on the anterior, dorsal side of the embryo that extend outward in antennalike fashion (Fig. 3E). Changes in the dorsal appendages occurred in both FSA5 and FS27, more frequently in the former. There was a range of egg morphs: in some eggs shorter and fatter appendages extended laterally around the embryo and sometimes fused at the base in the ventral midline, forming a ring around the embryo (Fig. 3F). Variable expressivity was apparent, in that a single female laid different morphs in a 24-h period. The appendages adhered to each other in many of these embryos and sometimes occluded the micropyle (point of sperm entry), which may have been a factor in the increased number of apparently unfertilized eggs observed in these mutants. In some cases the egg shape itself differed, with a collar-like region at the anterior end (Fig. 3F); and in a few, this end was missing, so that the appendages came off the top of a shorter egg. In the other class of defect, the chorion was translucent instead of opaque as in wild type; since the appendages were present, this probably means thin or missing exochorion.

At least one mutation affecting DNA replication (gnu) shows a mutant phenotype in the absence of fertilization (Freeman and Glover, 1987), so we looked at unfertilized eggs and found that the ABC mutant phenotype was not produced. No changes in the cortical actin or tubulin patterns were seen in unfertilized eggs laid by virgin females (mutants or control) from 0–24-h collections, nor was there evidence of mitotic apparatus except in polar bodies. Degenerative changes in the cytoplasm were visible in many of the eggs; these appeared earlier and were more severe in unfertilized ABC eggs than in unfertilized control eggs, suggesting that abc females hold their eggs. The large asters associated with degeneration in fertilized FSPL eggs were never seen in unfertilized eggs from either group (data not shown).

A developmental profile (Fig. 4) was obtained from four successive 1-h collections of embryos by determining the proportions of embryos at different nuclear division stages: in the first group ages (counted from the time of egg deposition) in the population ranged from 0.5–1.5 h and in the last, 3.5–4.5 h. Development was slower in FSPL: there was a preponderance of earlier stages and progression to the 32-nuclei stage took longer than in wild type. By 2.5 h, wild-type embryos had most of their nuclei at the surface, but most FSPL embryos still had 16 or fewer nuclei. By 4h most wild types were gastrulating, in contrast to the mutant embryos, of which about a third had nuclei at the surface, over half still had less than 16 nuclei and a very few were attempting gastrulation.

About 2% (17/850) of these FSPL embryos lived long enough to make some cuticle (at 13–16 h), but in each one it was aberrant. Some of the denticle belts were missing and parts of others were fused, twisted, and/or partial; there was no obvious pattern to the defects when different embryos were compared (data not shown). Because FS27 and FSA5 embryos cease nuclear division very early developmental profiles were not performed.

The three alleles can be ranked in terms of increasing severity of defect: fsPL<fs27<fsA5. In a 0–4-h collection of embryos, most FSPL have 16 or fewer nuclei, FS27 have 8 or fewer, and FSA5, 4 or fewer. There is a subgroup of embryos with nuclei that reach the surface, ∼30% in FSPL and ∼2% in the other two mutants. In FS27 embryos there are rarely >32 nuclei and in FSA5 rarely >16; thus developmental arrest usually has occurred by the 5th or 6th nuclear division. This same ranking is found in the chorion defects.

Because actin is involved in cytoplasmic movement and nuclear migration, as tubulin is in the mitotic apparatus, we used antibodies to them in seeking clues to their behavior in the mutants. There is a transition in wild-type embryos from a surface lawn of actin to actin caps and rings around the nuclei during different mitotic stages once the nuclei have surfaced. Tubulin-containing structures in the cortex are also reorganized during migration (Karr and Alberts, 1986). In the rare cases when nuclei in the ABC embryos reach the surface, both the actin and tubulin surface patterns change as they do in wild type; the pattern is less regular, however, and depending on the local condition of nuclei may be at different stages in different parts of an embryo. Otherwise there is no change in the surface pattern for either actin or tubulin (data not shown).

Because antibodies to tubulin visualize the mitotic apparatus of nuclei, mitotic stages and various anomalies could be discerned. Normal spindles and behavior were common in the mutants. Those embryos showing mitotic asynchrony (Fig. 2D) in regions with surface nuclei had contiguous patches (not individual nuclei) in different stages. Multiple spindles from a common pole or halfspindles occurred, and the latter might have one or no asters (Fig. 5F). Spindles were frequently in disarray (differing in size, stage, integrity) in the part of an embryo lacking surface nuclei.

Because centrosome abnormalities occur in some other maternal-effect mutants, we examined the ABC mutants using the anti-centrosome antibody Rbl88 (Whitfield et al. 1988), which reacts with both centrosomes and nuclear membranes (Fig. 5A). Centrosome behavior in these mutants seemed normal with no large-scale independent replication divorced from nuclear replication as in gnu (Freeman and Glover, 1987) or asp (Gonzalez et al. 1990), but centrosomal division was sometimes asynchronous (Fig. 5B). Small tracts of centrosomes and asters without visible nuclei were common in mutant embryos that reached blastoderm (Fig. 5B,D,F). In FSPL the brightly stained DNA wads (Fig. 5D) often had several spindles and numerous (8–22) centrosomes. Degeneration occurred earlier in mutant embryos, including loss of centrosomes from nuclei (or vice versa), so that clusters of centrosomes without nuclei or with bits of chromatin appeared. The numbers of spindle anomalies and DNA wads were higher in older FSPL embryos and may be part of the degenerative process. Very large asters developed on the surface, typically not associated with nuclei. In contrast, wild-type embryos of the same ages contained no multiple spindles, no DNA wads, no large surface asters, and no centrosomes without nuclei (Fig. 5A-C).

The abc locus is one of a few maternal-effect genes that appear to affect the cell cycle in the early Drosophila embryo (see review by Glover, 1989). Early nuclei in the F₂ ABC embryos behave differently from wild type: cleavage usually ends prematurely and often with only a few large nuclei, nuclei do not always separate cleanly, some nuclei stain more intensely than others, distribution of nuclei may be abnormal, unusual chromatin clumps develop, mitosis is often asynchronous, development is slower, syncytial blastoderm is abnormal; in addition, the chorion is defective. Thus the mutated abc gene product made by hemizygous (and homozygous fsPL) mothers is inadequate to support development up to the larval stage, and in most of the embryos not even up to gastrulation. Therefore this gene is essential in the initial stages of development. Since nuclear behavior is often aberrant from the first divisions, we consider it likely that the embryonic nuclear cycle is being disrupted and that the gene is involved in some aspect of DNA replication or nuclear division.

Mitotic divisions in homozygous larval neural ganglia appear normal and no chromosomal defects are apparent, evidence that if the gene acts in cell division its main role is maternal rather than zygotic. Many F₂ embryos with either of the stronger alleles appear not to develop and most of the rest arrest earlier than FSPL. The large nuclei characteristic of early arrest appear to be in interphase, suggesting that progress through this stage is not completed. Thus this gene might act in DNA synthesis or in preparing for or triggering mitosis (see reviews by Glover, 1989; Murray and Kirschner, 1989).

A defect in DNA synthesis could explain the seemingly disparate phenotypic characteristics of chorion defects and early developmental arrest if there is under-replication of DNA in the follicle cells and/or nurse cells, which are polyploid and interact in producing the mature oocyte. Under-replication could cause insufficient amounts of the eggshell and internal oocyte components to be produced. The chorion genes themselves are not underamplified when compared with an internal control, rDNA levels (Underwood et al. 1990), but this does not rule out lower ploidy in follicle cells.

The connections between nuclei seen frequently in the F₂ embryos may be telophase bridges, reflecting difficulties involving the mitotic apparatus, sister chromatid separation, or incomplete DNA replication. An apparent phenocopy of this bridging characteristic has been produced by aphidicolin, a DNA synthesis inhibitor that interferes with DNA polymerase a(Ikegami et al. 1979), in sea urchin embryos (Brachet and DePetrocellis, 1981) and Drosophila embryos (Raff and Glover, 1988). The telophase bridges between nuclei have been attributed to the spindles’ difficulty in separating incompletely replicated DNA (Raff and Glover, 1988). Zahner and Cheney (1990) interpret nuclear clustering in their quartet embryos as indicative of chromosome separation problems. Failure of chromosome separation with continued DNA replication is a cause of endoreduplication. The increased staining seen in a subset of nuclei in some ABC embryos would suggest endoreduplication.

The slower development we see in FSPL might be related to the cell-cycle delay observed in sea-urchin (Ikegami et al. 1979), starfish (Nagano et al. 1981) and Drosophila (Raff and Glover, 1988) embryos treated with aphidicolin. While the polymerase-aphidicolin interaction affects DNA replication, it also causes an unexplained delay in mitosis (Raff and Glover, 1988). Abnormal tubulin can also retard mitosis (Sluder, 1979; Sluder and Begg, 1983). The frequent failure of nuclei in the ABC mutant embryos to become arranged in a sphere and then an ellipse during the first 5 cleavages (as wild type does: see Zalokar and Erk, 1976) suggests some type of distribution/ migration problem. Alternatively, it could be explained by a developmental delay in nuclear division: the embryos are ‘older’ and try to allocate the nuclei appropriately, but due to delay in nuclear division and/or failure of separation (both of which may be related to incomplete DNA replication) there are too few nuclei to show the normal pattern.

In FSPL embryos, the abnormal clumps may be yolk nuclei that appear relatively premature only because the other nuclei are delayed in reaching the surface. Since wild-type yolk nuclei continue to divide in synchrony with the rest of the nuclei in cleavages 8–10 (Foe and Alberts, 1983), their polyploid state must develop after that. A second possibility is that the brightly staining, abnormal chromatin represents early polyploidy in yolk nuclei due to separation failure.

The ABC phenotype can be seen only after fertilization, in contrast to gnu (Freeman and Glover, 1987). Centrosomes in gnu embryos replicate independently of nuclei, migrate to the surface where they continue to replicate without nuclei, and show a uniform distribution (Raff and Glover, 1988). FSPL embryos have centrosome replication occurring in conjunction with nuclear replication and distribution of centrosomes is often nonunifonn, probably reflecting aberrant distribution of nuclei and nuclear breakdown. The occasional small tracts of centrosomes associated with asters but not visibly with DNA may also represent nuclear breakdown. Additionally, free centrosomes have been seen in other mitotic mutants that display syncytial embryonic phenotypes (e.g. asp, see Gonzalez et al. 1990; aurora, Leibowitz and Glover, personal communication).

The aberrant patterns of surface nuclei in FSPL seem to reflect loss of different groups of nuclei from the surface (on the basis of observations of living embryos). The apparent randomness of this variable distribution is reminiscent of gynander patterns, where a stochastic event can have a variable result depending on the point in development and the random orientation of the spindle during and following the event (Garcia-Bellido, 1979). Such an event could also explain why in early embryos some nuclei look normal and some abnormal from the beginning. The variable phenotype may be a generic response to such a stochastic event: a similar phenotype has been observed in several lines with female-sterile mutations at other loci on the second chromosome (K. Harding and E. Wieschaus, personal communication). Alternatively, such genes may act in the same pathway. Other explanations for variable phenotypes in maternal-effect mutants are environmental factors such as temperature, food and female age (Schüpbach and Wieschaus, 1989) and leaky mutations (Gatti and Baker, 1989).

Because the earliest nuclear cycles are rapid and internal, they are not easy to dissect. Phenotypic analysis of mutants defective in the cell cycle is currently the major means by which one can approach this complex process. Here we have presented such an analysis of one of a few known Drosophila maternal-effect genes in which mutations disrupt the cell cycle during early cleavage. Many loci must be active in the pathways related to early nuclear proliferation, and only through a detailed analysis of mutations in such genes can we hope to understand the processes involved in this aspect of early development. Further pursuit of the abc gene is in progress in our laboratory.

The authors are grateful for helpful discussions with T. Karr, W. Saxton and C. Cheney; for criticism of the manuscript by D. Glover, T. Karr, M. E. Newport, D. Otteson and R. Woodruff; for the gifts of taxol from M. Suffhess, of the centrosome antibody from D. Glover and W. Whitfield, and of the original PL63 allele from T. Schupbach; and for technical assistance from J. Insley and T. Bergan. Support for this work came from grants by the National Science Foundation (DCB-8718307) and the OBOR Selective Excellence Programs for Research Challenge and Academic Challenge.