Morphogenesis of the Drosophila fusome and its implications for oocyte specification

Gametes in many organisms develop within small clusters or cysts of germline-derived cells (reviewed in Telfer, 1975; Büning, 1994; de Cuevas et al., 1997). Germ cell cysts are formed by a process of mitosis with incomplete cytokinesis, which keeps the daughter cells linked by stable intercellular bridges called ring canals. In most male cysts, the cells remain synchronized and share a common fate throughout their development into spermatocytes. In female cysts, however, generally only one cell develops into an oocyte; the others become nurse cells, which synthesize materials that are transported through the ring canals into the oocyte, to support its growth. Organisms that use this strategy, such as higher insects, are able to produce relatively large eggs much more quickly than those without persistent nurse cells.

In the Drosophila ovary, cysts arise from germline stem cells located at the tip of each ovariole, in the germarium (Fig. 1; see Spradling, 1993, for a review of oogenesis). These stem cells divide asymmetrically to produce a new stem cell and a cystoblast, which then goes through four synchronized rounds of mitosis to form a cluster of 16 cells, or cystocytes. Because each division plane is oriented relative to the others during these mitoses, the 16 cystocytes are always interconnected in a stereotypical pattern, with two cells containing four ring canals each, and the others containing three, two or one. Following cyst formation, oocyte-specific transcripts, proteins and other cytoplasmic components accumulate in one cell; invariably, it is one of the two cells with four ring canals. This accumulation, which depends on polarized, microtubule-based transport, appears to be required for oocyte differentiation (Mahowald and Strassheim, 1970; Koch and Spitzer, 1983; Suter and Steward, 1991; Theurkauf et al., 1993). The transported oocyte determinants have not yet been identified, but they are likely to include factors that establish and maintain meiosis in the oocyte (Lilly and Spradling, 1996).

How the oocyte is specified is controversial. One model is that oocyte specification occurs after the cyst mitoses are complete: first, both of the cells with four ring canals are selected as ‘pro-oocytes’, but later one is chosen as the oocyte while the other reverts to the nurse cell fate (Carpenter, 1994; González-Reyes et al., 1997). This view is based on the observation that the two cells with four ring canals are initially difficult to distinguish; both enter meiosis and form synaptonemal complexes, and both accumulate factors transported from the nurse cells (Mahowald and Strassheim, 1970; Carpenter, 1975, 1979). How the asymmetry between the two pro-oocytes could arise is unclear. Another model is that the oocyte is specified at the first cyst division, by the establishment of some system of polarity that is maintained in subsequent divisions (Lin et al., 1994; Theurkauf, 1994). Some support for this model comes from studies that show the fusome, a prominent cytoplasmic organelle, associates with only one pole of the spindle at every cyst mitosis (Storto and King, 1989; Lin and Spradling, 1995; McGrail and Hays, 1997). Thus, each cystocyte division is inherently asymmetric.

Fusomes are found in developing cysts in Drosophila and many other insects (reviewed in Telfer, 1975; Büning, 1994; de Cuevas et al., 1997; McKearin, 1997). Ovarian fusomes were first identified in stained Dytiscus (diving beetle) specimens (Giardina, 1901) and later in electron micrographs of Drosophila germaria, where the organelle appears as a distinct and highly vesiculated region of cytoplasm (Koch and King, 1966; Mahowald, 1971; Telfer, 1975). Cystoblasts contain a small spherical fusome, which grows during cyst formation into a large branched structure that extends through the ring canals into every cell. Once the cyst has stopped dividing, the fusome begins to break down, usually disappearing soon after the cyst leaves the germarium. In Drosophila, a spherical fusome, also called a spectrosome, is found in germline stem cells (Lin et al., 1994). Although the association of the fusome with one pole of the mitotic spindle has been well documented in both stem cells and cysts (Storto and King, 1989; Lin and Spradling, 1995, 1997; Deng and Lin, 1997; McGrail and Hays, 1997), its growth and behavior at other points in the cell cycle have not been analyzed.

Recent studies have identified several components of the fusome and have begun to shed light on its function. Fusomes in Drosophila females contain the membrane skeletal proteinsα-spectrin, β-spectrin, ankyrin and the adducin-like hu-li tai shao (hts) protein (Lin et al., 1994; de Cuevas et al., 1996). Mutations in the genes encoding hts and α-spectrin essentially abolish the fusome and cause the formation of cysts with fewer than 16 cells, which usually lack an oocyte (Yue and Spradling, 1992; de Cuevas et al., 1996). The microtubule motor dynein associates with the fusome in a cell cycle-dependent manner and is also required for fusome integrity and cyst division (McGrail and Hays, 1997). Fusomes in Dynein heavy chain (Dhc64C) mutants are misshapen and fragmented, and Dhc64C cysts contain too few cells and no oocyte. Another fusome component, the protein expressed by the bag-of-marbles (bam) gene, is thought to be involved in forming the membranous network of the fusome; it is required for cystoblast differentiation and may regulate cyst divisions (McKearin and Ohlstein, 1995; McKearin, 1997). Finally, the cell cycle regulator cyclin A also associates with the fusome and is involved in regulating cyst formation (M. Lilly, M.C. and A.C.S., unpublished data). Taken together, these results show that the fusome is an integral part of the mechanisms that guide cyst construction. Moreover, they suggest that the fusome not only is a marker of cyst polarity, but also participates directly in establishing or maintaining this polarity.

To gain a greater understanding of how and when cysts become polarized, we have closely examined fusome morphogenesis throughout cyst formation, from stem cell to 16-cell cyst. Our analysis shows that the fusome grows in all cells by a regular, cyclic process of accumulation and fusion of fusome pieces. Furthermore, the fusome remains asymmetrically distributed within the cyst throughout this time, supporting the idea that the oocyte is specified at the first cystocyte division.

All flies were maintained at 22-25°C on standard medium; they were transferred to bottles containing wet yeast for 1-2 days before dissections.

Ovaries were dissected in Grace’s medium and incubated in Grace’s medium plus 10 μM bromodeoxyuridine (BrdU) for 1 hour at room temperature. They were washed for 5 minutes in Grace’s medium and 10 minutes in PBS (10 mM NaH₂PO₄/Na₂HPO₄, pH 7.4, 175 mM NaCl), fixed for 10-20 minutes in Buffer B (16.7 mM KH₂PO₄/K₂HPO₄, pH 6.8, 75 mM KCl, 25 mM NaCl, 3.3 Mm MgCl₂) plus 5% formaldehyde (Ted Pella Inc.), washed for 30 minutes in PBS plus 0.6% Triton X-100 and for 10 minutes in DNase I buffer (66 mM Tris, pH 7.5, 5 mM MgCl₂, 1 mM 2-mercaptoethanol), incubated in 0.5 ml DNase I buffer plus 25 u DNase I (Boehringer Mannheim Biochemicals) for 30 minutes at 37°C, and washed for 30 minutes in PBS plus 0.1% Triton X-100. Ovaries were then labeled with anti-BrdU and other antibodies (see below).

Affinity-purified rabbit anti-anillin antibodies (Field and Alberts, 1995), unlabeled or directly labeled with Oregon Green, were a gift from C. Field and were used at a concentration of 1-5 μg/ml. Bam protein was labeled using a mouse anti-BamF antiserum (McKearin and Ohlstein, 1995), which was a gift from D. McKearin, at a 1:100 dilution. Mouse anti-BrdU antibodies were purchased from Becton-Dickinson and used at a 1:20 dilution. Anti-α-spectrin, anti-hts and secondary antibodies are described in de Cuevas et al. (1996).

Ovaries were dissected, fixed and stained as described previously (de Cuevas et al., 1996). They were mounted in PBS plus 50% glycerol and 1 mg/ml p-phenylene diamine (Sigma) for microscopy. Stained ovaries were examined by epifluorescence and confocal microscopy on a Leica TCS NT or on a Zeiss Laser Scan Microscope. Figs 2F, 2I, 3C-E and 7A-C are single optical sections; all other confocal images are stacks of multiple sections. DNA stained with 4,6-diamidino-2-phenylindole (DAPI) was photographed with a Princeton Instruments cooled CCD camera.

To identify dividing germline cells, we labeled ovaries with antibodies that recognize the actin-binding protein anillin. In Drosophila embryos and tissue culture cells, anillin is expressed in actively dividing cells and is localized in a cell-cycle-dependent manner (Field and Alberts, 1995). During interphase, it accumulates in the nucleus, but in mitosis, following nuclear envelope breakdown, it is released into the cytoplasm and moves out to the cortex. At telophase, anillin becomes highly enriched in the cleavage furrow, where it remains until the connection between the sister cells is severed. It is usually not detected in cells that have left the cell cycle. In germline stem cells and cystocytes, anillin follows the same cell-cycle-dependent pattern of localization; however, it persists in the ring canals that link the cystocytes long after the cells have stopped dividing, at least until the cyst has left the germarium (Fig. 2A; see also Fig. 8 in Field and Alberts, 1995). We have found that anillin also localizes to a transient ring canal that arises when a stem cell divides to produce a daughter stem cell and a cystoblast. When it is first formed, this ring canal is similar in size and shape to the ring canals that link dividing cystocytes (Fig. 2A,B). It is not a permanent structure, however; before either cell divides again, the ring canal shrinks in diameter and the hole in its center disappears, severing the connection between the two cells (Fig. 2C). Usually no detectable anillin remains at the site of the severed connection, though occasionally a small spot can be found on the membrane of a dividing cystoblast (not shown). This stem cell ring canal was identified previously in electron micrographs of Drosophila germaria (Carpenter, 1981), but to our knowledge anillin is its first molecular component to be identified.

Having a marker for the stem cell ring canal has allowed us to identify stem cells that have recently completed mitosis, whose fusomes are likely to be in the process of segregating between the two daughter cells. To analyze the behavior of the fusome in these cells, we fixed ovaries and triple-stained them with anti-anillin antibodies, anti-hts antibodies (which label the fusome) and the DNA dye DAPI; we then examined them by immunofluorescence and confocal microscopy. Stem cells were identified by their location in the germarium, directly abutting the base of the terminal filament. We determined the cell cycle phase of a stem cell from its nuclear morphology, from the location of anillin within the cell, and from the presence or absence of a ring canal. The results of this analysis are shown in Fig. 2F-M (DAPI not shown). In late interphase, when the stem cell has a high level of anillin in its nucleus and there is no ring canal attaching it to any other cell, the fusome is spherical in shape and is located at the anterior tip of the stem cell, adjacent to the base of the terminal filament (Fig. 2F). Throughout mitosis, when anillin is enriched at the cell cortex, the fusome stays in this location and usually remains spherical (Fig. 2G), although occasionally it flattens out along the membrane of the stem cell, forming a wedge-shaped structure (not shown). Regardless of its shape, however, in most stem cells the fusome appears to stay the same size throughout mitosis; we did not see an eight-fold reduction in its volume, as reported by Deng and Lin (1997). Previous studies have shown that, at metaphase and anaphase, the fusome is also adjacent to the anterior pole of the mitotic spindle (Deng and Lin, 1997; Lin and Spradling, 1997).

After mitosis, when the nuclear membrane has reformed and the cleavage furrow is well advanced, the fusome begins to migrate posteriorly towards the cleavage furrow (Fig. 2H,I). At the same time, a small amount of fusomal material accumulates in the nascent ring canal, forming a ‘plug’ in the ring. Although we could not determine its origin(s) from this experiment, we suggest that the plug is formed at least in part from newly synthesized material, since the original spherical fusome does not become noticeably smaller as it forms. As interphase progresses, the plug enlarges and the original fusome elongates along the anterior-posterior axis of the stem cell, eventually contacting the growing plug (Fig. 2J). The resulting single fusome then reorganizes and/or accumulates new material until it forms a thick bar-shaped structure that extends through the ring canal (Fig. 2K). Later, as the ring canal closes, it appears to squeeze the bar-shaped fusome into two pieces (Fig. 2L). Because of the appearance of the fusome, we call this the ‘exclamation point’ stage. At this stage, the fusome is usually distributed unequally between the two cells; thus about one-third of it ends up in the cystoblast, and the remainder in the stem cell. The bar-shaped fusome and its unequal distribution between the two cells were previously reported by Deng and Lin (1997).

When there is no anillin at the membrane (and presumably no connection) between the stem cell and cystoblast, the stem cell fusome returns to the anterior tip of the cell and regains its spherical shape (Fig. 2M). In stem cells that appear to have separated only recently from a cystoblast, the fusome looks smaller than in other stem cells (compare Fig. 2F and M); hence it might accumulate more material before the stem cell divides again. In some stem cells, the fusome appears to remain connected to the anterior tip of the cell throughout interphase. After the bulk of the fusome has migrated posteriorly, a thin ‘thread’ of material is visible between it and the anterior membrane. The fusome can also separate into two pieces; one piece migrates posteriorly and fuses with the new fusome plug, while the other piece remains at the anterior membrane of the stem cell. The behavior of the fusome in these cases suggests that there is a specific cortical attachment site for the fusome, from which it does not always detach completely.

Because most germaria have at least one stem cell, and sometimes several, with a ring canal and an elongated fusome, we reasoned that these stages of the fusome cycle must occupy a fairly large portion of the cell cycle following mitosis. To determine how long the transient ring canal persists, we incubated ovaries in the nucleotide analog BrdU, which incorporates into S-phase nuclei; we then fixed and triple-labeled them with anti-BrdU and anti-anillin antibodies, and with DAPI (Fig. 2D,E). In about 8-10 ovaries, we found 43 stem cells that were labeled with BrdU; each one was connected to a cystoblast by a ring canal that was fully open. Furthermore, in every case, the connected cystoblast was also labeled with BrdU. No single, unconnected stem cells or cystoblasts, and no stem cell-cystoblast pairs connected by a closed ring canal, were labeled with BrdU. These results show that the ring canal between stem cell and cystoblast stays open until after S-phase of the following cell cycle is complete in both cells. Thus, we conclude that the exclamation point stage of the fusome cycle occurs in G2. We have confirmed this result by examining ovaries double-labeled with anti-BrdU and anti-hts antibodies (not shown).

The results of these experiments also suggest that the stem cell and cystoblast enter and exit S-phase together. By examining the BrdU-labeled stem cell and cystoblast nuclei more closely, we determined that the BrdU was distributed over the same portion of the nucleus – over euchromatin, over heterochromatin, or over both – in every labeled pair of cells. Thus, we conclude that the cell cycles of the stem cell and cystoblast are synchronized through S-phase. Our findings are supported by those of Carpenter (1981), who labeled ovaries with [³H]thymidine, which incorporates into S-phase nuclei, and examined them by EM autoradiography. She identified 20 stem cell-cystoblast pairs with a connecting ring canal; in 4 of these pairs, both stem cell and cystoblast were labeled, but neither cell was labeled in the other pairs. She also found 34 single stem cells and 29 single cystoblasts, none of which was labeled. From these observations, she concluded that the stem cell and cystoblast remain connected and synchronized at least through S-phase. We suggest that they become asynchronous soon after the cells are separated, however, since we almost never saw a stem cell and a cystoblast in mitosis at the same time. We also found several germaria each with two labeled stem cell-cystoblast pairs; thus, two stem cells in one germarium are not prohibited from entering S-phase at the same time.

We next examined the growth and behavior of the fusome in dividing cystoblasts, in ovaries that were triple-labeled with anti-anillin antibodies, anti-hts antibodies and DAPI (Fig. 3). Cystoblasts were identified by their position in the germarium: any single cell without a ring canal, which was separated from the base of the terminal filament by at least one other cell, was considered a cystoblast. Its cell cycle phase was determined by the location of anillin and by its nuclear morphology. A new cystoblast, which has separated from the stem cell but not yet entered mitosis, contains a spherical fusome slightly smaller than a stem cell fusome; it is usually located in the anterior half of the cell, near the plasma membrane (Figs 2M, 3A). In mitotic cystoblasts, the fusome keeps its spherical shape and size but is often located in the posterior half of the cell (not shown). Thus, at cytokinesis, the fusome is often retained by the more posterior of the two daughter cells (Fig. 3B). This finding contradicts that of Lin and Spradling (1995), who stated that the fusome remains in the anterior-most cell. We did not determine whether the fusome changes its location by migrating through the cystoblast before mitosis begins, or whether the entire cell rotates; we suggest, however, that the fusome migrates towards the centrosome, which is likely to be located posteriorly.

Following mitosis, as in the stem cell, a small plug of fusomal material accumulates in the arrested cleavage furrow (Fig. 3C). As interphase progresses, the original fusome and plug move closer together, additional fusomal material accumulates between them, and the two pieces of fusome eventually fuse, forming a snowman-shaped structure with its ‘neck’ in the ring canal (Fig. 3D). In the cystoblast, however, the original fusome does not elongate towards the plug as they fuse. Thus, we could not determine if the original fusome moves towards the plug, or if the plug and ring canal move towards the original fusome. Because it is formed from two smaller fusomes, the fused fusome has also

been called a ‘polyfusome’ (King et al., 1982). After fusion, the cell that lacked a fusome at cytokinesis contains less fusomal material than its sister cell and, apparently, it does not accumulate more during the rest of interphase. By the next mitosis, which forms four cells from two, the fusome is still asymmetrically distributed within the cyst (Fig. 3E).

In 4-, 8- and 16-cell cysts, the fusome grows in a similar manner as in 2-cell cysts: after each round of mitosis, plugs form in each nascent ring canal and gradually fuse with the original fusome (Fig. 4A-G). Fusion takes place as the plugs move towards the central fusome and as fusomal material accumulates in the gaps between them. All plugs fuse before the next round of mitosis begins, but they do not always fuse at the same rate; we saw many cysts in which some plugs had already fused while others were still separated from the central fusome by a large distance (eg. Fig. 4E,F). In 16-cell cysts, fusion is always complete before the cyst becomes lens-shaped and the fusome begins to disaggregate. Thus, 16-cell cysts whose fusomes are in the process of fusing are located in region 2a of the germarium. It is likely that some fusomal material is newly synthesized or assembled as the cyst grows, since fusomes in 16-cell cysts are much bigger than cystoblast fusomes.

As in 2-cell cysts, after every round of fusome fusion, one cell appears to contain more fusomal material than its sister cystocytes; this cell is always one of the two with the greatest number of ring canals (Fig. 4A-G). Based on this observation, the cell with the most fusomal material in a 16-cell cyst is probably the same cell that retained the fusome at cytokinesis in the cystoblast. Thus, the initial asymmetry of the fusome is maintained throughout cyst formation. In 4-, 8- and newly formed 16-cell cysts (region 2a), the cell with the most fusomal material is positioned at random in the cyst with respect to the axes of the germarium; we could not accurately determine its location in older 16-cell cysts (region 2b), however, because the fusome begins to break down before this point.

We also examined fusome morphogenesis in two mutants in which cyst polarity appears to be disrupted. Mutations in Bicaudal-D (Bic-D^PA66/R26) and egalitarian (egl^WU50/RC12) cause the formation of cysts that fail to accumulate oocyte-specific factors in a single cell and thus differentiate into 16 nurse cells (Suter et al., 1989; Suter and Steward, 1991; Schüpbach and Weischaus, 1991; Mach and Lehmann, 1997). Fusomes appear to form normally and are still asymmetrically distributed in these cysts, however, indicating that the polarizing mechanisms that underlie fusome formation are not disrupted by these mutations (Fig. 4H-J). Thus, the fact that the oocyte fails to differentiate in these mutants is apparently not caused by a lack of polarity in dividing cysts. Consistent with this result, one of the two cells with four ring canals still localizes to the posterior end of these mutant cysts in region 2b, indicating that some polarity is retained in older cysts as well (Suter and Steward, 1991; Ran et al., 1994; Carpenter, 1994).

After each round of mitosis, as the fusome plugs move towards the central fusome, their associated ring canals move with them. To quantify this movement, we measured the distances between new and old ring canals both before movement (in cysts with newly arrested cleavage furrows, which did not yet contain a distinct plug) and after movement (in cysts whose plugs had all fused with the central fusome). After plug fusion, the newest ring canal in each cell can still be distinguished by its smaller size. In cells with more than one old ring canal, we recorded only the distance from the new ring canal to the closest old ring canal. The results of this analysis show that new ring canals move about 2-3 μm closer to old ring canals after each round of mitosis (Table 1). A similar conclusion was reached by Storto and King (1989), based on measurements made from electron micrographs of a single dividing cyst.

We could not determine from this experiment if new ring canals actually flow through the plasma membrane, or if they are brought closer to old ring canals by removal of plasma membrane from one side of the cell and its replacement elsewhere. Regardless of how it happens, the centripetal ring canal ‘movement’ results in a change in the cyst’s geometry. This change occurs after every round of mitosis but is best illustrated in 4-cell cysts, as shown in Fig. 5. In newly formed 4-cell cysts, just after cytokinesis, the cells are arranged in a linear or V-shaped structure (Fig. 5A; see also 3E). In older 4-cell cysts, however, the cells radiate out like flower petals from the central clump of ring canals. This arrangement, called a ‘rosette’ (Telfer, 1975), can be seen most clearly in cysts that are just entering mitosis, when anillin outlines the cortex of the cells (Fig. 5B). Rosettes have been found in many other insect species and their formation appears to be a general characteristic of cyst morphogenesis (Telfer, 1975; Büning, 1994). The effect of rosette formation is to shorten the distances between cells that are not directly connected; thus, it is tempting to speculate that this change in cyst geometry facilitates synchronization or communication between cystocytes. We also suggest that ring canal movement and rosette formation might be facilitated by the reduced amounts of membrane skeletal proteins found at the plasma membrane of dividing cystocytes (Lin et al., 1994; de Cuevas et al., 1996).

To characterize fusome formation in more detail, we labeled ovaries with antibodies against two other fusome components,

α-spectrin and bag-of-marbles (bam) protein. Both of these proteins, like hts protein, localize to fusome plugs in nascent ring canals and are distributed asymmetrically within the cyst throughout its formation (Fig. 6A-D). We have not yet found any difference in composition between the plugs and central fusome, but it is possible that other components might associate with only one part of the fusome.

We also examined plug formation more closely with these antibodies. By examining images at higher magnification, we found that the plugs are formed initially by material that appears to approach the ring canal from both sides. Moreover, especially when labeled with anti-hts antibodies, this material appears to approach the ring canal in lines, as if it is travelling along fibers (Fig. 6E). Thus, we suggest that the fusome plugs are assembled from material that is transported to the site of the spindle midbody, around which the ring canal forms, perhaps along microtubules that are remnants of the mitotic spindle.

In the course of these experiments, we found that anillin is also present in the somatic ring canals that link ovarian follicle cells (Fig. 7A-D; see Robinson and Cooley, 1996, for a review of ring canals in follicle cells). These ring canals, which are first visible in region 2 of the germarium, increase in number as the follicle cells divide; thus by late oogenesis many follicle cells have multiple ring canals associated with them. Unlike germline ring canals, somatic ring canals are very small – about 0.5 μm in diameter – and do not enlarge as oogenesis progresses. They also continue to stain with anti-anillin antibodies throughout oogenesis. We did not see ring canals between the cells that comprise the interfollicular stalks (except when the stalks are forming, in stages 1-2; not shown) but they are present in the follicle cells that remain over the nurse cells after the others migrate posteriorly in stage 10 (Fig. 7D). Somatic ring canals in later egg chambers also contain actin (H. Frydman and A.C.S., unpublished data). In dividing follicle cells that were double-labeled with anti-anillin and anti-hts antibodies, we did not see any fusome-like structure spanning the ring canals.

The work presented in this paper strongly suggests that ovarian cysts are polarized at the first division, that this polarity is maintained during the subsequent rounds of division, and that the fusome is a sensitive indicator of polarity in dividing cysts. Thus, our observations support the idea that the oocyte is specified early, during the cyst divisions, rather than later, after the completion of the cyst mitoses. Previous studies had shown that the fusome associates with only one pole of the spindle in dividing cystoblasts (Lin and Spradling, 1995). Since the two cells that are produced by this division give rise to the two cells with four ring canals, one of which becomes the oocyte, this result suggested that the fusome might mark the future oocyte. It was not clear, however, how or even if the original asymmetry between these two cells was maintained as the cyst continued to divide. Our results provide the first evidence suggesting that the initial polarity of the cyst is maintained.

One important question that remains is whether the cell that originally inherits the fusome does in fact become the oocyte. One of the earliest markers of the oocyte is its accumulation of specific proteins including Bic-D and Cup (Suter and Steward, 1991; Keyes and Spradling, 1997). Unfortunately, by the time these proteins are localized to a single cystocyte, the fusome has begun to break down, the ring canals have migrated away from the center of the cyst, and it is not possible to determine which cell originally contained the most fusomal material (not shown). This question has been answered in another insect, however. In the diving water beetle Dytiscus, cyst formation is similar to that in Drosophila: cystoblasts go through four rounds of synchronous mitosis to form a cyst of 16 cells, one of which differentiates into an oocyte while the others become nurse cells. Unlike Drosophila, however, the Dytiscus oocyte can be distinguished from its siblings as early as the 2-cell stage because it contains a large ring of highly amplified rDNA (Giardina, 1901; see also Gall, 1996). Moreover, the cell that inherits the amplified rDNA ring at the first division of the cystoblast is the same cell that inherits the fusome (Fig. 8A). Thus, by analogy to Dytiscus, we propose that the cell that retains the most fusomal material in Drosophila is the one that becomes the oocyte.

What then is the mechanism of oocyte specification? One possibility is that the asymmetric distribution of the fusome reflects an underlying asymmetry in the cyst’s cytoskeleton. Although no obvious polarity has yet been found in the distribution of microtubules in dividing cysts (Theurkauf et al., 1993), other differences in the microtubule or actin-based cytoskeleton might be present and might contribute to the formation of the polarized microtubule array that directs transport into the oocyte. Consistent with the view that the asymmetric fusome is indicative of a polarized cytoskeleton, fusomes in dividing male cysts, which are not likely to be polarized, appear to be more evenly distributed (not shown). Another possibility is that an oocyte-specifying factor that co-segregates with the fusome might be inherited by the future oocyte at each round of cyst division. For example, the centrosome inherited by the future oocyte at the each division could be distinct in some way (Theurkauf, 1994). The fusome, which associates with one pole of the spindle at metaphase, would thus be segregating with this marked centrosome at the first division of the cystoblast.

To explain why both of the cells with four ring canals (the two ‘pro-oocytes’) initially enter meiosis and accumulate transported factors, we suggest that transport between these two cells is delayed. As transport within the rest of the cyst begins, if the ring canal connecting these two cells is somehow blocked, then oocyte determinants would pile up in both cells and both would initially appear to differentiate as oocytes. The fusome itself might be involved in blocking this ring canal; the larger piece of fusome associated with the oocyte might simply take longer to break down than the fusome in the rest of the cyst.

Our work, together with previous studies, shows that cyst and fusome growth are highly regular, polarized processes (Fig. 8B). In each division cycle, spindles orient with one pole in the fusome (Storto and King, 1989; Lin and Spradling, 1995, 1997; Deng and Lin, 1997; McGrail and Hays, 1997), fusome plugs accumulate in the new ring canals, at the site of the spindle midbody, and ring canals and fusome plugs move centripetally, changing the cyst from a linear into a rosette-shaped structure. In addition, our observations suggest several mechanisms that might mediate these events. We have shown that fusome plugs are formed from material that appears to approach from both sides of the ring canals. If the spindle midbody contains the plus ends of microtubules, then a plus-end directed, microtubule-based motor might be involved in transporting fusome vesicles to the ring canals. A different motor might be involved in moving the fusome plugs, together with their associated ring canals, towards the central fusome. Since this movement is directed towards the centrosome, which remains associated with the original fusome in each cell, it is likely to be mediated by a minus-end directed motor. One such motor, dynein, localizes to the fusome and is required for fusome assembly; mutations in the Dynein heavy chain (Dhc64C) gene cause the formation of fragmented fusomes (McGrail and Hays, 1997). Dynein is unlikely to play a direct role in transporting fusome plugs, however, because it apparently does not localize to the fusome in interphase, when this movement occurs. Instead, it appears to be required for anchoring spindle poles in the fusome during mitosis (McGrail and Hays, 1997). Perhaps, in the absence of this mitotic orientation, interphase microtubules are also misoriented and the fusome plugs are unable to move to the center of the cyst.

Our results are supported by numerous studies of fusomes and cysts in other insects (reviewed in Telfer, 1975; Büning, 1994). These studies, based on light and electron microscopic observations of dividing cysts, show that cystocytes divide in synchrony, spindles orient with one pole adjacent to the fusome, vesicles and fibrils characteristic of the fusome accumulate within new ring canals, at the site of the spindle midbody, and many cysts form rosettes. Because these features of fusome and cyst formation are common in other insects, we suggest that the asymmetric distribution of the fusome, which we have now documented in Drosophila, may also be a general feature of ovarian cyst formation. Thus, it is tempting to speculate that oocytes are specified early in all organisms that form cysts, by the establishment of cyst polarity at the first division and its maintenance in subsequent divisions.

We thank Dan Branton, Howard Lipshitz, Dennis McKearin and especially Chris Field, for providing antibodies. We also thank Jennifer Lippincott-Schwartz, Dennis McKearin, Yixian Zheng and members of the Spradling laboratory for helpful discussions, and Horacio Frydman, Nicole Grieder, Mary Lilly, Phil Newmark and Melissa Pepling for reading the manuscript. This work was supported by the Howard Hughes Medical Institute. M. C. dedicates this work to Ian Sussex, for his encouragement and inspiration.