ABSTRACT
The Drosophila ovary consists of repeated units, the ovarioles, where oogenesis takes place. The repetitive structure of the ovary develops de novo from a mesenchymal cell mass, a process that is initiated by the formation of a two-dimensional array of cell stacks, called terminal filaments, during the third larval instar. We have studied the morphogenetic process leading to the formation of terminal filaments and find that this involves recruitment, intercalation and sorting of terminal filament cells. Two other types of cell stacks that participate in ovary morphogenesis, the basal stalks and interfollicular stalks, also form by cell rearrangement utilizing a convergence and extension mechanism. Terminal filament formation depends on the Bric à brac protein, which is expressed in the nuclei of terminal filament cells and is cell autonomously required. Disruption of terminal filament formation, together with defects of basal and interfollicular stalk development, leads to disruption of ovariole formation and female sterility in bric à brac mutants.
INTRODUCTION
The distinctive and reproducible structure, shape and size of an organ or a whole animal is controlled by the morphogenetic properties of the participating cells. Large scale changes in the shape of a tissue originate from local changes in cell behavior. Changes in cell shape, cell rearrangement and migration have been shown to drive morphogenetic movements during development. Examples in Drosophila include gastrulation (Leptin and Grunewald, 1990; Costa et al., 1994), germband extension (Irvine and Wieschaus, 1994), endoderm migration (Reuter et al., 1993; Tepass and Hartenstein, 1994) and imaginal disc evagination (Condic et al., 1991). Most of the work on morphogenetic mechanisms in Drosophila has been focused on epithelial sheets. However, studies predominantly done on vertebrates show that mesenchymal cells also participate actively and autonomously in morphogenetic movements and pattern formation during organogenesis (for review see Bard, 1990). One organ where the morphogenetic abilities of mesenchymal cells can be studied genetically is the Drosophila ovary.
The Drosophila ovary possesses about twenty ovarioles, repeated units that serve as an assembly line for oogenesis (for review see Snodgrass, 1935; King, 1970; Mahowald and Kambysellis, 1980; Spradling, 1993). Each ovariole contains a germarium at its anterior end where oogonia are located, and the 16-germ-cell cyst is formed and surrounded by a follicular epithelium (Fig. 1). The germarium is followed posteriorly by a string of follicles of advancing developmental stages, each of which contains a differentiating oocyte accompanied by fifteen nurse cells. The ovarioles are separated from each other by epithelial sheaths. The different elements of the ovariole are linked together by three specialized cell groups of mesodermal origin, the terminal filament, the basal stalk and the interfollicular stalk. Morphologically similar, each of these is a stack of aligned disc-shaped cells. The terminal filament connects the germarium with the surrounding ovariole sheath, the interfollicular stalks interconnect the follicles, and the basal stalk, which is a transient structure during the pupal stages, attaches the most advanced follicle to a basal cell population at the posterior end of the ovary (King, 1970; Fig. 1).
When the Drosophila gonads form during embryogenesis, they have a simple structure containing approximately ten primordial germ cells which are intermingled with and surrounded by mesodermal cells (for review see Campos-Ortega and Hartenstein, 1985). During most of the larval stages, the Drosophila ovary consists of a growing spherical mass of mesenchymal somatic and primordial germ cells (Kerkis, 1930), and is transformed through complex morphogenetic events into an organ of repeated units, the ovarioles, during the late larval and early pupal stages (Kerkis, 1933; King et al., 1968; King, 1970). The morphogenetic process that leads to the development of the ovarioles starts with the formation of a field of cell stacks, the two-dimensional array of terminal filaments. Subsequently, anteriorly located somatic cells migrate posteriorly, initially guided by the regular grid provided by the terminal filaments, and subdivide the ovary into ovarioles.
We have studied the development of the terminal filaments and their regular spatial arrangement, as well as the formation of the structurally related basal and interfollicular stalks. We find that they develop through recruitment and cell intercalation of mesenchymal cells, a process resembling the formation of the notochord by convergence and extension in a chordate embryo (for review see Keller et al., 1992). Furthermore, we show that the pattern forming process through which multiple terminal filaments arise from a common cell population requires cell sorting and we propose a model describing how this might occur. In addition, we find that the Drosophila gene bric à brac (bab) is cell autonomously required for the formation of terminal filaments and is crucial for normal morphogenesis of the ovary. bric à brac has previously been shown to be required for proper segmentation and specification in leg primordia, and encodes a nuclear protein which contains a BTB domain, a conserved domain that is found in several transcriptional regulators (Godt et al., 1993; Zollman et al., 1994).
MATERIALS AND METHODS
Drosophila stocks
The P-element-induced bab alleles babP, babPRDS, babPR24 and babPR72, and the deficiency Df(3L)babPG were linked with e ry ca and maintained as TM6B, Tb e ca balanced stocks (Godt et al., 1993). The babA128ry chromosome (Bellen et al., 1989; Godt et al., 1993) is homozygous viable and fertile. The enhancer trap lines P[ry+,lacZ] AD47 and P[ry+,lacZ] XA42 (Laski, unpublished), and P[w+,lacZ] B1-93F (Ruohola et al., 1991) were used as genetic markers for terminal filament cells, and P[w+,lacZ] B1-93F also for interfollicular stalk cells. Flies used for the mosaic analysis are described below. Balancer chromosomes carried the dominant larval markers Tb or Bc and are described with other genetic markers in Lindsley and Zimm (1992). Oregon R was used as a wild-type stock. Flies were raised at 25°C on standard Drosophila medium.
Mosaic analysis of larval ovaries
Site-directed mitotic recombination was catalyzed by the heat-shock inducible FLP yeast recombinase at a FRT target element (Golic, 1991). The FRT2A insertion, which contains a w+ gene and maps proximal to bab was recombined with both babA128 and babPRDS which express β-galactosidase (β-gal) in terminal filament cells. For the experiment w; babA128FRT2A and w; babPRDSFRT2A/ TM6B, Tb females were crossed to w; FLP/+; FRT2A/+ males. As a control, females of the same genotype were crossed to w; +; FRT2A males. Eggs were collected for 3-5 hours in yeasted vials and the developing animals were kept at 25°C before and after the 2 hours heat shock at 38°C (air incubator). The heat shock was set 48, 60, 72 or 96 hours after oviposition, which corresponds to the early second, mid second, early third and mid third larval instar, respectively. 25% of all female larvae (babA128 experiment) or of those that do not have the dominant Tb marker (babPRDS experiment) are of the genotype w; FLP/+; bab FRT2A/+ FRT2A which enables mitotic recombination between non-sister chromatids. In the case of a heat shock at the early second larval instar, the probability (P) that a mosaic patch in an ovary is due to a single mitotic recombination event was calculated according to the equation: P+P2+P3…=P/(1-P)=0.18, the percentage of mosaic ovaries (see Table 1). For this equation P=0.1525, therefore 85% of the mosaic ovaries (0.1525/0.18) are due to a single recombination event. Ovaries were dissected from early prepupae (when terminal filament differentiation is complete) and stained with an anti-β-gal antibody.
Ovary preparation
Adult ovaries were dissected from 2-4 days old well-fed females. To isolate ovaries of specific developmental stages, eggs were collected on apple juice agar plates for 2 hours, and the hatched larvae were fed fresh yeast while growing at 25°C, selected at the second molt or at puparium formation and kept on fresh plates until dissection. Female larvae were recognized by the size of their gonads (Ashburner, 1989). Ovaries were dissected in phosphate-buffered saline (PBS) by opening the larva in the middle of the abdomen and pulling out the fat body to which the ovaries are connected. The fat body, or as in the case of pupae the whole inner content of the posterior body half, was directly transferred into plastic mesh baskets in microtiter plates with PBS on ice, where fixation and antibody incubations were performed. The final staining reaction was done in a glass dish. The ovaries were separated from the fat body with a fine needle in the medium in which they were mounted.
Stainings of ovaries
Antibody stainings were done as described in Godt et al. (1993). Primary antibodies were: anti-Armadillo (1: 200; monoclonal line N2-7A1; Peifer et al., 1994); anti-Bab r2 (1: 1600; Godt et al., 1993); anti-β-gal (1: 3000; polyclonal; Cappel); anti-Fasciclin III (1: 5; monoclonal line 2D5; Patel et al., 1987). Secondary antibodies were either FITC-conjugated or Biotin-conjugated (Jackson Laboratories) and were preabsorbed to a mixture of biobeads (Bio-Rad, Kuo et al., 1982) and embryos for 1 hour before usage.
To monitor β-gal activity, ovaries were fixed in 0.5% glutaraldehyde in PBS for 10 minutes, washed in PBS and incubated in prewarmed X-gal staining solution (Bellen et al., 1989) containing 0.1% X-gal (Bachem) at 37°C for 2-3 hours. For detection of actin filaments, formaldehyde fixed ovaries were incubated in 10 units/ ml rhodamine-phalloidin (Molecular Probes) in PBS overnight at 4°C. Cell nuclei were visualized by treatment with 1 μg/ ml diamino-2-phenylindole (DAPI, Sigma) in PBS for 30 minutes after X-gal staining or directly after fixation.
Ovaries were mounted in 50% glycerol-PBS for classic microscopy, in 80% glycerol-PBS for confocal microscopy and in an Epon/Araldite mixture for histological sectioning (4 μm). Pictures were taken with a Zeiss Axioskop microscope equipped with Nomarski and fluorescence optics. Confocal images were obtained using a Bio-Rad MRC-600 Confocal laser scanner, mounted on a Nikon Diaphot microscope, and CoMOS software (Biorad).
RESULTS
Development of cell stacks during ovary morphogenesis
During the third larval instar distinct mesodermal cell populations become apparent in the ovary. By puparium formation the different cell populations are arranged in layers along the anterior-posterior axis (King, 1970; Fig. 1A). There are three cell populations in the anterior region of the ovary. The anterior most cells, which we refer to as cap cells, participate in the formation of the peritoneal sheath which envelops the whole ovary. The apical cells migrate between the terminal filaments and form the epithelial sheaths that divide the ovary into ovarioles. Terminal filaments are already differentiated cell stacks. The central region is occupied by the germ cells intermingled with somatic cells that are believed to give rise to the precursors of the follicle cells and the interfollicular stalk cells in the germarium. The posterior region consists of at least two distinct cell populations: the basal stalk primordium and the basal cells that will form the calyx of the oviduct. During the larval and prepupal stages the fat body is attached to the lateral side of an ovary.
Three types of cell stacks are found in a pupal ovary (Fig.1B), and each forms at a different stage of ovary development. The terminal filaments (TFs), stacks of 8-9 cells, form during the third larval instar. King (1970) mentions that short TFs have already formed at mid third larval instar. The basal stalks (BSs), stacks of approximately 30 cells, are made during the early pupal phase; and the interfollicular stalks (IFSs), each of which comprises 6-8 cells, are generated from the mid pupal stage onward.
Terminal filaments develop successively along the anterior-posterior and medial-lateral axis
At onset of metamorphosis all TFs (around 20) have differentiated, each of them a stack of 8-9 aligned disc-shaped cells that are oriented along the anterior-posterior axis of the ovary. The stacks are arranged in a plane along the dorsal-ventral and medial-lateral axis. The Bab protein is expressed in the nuclei of TF cells and in 2-3 somatic cells that are located at the base of each TF but are not aligned with the TF cells (Fig. 2A). Bab is expressed in TF cells before they differentiate to disc-shaped cells and we have used an anti-Bab antibody to study the process of TF formation.
TFs form in the third larval instar beginning approximately 12 hours after the second molt and finishing 36 hours later shortly before puparium formation. Stacks emerge in a twodimensional pattern which is oriented along the medial-lateral and dorsal-ventral axis. The first two stacks begin to form on the medial side of the ovary (Fig. 2B). Slightly later a crescent-like staining is seen across the medial half of the ovary (Fig. 2C). A short stack of 5-6 aligned cells with flattened nuclei has formed on the medial side with cell clusters of decreasing size and irregular shape located more laterally. During the following 24 hours, new TFs appear more and more laterally and the older stacks increase in size (Fig. 3E). Approximately 32 hours after the second molt, the last TFs begin to develop at the lateral side of the ovary. These observations show that TFs form progressively from the medial to the lateral side of the ovary and that individual TFs develop gradually through an increase of their cell number (illustrated in Fig. 3H). This is consistent with observations from enhancer trap lines which label TF cells (Sahut, Godt, Laski and Couderc, unpublished data).
TF formation can be divided into four stages. These stages are most easily distinguished along the dorsalventral axis where rows of TFs develop simultaneously (summarized in Fig. 3H).
(Stage 1) TF cells have a spindle-like shape and are not organized in clusters but appear as a uniform cell layer along the dorsal-ventral axis (Fig. 3A). This cell layer lies between the anteriorly located apical cells and the posteriorly located germ cells. The beginning of TF formation is marked by the appearance of a furrow around the TF region, which indicates that TF cells leave the surface and move to the interior of the ovary (Fig. 3A,B).
(Stage 2) Clusters of each 5-6 strongly stained TF cells appear next to each other (Fig. 3B,F). The formation of clusters from the uniform cell population seems to occur by cell intercalation. These clusters have an eye-shaped appearance and differ from a mature TF by a smaller cell number, spindle-shaped cells and non-aligned cell nuclei.
(Stage 3) The clusters increase in size to the final number of 8-9 TF cells (Fig. 3C,G), which we show below to occur by cell recruitment. The process is initiated by the appearance of weakly stained TF cells posterior to the clusters (Fig. 3F). These cells join individual clusters leading to clusters of increased size.
(Stage 4) TF cells change their shape, their tapered ends becoming blunt and their nuclei almost perfectly aligned, resulting in a mature TF (Fig. 3D).
Changes in cell shape are accompanied by alterations in the distribution of actin filaments and Armadillo during terminal filament formation
To understand better how distinct TF cell clusters are formed, ovaries were stained with phalloidin, which binds to actin filaments. Before cluster formation begins, TF cells and apical cells, which are located in a layer anterior to the TF cells, cannot be distinguished from one another in phalloidin stainings because they both have a similar spindle-like shape (Fig. 4A). In early stage 2, the TF cells of prospective neighboring clusters are interdigitated with each other along the dorsal-ventral axis (Fig. 4B). Actin filaments are homogeneously distributed beneath the surface of the cells. To form distinct clusters, the cells of adjacent stacks sort out from one another. This sorting process is immediately followed by a strong bundling of the tapered ends of these cells, which leads to the eye-shaped appearance of the clusters. The bundling is accompanied by a strong accumulation of actin filaments in these ends where the cells from adjacent clusters contact each other (Fig. 4C). The cells of individual stacks remain elongated and bundled while new TF cells are added posteriorly to the stack (Fig. 4D). These additional TF cells also intercalate and sort out when they join a particular cluster. When complete TF cell clusters of 8-9 cells have formed, the apical cells begin their migration between the TFs (stage 3; Fig. 4D). The transformation from the eye-shaped clusters into mature TFs occurs when the TFs become separated from each other by the invading apical cells. At this time the TF cells change from a spindle shape into a disc shape with a blunt margin all around their circumference (stage 4; Fig. 4E). Actin filaments remain concentrated along the margins. Our results suggest that cell intercalation is the driving force in the formation of distinct TF cell clusters along the dorsal-ventral axis. In contrast, there appears to be no overlap of cells between successively emerging TF clusters along the medial-lateral axis. Even small clusters are clearly separated from each other along this axis (Fig. 4A).
The strong bundling of the TF cells and the striking actin filament distribution suggests that the tapered ends of the TF cells may be highly adhesive to each other. To investigate this possibility, we have examined the distribution of Armadillo (Arm) protein during TF formation. Arm is closely related to the vertebrate β-catenin and plakoglobin and is a component of cell-cell adherens junctions (Peifer and Wieschaus, 1990; Peifer, 1993). Most of the Arm protein is seen along the cell membranes in anti-Arm-stained larval ovaries. In an eye-shaped TF the Arm protein is found in spots along the whole surface of the TF cells but is concentrated at the tapered ends of the TF cells like the actin filaments (Fig. 5A). Interestingly, the protein becomes redistributed when a mature TF forms. Arm is still found along the membranes between the TF cells but its concentration along the margins is drastically reduced (Fig. 5B).
Terminal filaments are polyclonal in origin
During the development of a TF, its cell number increases gradually. This raises the question whether the 8-9 cells of a stack derive from a single precursor cell or whether TFs form by cell recruitment. Our morphological analysis suggests that cell rearrangement is involved in TF development which is consistent with a polyclonal origin for the cells of a TF. A genetic mosaic experiment was performed to test this hypothesis. Clones in the TFs were created by somatic recombination with the yeast FLP recombinase-FRT system (Golic, 1991; see Material and methods). The clones were genetically labeled using the TF cell marker babA128, a P[lacZ] insertion mutation that does not cause defects in ovary development. Cells of mosaic patches contain zero copies (unlabeled cells) and two copies of lacZ.
24 mosaic ovaries were examined, each deriving from a somatic recombination event in the early second larval instar, a day before onset of TF formation (Table 1). Two observations were made. First, the mosaic ovaries contain stacks where stained and unstained TF cells are mixed (Fig. 6A,E). The distribution pattern of stained and unstained stacks is different from stack to stack and appears random. Second, unstained cells are distributed over several TFs in all the mosaic ovaries (Fig. 6A,E). The mosaic stacks of an ovary are not scattered but lie together in a group. Because the frequency of mosaic ovaries is relatively low in this sample (see Table 1), we estimate that in approximately 85% of the mosaic ovaries all the unstained cells derive from a single mitotic recombination event (see Material and methods). This indicates that cells from a single clone are distributed between different TF stacks. Both observations show that TFs are of polyclonal origin and suggest that TFs form by a cell recruitment mechanism. Furthermore, this experiment allows an estimate of the number of TF progenitor cells at the early second larval instar. Taking into account an average number of 164 TF cells per ovary (20 ovarioles with about 8 TF cells each) and 10-18% unstained cells in a mosaic ovary (8 mosaic ovaries evaluated), an induced clone (all cells with either 0 or 2 copies of lacZ) occupies a fifth to a third of the TF territory. This suggests that there exist 3-5 TF progenitor cells in the early second larval instar.
Clones are produced at higher frequency by heat shock at early third larval instar when TFs start to form and at mid third larval instar when TF formation is fully in progress. In a control experiment where the FLP gene was absent, mosaic ovaries were not detected (Table 1). The position of a labeled clone in a mosaic ovary depends on the time point of mitotic recombination. In mosaic ovaries induced at early third larval instar, clonal cells are primarily found in stacks on the medial side of the ovary or extending from medial to lateral (14 out of 18 mosaic ovaries). In contrast, clones induced during mid third larval instar were found predominantly located on the very lateral side of the ovary (11 out of 16 mosaic ovaries; Fig. 6C,D). The difference in the location of early and mid third larval instar induced clonal cells apparently reflects the developmental gradient in TF formation. Mosaic ovaries produced at third larval instar, like at second larval instar, contain TFs consisting of stained and unstained cells. Because of the high frequency of mosaic ovaries induced at third larval instar, the unstained cells likely correspond to multiple recombination events. The late-induced clones indicate that the gradual development of a TF is based on a sequential recruitment of TF cells, which do not come from a large preformed cell pool but rather from TF precursor cells that continue to divide in parallel to the process of TF formation.
Basal stalks and interfollicular stalks form by cell intercalation
BSs and the first IFSs appear in the ovary during the pupal stage. The expression of Fasciclin III on the surface of IFS cells (Brower et al., 1981) and BS cells was used to follow their development. At the end of the third larval instar, the BS primordium is located between the germ cells and the basal cell population (Figs 1A, 11E). When the apical cells that will form the epithelial sheaths migrate from anterior to posterior during the first day after puparium formation, they subdivide the germ cells as well as BS cells into small subpopulations, and a thick basement membrane (tunica propria) is laid down at the interface of the intruding apical cells and the enclosed cells (King, 1970). The BS cells have a rounded cell shape during this period (Fig. 7C), and their cell number increases (data not shown). During the second day after puparium formation, the BS subpopulations, each about three cell diameters wide, transform into long stalks, each only one cell diameter wide (Fig. 7). The narrowing and elongation of the BS primordia is accompanied by striking changes in the shape of the BS cells. The BS cells flatten, in the same direction (anterior-posterior) into which the whole cell population extends (Fig. 7D). BS cells apparently intercalate with each other. Cells along the tissue boundary are wedge-shaped during this morphogenetic process (Fig. 7E), with the blunt side contacting the basement membrane. At the end of the process, the BS cells are surrounded by basement membrane along its whole circumference and they become disc-shaped (Fig. 7F). The BS is only a transient structure and begins to degenerate during the third day after puparium formation.
IFSs connect the follicles of an ovariole and appear as soon as follicles separate from the germarium, a process that starts about 2.5 days after puparium formation and continues through the adult life. When IFS cells become recognizable as a distinct cell group between the germarium and a budded follicle, they form a short and thick stalk (Fig. 8A). Similar to BS development, intercalation of the wedge-shaped IFS cells and a subsequent change in cell shape leads to the formation of a single row of disc-shaped cells (Fig. 8B,C). A group of BS or IFS cells that will give rise to a single stalk is already enclosed by epithelial sheath cells prior to stalk formation. This suggests that the individual BSs or IFSs develop independently from each other, in contrast to the formation of TFs, where multiple stacks develop from an undivided common cell population.
bab mutant ovaries lack ovarioles
Mutations in the gene bab cause a mutant phenotype in legs (Godt et al., 1993) and in ovaries. bab mutant females are sterile and have ovaries of an aberrant morphology (Fig. 9). In comparison to the cone-shaped wild-type adult ovary, the bab mutant ovary is very small and abnormally shaped. Ovaries that are homozygous for strong hypomorphic alleles, such as babPRDS, babPR24 and babPR72, which do not express the Bab protein at a detectable level, have a severely reduced ovariole number or lack ovarioles altogether. A dominant haploinsufficient effect of bab mutations, previously described for legs, has not been observed for ovaries. The bab mutant adult ovary phenotype is complex. This study analyzes the requirement of bab on the development of TFs, BSs and IFSs. Additional aspects of the bab mutant phenotype will be described elsewhere (Sahut, Godt, Laski and Couderc, unpublished data).
bab mutations disrupt the formation of terminal filaments
Ovary defects in bab mutants are first detected in the third larval instar. Wild-type ovaries take on an oval shape when the TFs develop. In contrast, bab mutant ovaries maintain a nearly round shape, and analysis shows that TFs fail to form. Enhancer traps that express lacZ in TFs were used to study the bab mutant ovary phenotype. These included the strong allele babPRDS, as well as AD47, XA42 and B1-93F, which map to other loci and which were used in the background of the β-gal negative babPR24 and babPR72 mutations. The analysis shows that TF cells are affected in a strong hypomorphic bab mutant to a variable degree (Fig. 10A,B,C). (1) The number of cells that express the four TF markers is severely reduced in the TF region as compared to wild-type. (2) Most of these cells have aberrant morphogenetic properties. They have a rounded cell shape and form loose aggregates that do not differentiate into stacks. (3) A very few TF cells have a flattened cell shape and form short and irregular stacks of 2-5 cells. This probably reflects some residual bab activity in the examined bab mutants; however, study of bab null alleles will be necessary to exclude the possibility that TF formation is to a minor degree independent of bab function. (4) babPRDS hemizygous ovaries have cells in the posterior region that express lacZ, which is not seen in babPRDS heterozygous ovaries (Fig. 10A,B). These cells show a lower staining level than the anterior cells, have a rounded shape and show signs of degeneration. Because the posterior cells also express a TF marker unrelated to bab (B1-93F; Fig. 10C) the simplest assumption is that these are TF cells that are misplaced due to cellular defects which would also explain the strongly reduced number of TF cells in the anterior region. Alternatively, TF-specific genes might be ectopically expressed in another cell population in bab mutant ovaries and the reduced number of TF cells might be due to cell death or an altered cell fate.
bab is required cell autonomously for terminal filament development
To study whether bab is autonomously required in TF cells, a mosaic analysis was conducted using the strong allele babPRDS which also serves as a TF cell marker (see Material and methods). babPRDS homozygous cells stain detectably stronger in an anti-β-gal assay than heterozygous cells and can therefore be easily identified. Clones were induced at early or mid second or early third larval instar, and 11, 58, and 15 mosaic ovaries were recovered, respectively. Unstained cells, which lost the marker and are wild type, and light-colored heterozygous cells have a flattened shape and form normal TFs in the mosaic ovaries (Fig. 10D-G). Most of the darker stained bab mutant cells have a rounded cell nucleus, are scattered or form loose aggregates in the TF region, but do not form normal stacks (Fig. 10D,E,I). Occasionally, dark stained cells were found integrated into a TF. This finding is not surprising because in non-mosaic bab mutant ovaries a few TF cells form rudimentary stacks. Interestingly, the alignment of TF cells in a mosaic stack is disturbed at the point where a mutant cell is located, which can lead to a kink in the TF (Fig. 10E,H). Furthermore, the mosaic ovaries contain in addition to the mutant cells in the TF region lacZ expressing cells ectopically in a more posterior region of the ovary similar to non-mosaic bab mutant ovaries (Fig. 10D). These observations show that bab mutant TF cells display autonomously aberrant morphogenetic features.
bab mutations affect the development of basal stalks and interfollicular stalks
Disruption of TF formation is the earliest defect observed in bab mutant ovaries. Subsequently, defects develop involving other somatic cell populations. Staining with anti-Fasciclin III reveals defects in the BS primordium from puparium formation onwards. In wild type, Fasciclin III is expressed in the BS primordium and the apical cell population at puparium formation (Brower et al., 1980 and 1981; Fig. 11E). In strong bab mutant ovaries, no cells or only a small number of cells are stained in the BS region and the stained apical population is strongly reduced in size, which causes the remaining TF cells to be located more anteriorly than in wild type (Fig. 11F). Increased cell death is observed in both regions (data not shown).
In wild-type adult ovarioles, the follicles are separated from each other by interfollicular stalks (Fig. 11A,B). The remaining ovarioles in bab mutant adult ovaries contain follicles of different developmental stages, which are partially or completely fused and of irregular shape (Fig. 11C). To identify IFS cells in the disorganized ovaries of bab mutants, the enhancer trap B1-93F was used as a genetic marker (Ruohola et al., 1991; Fig. 11D). Stained cells are found between adjacent follicles of bab mutant ovarioles. These cells however, are not aligned in a single row but are organized as an irregular band (Fig. 11C,D). The larger number of β-gal-positive cells per cluster, as compared to a wild-type IFS, seems not to rely on an altered cell fate decision between IFS and polar follicle cells, which have been shown to derive from common precursors (Ruohola et al., 1991) (data not shown). Even though the IFS cells do exist in the rudimentary ovarioles of bab mutant ovaries, they are not able to arrange into a stack. This phenotypic trait is similar to the failure to form TFs in the larval ovary.
DISCUSSION
Ovariole development begins with formation of the TFs in the ovary during the third larval instar. We analyzed the development of the two-dimensional array of TFs and compare this process with the formation of the BSs and IFSs. Furthermore, we show that bab is crucial for TF formation and the subsequent morphogenesis of the ovary.
Cell stacks in the ovary form by cell recruitment
TFs together with BSs and IFSs represent a specialized tissue type containing disc-shaped cells organized in an one-dimensional stack. Single rows of cells could form by different mechanisms. They could form by polar cell division of a stem cell, which produces a string of daughter cells as is seen with germinal bandlets derived from teloblasts in a leech embryo (Weisblat and Shankland, 1985), or they could form by consecutive divisions of aligned cells. In both cases, cells are directly clonally related and cell clones are found in specific and repeated spatial patterns. Alternatively, single rows of cells could form by recruitment of primordial cells. An example is the formation of the notochord in ascidian embryos (Nishida and Satoh, 1985). We have shown that TFs are composed of cells of different lineages that are randomly mixed. This rules out the possibility that a TF derives from a single progenitor cell and argues strongly for a recruitment mechanism. Clonally mixed cell stacks also indicate that cell rearrangement plays a role in the formation of TFs. We believe that the BSs and IFSs also form by cell recruitment because both stalks develop through the reorganization of a preexisting cell population.
Cell stacks form through cell rearrangement
Cell rearrangement is a powerful morphogenetic mechanism through which local cell behavior can result in the change of the shape of a whole tissue. It is responsible for convergence and extension, which drives gastrulation of sea urchin (Hardin, 1989), zebrafish (Warga and Kimmel, 1990) and Xenopus (Wilson and Keller, 1991) and leads to the transformation of the spherical shape of the egg into the axial structure of a chordate embryo (for review see Keller et al., 1992) and to the analogous extension of the germband in the Drosophila embryo (Irvine and Wieschaus, 1994). The basis of the convergence of a cell population is a change towards a flattened cell shape and towards a bidirectional motility of the involved cells, which enables the cells to intercalate between each other. The intercalation leads to an extension of the tissue perpendicular to the cell movement. A hallmark for this type of morphogenetic movement is the apparently paradoxical situation that cells flatten in the same direction in which the whole tissue will eventually extend. This is different from the process where the change in tissue shape reflects the change in cell shape as shown for the evagination of the imaginal discs in Drosophila (Condic et al., 1991). By analyzing the different developmental stages in BS and IFS formation, we could show that these single rows of cells derive from a primordium which is initially much shorter and several cell diameters wide. During the transition to a single row, cells achieve a flattened shape and intercalate perpendicular to the direction the tissue will extend.
If cells are freely motile in two directions their movements would equilibrate each other and the tissue would not change in shape. However, convergence can be achieved by a restriction in the motility of the cells along the tissue boundary, so that these cells can not leave the boundary, as it has been the-oretically proposed (Weliky et al., 1991) and experimentally demonstrated in tissue explants of Xenopus gastrulae (Wilson and Keller, 1991). The primordium of each BS or IFS is enclosed by a thick basement membrane prior to stalk formation (King, 1970). During stalk formation, the BS and IFS cells that flank the tissue boundary have a wedge-shaped appearance suggesting that these cells have established stable contact to the basement membrane with one side and use their free protrusive side for intercalation. When a single row of cells is formed, each cell is in contact with the basement membrane along its circumference and as a consequence takes on a disc-shaped appearance with a blunt margin all around. The formation of the BS and IFS in the ovary of Drosophila appears to be an example of classic convergence and extension (Fig. 12A).
Terminal filament formation involves cell sorting
TFs develop from an apparently uniform cell population that has not been divided into subpopulations prior to stalk formation. Therefore, in contrast to BS and IFS development, the formation of a single TF can not be regarded as a process independent from the formation of the neighboring TFs. The two-dimensional array of TFs arises through a morphogenetic process in which TFs develop progressively along the mediallateral axis of the ovary but appear simultaneously along the dorsal-ventral axis. Along the dorsal-ventral axis, an initial movement of TF cells leads to aggregates where cells of future neighboring stacks are interdigitated. Subsequently the TF cells sort out by reverse intercalation and form separated clusters. Based on tissue cultures, it has been proposed that sorting-out of different cell types is due to differences in their adhesive properties (Townes and Holtfreter, 1955; Steinberg, 1970), and it has been shown that cells sort out from one another differing in the expression of different cadherin adhesion molecules (Nose et al., 1988; Friedlander et al., 1989) or in the expression of different levels of a single cadherin type (Friedlander et al., 1989; Steinberg and Takeichi, 1994). A sorting of identical cells, as seen with TF cells, might rely on a differential distribution of cell adhesion molecules on the surface of individual cells. The most striking feature in TF formation is the appearance of an eye-shaped cell cluster immediately after the cell sorting process. Eye cluster formation results from a strong bundling of the TF cell margins, which might prevent further intercalation movements so that the TF cells become trapped in a particular cluster. The strong bundling of the TF is accompanied by a high concentration of actin filaments and the β-catenin and plakoglobin-related protein Arm in the tapered ends of the cells. In mammalian cell cultures accumulation of β-catenin and plakoglobin at the cell surface correlates with an increased degree of cadherin-mediated adhesion between cells (Bradley et al., 1993; Hinck et al., 1994), and the arm mutant phenotype in oogenesis is consistent with the proposed role of Arm in cell adhesion (Peifer et al., 1993). Taken together, it suggests that the margins of TF cells exert a highly attractive force upon each other. If the adhesion between the margins of the TF cells is higher than the adhesion along the remaining cell surface, it is possible that a TF cell intercalates until its tapered margin has gained contact to the margins of other TF cells, which leads to a sorting of the cells until they are organized into distinct cell clusters. That a moderate level of adhesion promotes cell motility but a high level of adhesivity prevents cell movements has been shown for cell-substrate interactions (e.g. DiMilla et al., 1993). In summary, we propose that differences in the adhesive properties along the surface of an individual cell play a critical role in the pattern forming process that facilitates the development of multiple TFs from a common cell pool (Fig. 12B).
There is no indication that non TF cells are involved in the development of distinct TF cell clusters. Distinct clusters have already formed before the apical cells penetrate the TF cell region suggesting that the formation of the two-dimensional array of TF cell stacks is based on the morphogenetic properties of TF cells. However, the transformation of the eye-shaped cluster towards a mature TF might be related to the invasion of the apical cells and the formation of a basement membrane which provides a new substrate for the TF cell margins. The change towards a disc-like cell shape with a circular blunt margin and the relocalization of Arm, but not of F-actin, away from these margins suggests a transition from cell-cell towards cell-substrate adhesion.
bab is required for ovary morphogenesis
bab plays a central role in ovary morphogenesis and is the first gene described to control this process. bab mutant females are sterile and have small and disorganized ovaries. The absence of ovarioles in bab mutant ovaries results from developmental defects in morphogenesis during the third larval instar and early pupal stages. The primary defect appears to be the failure to form TFs, which correlates with the specific expression of the Bab protein in the nuclei of TF cells, beginning with their appearance during the third larval instar.
Our phenotypic analysis shows that bab is required for TF formation. bab mutant TF cells have an abnormal shape, are not able to form normal stacks, and appear to be partially located in ectopic positions in the ovary, indicating that they are affected in their morphogenetic properties. This phenotype is cell autonomously expressed. In addition to the TFs, other somatic cell populations in the ovary are also affected at the onset of metamorphosis. The BS primordium and the apical cell population are strongly reduced in size. Because the Bab protein is only detected in TF cells and appears to be cell autonomously required, one possibility is that the absence of TFs is responsible for these additional phenotypic traits. Alternatively, bab may be directly involved in the proper development of other somatic cell types in the ovary, including the IFS cells. The babP enhancer trap expresses β-gal not only in TFs but also at lower levels in BSs, in forming IFSs, as well as in some other somatic cells of the adult ovary (Sahut, Godt, Laski and Couderc, unpublished data). In addition, the recently identified BTB-II transcript that is related to bab, based on the same chromosomal map position, sequence homology and an over-lapping expression pattern (Zollman et al., 1994), is expressed in TFs, the apical cells and the BS primordium of a prepupal ovary (unpublished observations), and bab mutations may affect both transcripts.
A characteristic defect of oogenesis in bab mutants is fused follicles. This is at least partially attributable to a failure in IFS formation, although the IFS cells, identified with a genetic cell marker, are present at their proper location. IFS cells, like TF cells, are not able to arrange into stacks. This is different from the phenotype of mutations in the neurogenic genes, where the fusion of follicles results from the absence of IFS cells due to an altered cell fate (Ruohola et al., 1991). Therefore, if bab acts directly in IFS development it is expected to function down-stream of the neurogenic genes in this process.
ACKNOWLEDGEMENTS
We thank Jean-Louis Couderc, Ulrich Tepass and the UCLA Embryology Club for refreshing discussions and Ulrich Tepass, John Fessler and Patricia Green for very helpful comments on the manuscript. We are grateful to Wen Juan Yu for excellent technical assistance. We thank Cahir O’Kane for the babA128 line, Hannele Ruohola-Baker and Yuh Nuang Jan for providing the B1-93F, FLP and FRT2A fly stocks, Corey S. Goodman for generous supply of Fasciclin III antibody, and Mark Peifer for the gift of Armadillo antibody. F. A. L. is a Pew Scholar in the Biomedical Sciences. Funding was also provided by the NIH.