Production of gurken in the nurse cells is sufficient for axis determination in the Drosophila oocyte

In Drosophila, as in many invertebrates and some vertebrates, the fundamental asymmetries that define the embryonic body axes are established during oogenesis through the localization of maternally contributed cytoplasmic determinants within the egg. The restricted distribution of these factors is often achieved through localization of maternal transcripts within the developing oocyte, and differential translational regulation of the localized and unlocalized mRNAs. Polarization of both the anteroposterior (AP)and dorsoventral (DV) axes of the Drosophila embryo requires the precise localization of a single cytoplasmic determinant: the product of the gurken (grk) gene(Cooperstock and Lipshitz,2001; Roth, 2003; Van Buskirk and Schüpbach,1999). While the molecular mechanism of Grk function and the requirement for its proper localization have been well established,elucidation of the mechanism(s) regulating Grk localization requires an understanding of the source of grk transcripts. To address this issue, we have taken a genetic approach to investigate the site of grk production within the Drosophila egg chamber.

Each Drosophila oocyte develops within a cyst of 16 germline cells, which is surrounded by an epithelium of somatic follicle cells to form a developmental unit called an egg chamber(Spradling, 1993). Each 16-cell cyst is derived from a single cell, the cystoblast, through four synchronized mitoses. Cytokinesis is incomplete during these divisions,forming a stereotypic pattern of cytoplasmic bridges, called ring canals,between the cells of the cyst (see Fig. 1A). One of the cells with four ring canals develops as the oocyte, and comes to occupy the posteriormost position in the cyst, while the other 15 cells become nurse cells, which supply essential components to the oocyte through the ring canals. The AP axis of the egg chamber is established when the oocyte adopts its posterior position within the cyst. The final AP polarity of the oocyte itself is generated during mid-oogenesis through a localized signal from the oocyte to the overlying follicle cells. In later stages, another localized signaling event between the oocyte and follicle cells defines the DV axis of the egg chamber(Huynh and St Johnston, 2004; Van Buskirk and Schüpbach,1999).

In each of these oocyte-patterning steps, the signal from the oocyte is encoded by the grk gene. The Grk protein, which is related to transforming growth factor-α, functions as a spatially restricted ligand secreted by the oocyte to activate the Drosophila epidermal growth factor receptor (Egfr) in the follicle cells(Neuman-Silberberg and Schüpbach,1993), resulting in localized Egfr signaling, which generates axial polarity. In early oogenesis, grk mRNA is localized to the posterior cortex of the oocyte, adjacent to the oocyte nucleus. As the oocyte is small at this stage, the Grk signal is restricted to the overlying follicle cells at the posterior of the egg chamber and induces them to adopt a posterior fate, thus distinguishing them from anterior cells and establishing the AP polarity of the follicular epithelium(Gonzalez-Reyes et al., 1995; Gonzalez-Reyes and St Johnston,1998; Roth et al.,1995). The posterior follicle cells in turn provide a signal that induces a reorganization of the oocyte cytoskeleton and thereby confers appropriate polarity on the oocyte AP axis. One consequence of this reorganization is the microtubule-dependent migration of the oocyte nucleus to the anterior cortex of the oocyte, where its position along the oocyte circumference provides the first detectable asymmetry along the DV axis. Both grk mRNA and protein remain closely associated with the oocyte nucleus, and at this new anterior position they provide a localized signal to a second population of follicle cells, resulting in the induction of dorsal follicle cell fates and establishing the DV axis of the egg chamber.

Proper localization of Grk is crucial for the generation of axial polarity. Mutations in the maelstrom gene alter grk mRNA localization in early stages and result in defective posterior follicle cell fate determination (Clegg et al.,1997). In ovaries overexpressing grk or lacking the function of squid (sqd) or fs(1)K10, grk is not restricted dorsally but is instead distributed throughout the anterior cortex of the oocyte, resulting in expanded induction of dorsal follicle cell fates(Neuman-Silberberg and Schüpbach,1993). The asymmetric localization of grk within the oocyte requires the association of grk transcripts with hnRNPs and the activity of the microtubule motor proteins Dynein and Kinesin(Brendza et al., 2002; Duncan and Warrior, 2002; Januschke et al., 2002; MacDougall et al., 2003). However, how these factors function to achieve the unique dorsal anterior localization of grk is not well understood.

The close association of grk mRNA and protein with the oocyte nucleus has led to the proposal that anchoring of the message upon export from the oocyte nucleus mediates or contributes to its localization(Goodrich et al., 2004; Norvell et al., 1999; Palacios and St Johnston,2001; Saunders and Cohen,1999). This model has been supported by evidence suggesting that grk is transcribed primarily or exclusively in the oocyte nucleus,derived from experiments designed to exclude the possibility of transport from the nurse cells to the oocyte (Saunders and Cohen, 1999). Such a mechanism would require that grkbe a rare gene transcribed in the oocyte nucleus, which is otherwise transcriptionally quiescent and arrested in meiotic prophase(King and Burnett, 1959; Spradling, 1993). Alternatively, grk may be transcribed in the nurse cells and transported to the oocyte, as are transcripts encoding other spatially restricted determinants, such as bicoid and oskar(osk), found at the anterior and posterior poles, respectively(Johnstone and Lasko, 2001; Lipshitz and Smibert, 2000; Palacios and St Johnston,2001). Production of grk in the nurse cells would require an alternative model for localization, as a localized synthesis and retention mechanism would not apply to nurse-cell-derived transcripts. For example,microtubule-based transport has been shown to function in grklocalization within the oocyte. However, since the existence of a localization mechanism within the oocyte could equally be proposed to localize nurse-cell-derived messages or to maintain localization of oocyte-derived messages, this observation does not address the source of grktranscripts.

Using a novel application of a standard genetic technique to address this question, we generated egg chambers with mosaic germline cysts in which the oocyte lacked the ability to produce grk but the nurse cells retained grk function. This genetic approach provided a stringent functional test for grk production: if grk were produced exclusively in the oocyte nucleus, then mosaic egg chambers with a mutant oocyte would be predicted to exhibit patterning defects. Our results demonstrate that the nurse cells produce functional grk, and that their contribution is sufficient for proper Grk localization within the oocyte and establishment of the oocyte AP and DV axes. While our data do not exclude the oocyte nucleus as a potential additional source of grk transcripts, any such contribution is not required for axis determination. Our findings imply the existence of a mechanism for transport of grk from the nurse cells and its subsequent localization within the oocyte.

For mosaic analysis of grk function, we used w;P{neoFRT}40A (Xu and Rubin,1993) and w; grk^2B6 cn bw/CyO (gift from T. Schüpbach) to generate grk^2B6 P{neoFRT}40A. The grk^2B6 null allele, which contains a 442 bp deletion that removes the transcriptional start site and part of the promoter, produces no grk RNA or protein and is the strongest existing grk allele(Neuman-Silberberg and Schüpbach,1993; Neuman-Silberberg and Schüpbach, 1994; Thio et al., 2000). Clones marked by the absence of nuclear localized green fluorescent protein (GFPnls) were generated using y w P{hsFLP}122; P{Ubi-GFPnls}2-1 P{neoFRT}40A (provided by S. Luschnig). Strains for the lacO marker system were w; P{Hsp83-GFP::lacI} and w; P{lacO^256x}(Vazquez et al., 2001). We mobilized a third chromosome insert of w; P{lacO^256x} to generate an insert on 2L, then generated w; grk^2B6P{lacO^256x}2L P{neoFRT}40A. For visualization of mirrexpression, we used the P{lacW}l(2)mirr^6D1 enhancer trap(gift from D. Morisato).

Mosaic ovaries with the nuclear GFP clone marker were produced by FLP-FRT-mediated mitotic recombination (Xu and Rubin, 1993) in females of the following two genotypes: y w P{hsFLP}122; grk^2B6 P{neoFRT}40A/P{Ubi-GFPnls}2-1 P{neoFRT}40A and y w P{hsFLP}122; grk^2B6P{neoFRT}40A/P{Ubi-GFPnls}2-1 P{neoFRT}40A;P{lacW}l(2)mirr^6D1. To induce recombination, pupae were heatshocked in a water bath for 1 hour at 37°C on each of three consecutive days beginning 2 days after puparium formation. The vials were maintained at 25°C at all other times, and adult ovaries were harvested 10 days after the first heatshock.

To generate ovaries mosaic for the lacO transgene, recombination was induced as above in females of the following genotype: y w P{hsFLP}122; grk^2B6 P{lacO^256x}2L P{neoFRT}40A/P{NM}31E P{neoFRT}40A; P{Hsp83-GFP::lacI}. In addition, females were heatshocked at 37°C for 20 minutes, 7 hours before dissection to induce expression of GFP::LacI. To confirm that the number of foci of GFP fluorescence observed in the oocyte nucleus corresponds to the number of copies of the P{lacO^256x} transgene present, we induced GFP::LacI expression in non-mosaic females either homozygous or heterozygous for a transgene-containing chromosome. Multiple fluorescent foci are readily detected in all nurse cell nuclei, which are polyploid, but individual foci in the oocyte nucleus are more difficult to detect. In females with two copies of the lacO transgene, two foci were detected in 6/49 stage-8-9 egg chambers, and the remainder exhibited either one focus (8/49) or no foci (35/49). A single focus of GFP fluorescence was detected in the oocyte in 4/23 stage-8-9 egg chambers from females with a single copy of the transgene; importantly, two foci were never observed,indicating that the presence of two foci is diagnostic for homozygosity of the transgene-containing chromosome. Because the inefficiency of detection of the lacO transgene in the oocyte nucleus prevented the unambiguous diagnosis of its absence, we placed the lacO transgene in cis to the grk^2B6 allele to allow positive identification of homozygous mutant oocytes by the presence of two foci in the nucleus (see Fig. 1C). Foci were undetectable in nearly all egg chambers after stage 9, precluding the use of this marker for analysis of DV patterning.

Ovaries were dissected in phosphate-buffered saline (PBS) and fixed at room temperature for 20 minutes in 5% formaldehyde (EM grade; Polysciences, Inc),in PBS with 1% NP40, saturated with heptane, then washed three times for 10 minutes in PBS with 0.3% Triton-X100 (PBST). Ovaries were further permeabilized by incubating at room temperature for 1 hour in PBS with 1%Triton-X100, then blocked for 1 hour in PBST with 1% bovine serum albumin(BSA). Ovaries were then incubated overnight at 4°C in PBST with a 1:100 dilution of primary monoclonal antibody [anti-Grk, MAb1D12, or anti-β-galactosidase, MAb40-1a, concentrated, Developmental Studies Hybridoma Bank; anti-Broad Core (BR-C) MAb25E9, supernatant, gift of Greg Guild], washed three times for 20 minutes in PBST, incubated for 1 hour in PBST with 1% BSA, then incubated for 90 minutes with goat anti-mouse AlexaFluor^568nm (1:1000, Molecular Probes) in PBST. Samples were washed three times for 20 minutes in PBST, then incubated for 10 minutes in PBST with rhodamine-conjugated phalloidin (1:1000, Molecular Probes) and DAPI(1:1000, Molecular Probes). After manual removal of stage-14 egg chambers,samples were mounted using the Slowfade Light Antifade Kit (Molecular Probes).

To determine whether expression of grk in the oocyte nucleus is required for axial patterning, we generated egg chambers with germlines mosaic for a null allele of grk, grk^2B6, using the FLP/FRT system to drive site-specific mitotic recombination in heterozygous females(Fig. 1A). Recombination events occurring during the division of heterozygous germline stem cells result in germline clones, in which the germline cyst consists entirely of homozygous cells, whereas recombination during subsequent germline divisions gives rise to mosaic cysts, with both wild-type and mutant cells. To address the requirement for grk transcription in the oocyte nucleus, we analyzed mosaics in which some cells were homozygous for the grk^2B6mutation and others retained grk function. If grk were produced exclusively by the oocyte nucleus, then egg chambers lacking a functional copy of the grk gene in the oocyte would be predicted to exhibit the AP and DV patterning defects characteristic of grkhomozygotes, even in the presence of wild-type nurse cells.

As a genetic marker for these mosaics, we used a transgene expressing a nuclear form of GFP. This marker was placed in trans to the grk^2B6 allele, so that homozygous mutant cells were marked by the lack of GFP (Fig. 1B). Egg chambers with homozygous mutant germline clones were readily recognized by the absence of detectable GFP fluorescence throughout the germline, while mosaic cysts exhibited a combination of nuclei with and without GFP. Though well established as a marker for clonal analysis, we anticipated that the use of nuclear GFP as a genotypic marker in mosaic cysts would be less straightforward. Due to the transport of material from the nurse cells to the developing oocyte through the cytoplasmic bridges connecting the germline cells, any GFP in the oocyte nucleus could include a contribution from the nurse cells. Moreover, it was unclear whether the GFP transgene would be expressed in the transcriptionally inactive oocyte nucleus. To circumvent these issues, we avoided assessing the genotype of the oocyte directly by restricting our analysis to egg chambers in which half the cells were homozygous wild type, as determined by uniformly high levels of GFP, and half were homozygous mutant, as determined by the lack of GFP. Given that the germline mitoses occur in an invariant pattern, mitotic recombination in the first division will always produce precisely eight cells of each genotype. Therefore in egg chambers with eight homozygous wild-type nurse cells, with uniformly high nuclear GFP, and seven grk^2B6 homozygous nurse cells, with little or no GFP, we could infer clearly that the oocyte must be the eighth mutant germline cell. In the reciprocal mosaics, with uniformly high nuclear GFP in seven nurse cells and little or none in the remaining eight, we concluded that the oocyte must be the eighth homozygous wild-type cell.

The four consecutive mitoses that generate each germline cyst result in a configuration of germline cells in which the oocyte is directly connected to four of the nurse cells: one is its sister from the first mitosis and the other three are its daughters from the subsequent rounds (see Fig. 1A). In 130/131 examples of germline mosaics with half wild-type and half mutant cells, we noted that three of these four nurse cells shared the deduced oocyte genotype, while the fourth had the opposite genotype. This nearly invariant configuration confirms our assessment of oocyte genotypes and demonstrates that these mosaics are generated primarily by recombination events occurring during the first round of germline mitoses. Multiple recombination events in subsequent mitoses probably generated the single exceptional mosaic.

In parallel, to determine the genotype of the oocyte directly, we generated germline mosaics using a transgene containing 256 direct repeats of the lac operator (lacO) as a genetic marker(Robinett et al., 1996; Straight et al., 1996; Vazquez et al., 2001). The presence of the transgene was visualized by the binding of a nuclear GFP-tagged Lac repressor protein (GFP-LacI), which binds to the lacOtransgene and yields a discrete focus of nuclear fluorescence. In this system,the integrated lacO transgene itself functions as the genotypic marker, and is therefore strictly cell-autonomous. In ovaries from females heterozygous for the lacO transgene, a single focus was present in the oocyte nucleus, whereas in the nurse cell nuclei, which are highly polyploid with partially dispersed chromatids, multiple foci were visible. For generation of germline mosaics, we constructed a chromosome containing the lacO transgene in cis to the grk^2B6allele, so that homozygous grk mutant cells would be homozygous for the transgene as well. This configuration allowed us to recognize a homozygous mutant oocyte directly by the presence of two clear foci of GFP fluorescence within the oocyte nucleus (Fig. 1C).

We examined whether production of grk in the oocyte nucleus is required for AP patterning of the egg chamber by observing the position of the oocyte nucleus. As described above, the movement of the oocyte nucleus from a central position at the oocyte posterior to an asymmetric location at the anterior cortex requires the correct polarization of the oocyte microtubule network, which in turn depends upon induction of posterior follicle cell fates by Grk signaling in early oogenesis. In the absence of grk function,this sequence of events is not initiated and the oocyte nucleus remains at the posterior of the oocyte (Gonzalez-Reyes et al., 1995; Roth et al.,1995).

In egg chambers with germline clones, in which all germline cells were homozygous for the grk^2B6 allele and therefore lacked grk function, the oocyte nucleus was located at the posterior of the oocyte in 94/154 cases observed (61%; Fig. 2A). This frequency of mislocalization is consistent with previous analyses, which report posterior localization of the oocyte nucleus in 31-70%of egg chambers from homozygous grk mutant females, depending on the allelic combination (Gonzalez-Reyes et al.,1995; Roth et al.,1995). This defect confirms the effect of the grk^2B6 mutation on grk function and provides an important internal control for the analysis of egg chambers with mosaic germlines, which are recovered from the same females. We then analyzed GFP-marked germline mosaics consisting of eight homozygous wild-type and eight homozygous mutant cells. In 21/21 control mosaics, in which the oocyte was homozygous for the wild-type grk allele, the oocyte nucleus was properly localized at the anterior margin of the oocyte(Fig. 2B). Strikingly, anterior localization of the oocyte nucleus was also observed in 22/22 mosaics in which the oocyte was homozygous for the grk^2B6 allele(Fig. 2C). These observations indicate that transcription of grk in the oocyte nucleus is not required for oocyte AP polarity.

We confirmed these observations using the lacO/GFP-LacI system as a genotypic marker. Fig. 3shows a mosaic egg chamber in which the oocyte exhibited two foci of GFP-LacI fluorescence, indicating that it was homozygous for the grk^2B6 allele (see Fig. 1C). In addition, 13 nurse cells exhibited multiple foci and two nurse cells lacked foci. We can conclude from the lack of foci in two nurse cell nuclei that these cells were homozygous for the wild-type grkallele, and that this mosaic configuration arose from a single recombination event during the third cystoblast division or two events during the final division. Although the oocyte lacked any functional copies of the grkgene, the anterior migration of the oocyte nucleus was not affected. Proper localization of the oocyte nucleus was observed in 4/4 germline mosaic egg chambers with a grk mutant oocyte detected using this system,corroborating our results with the nuclear GFP marker.

Taken together, our analysis of germline mosaics indicates that synthesis of grk in the oocyte is not required for its function in determination of the polarity of the oocyte AP axis. We therefore conclude that in wild-type egg chambers grk transcription is not restricted to the oocyte nucleus and that production of grk in the nurse cells is sufficient for proper grk function in this process.

Coincident with the movement of the oocyte nucleus to the anterior cortex of the oocyte, grk mRNA and protein also relocalize, remaining closely associated with the oocyte nucleus and achieving an asymmetric anterior localization that defines the dorsal side of the oocyte and thus the DV axis of the egg chamber. While the data presented above indicate that the production of grk by the nurse cells is sufficient for AP patterning,this analysis does not address the possibility that, at later stages of oogenesis, transcription of grk in the oocyte nucleus is required for DV patterning. Indeed, it has been speculated that grk may be produced in all germline cells during early stages, for grk-mediated AP patterning, then produced primarily or exclusively in the oocyte nucleus in later stages, leading to the dorsally restricted distribution required for DV patterning (MacDougall et al.,2003; Norvell et al.,1999; Thio et al.,2000).

To investigate this possibility, we assessed dorsal follicle cell fate determination in grk mosaics. As a cell fate marker we used an enhancer trap inserted in the mirror locus, mirr-lacZ, which drives expression of a lacZ reporter gene in dorsal anterior follicle cells in response to Grk-Egfr signaling(Fig. 4A)(Jordan et al., 2000; Zhao et al., 2000). Egg chambers with germline clones homozygous for the grk^2B6mutation exhibited no follicle cell expression of mirr-lacZ,confirming that expression of this reporter is grk-dependent(Fig. 4B). By contrast, in germline mosaic egg chambers derived from the same females, expression of the mirr-lacZ marker was detected in dorsal anterior follicle cells, even when the oocyte contained no functional copies of the grk gene(Fig. 4C,D). We found no significant difference in the number of lacZ-positive follicle cells between stage-10 germline mosaics with a grk mutant oocyte(n=12) and germline mosaics with either a wild-type oocyte(n=7) or grk^2B6 heterozygous egg chambers(n=25), from which germline mosaics are derived. While these data reveal no gross alterations in DV patterning in the absence of a functional grk gene in the oocyte nucleus, because the mirr-lacZexpression pattern is dynamic and the number of positive cells varies even among stage-10 egg chambers from grk heterozygotes, any subtle changes in expression within this range would be difficult to detect.

As an additional marker of DV patterning, we analyzed the expression pattern of the Broad Complex (BR-C) protein. Although BR-C is initially expressed throughout the follicular epithelium, this expression pattern is refined during stage 10, in response to Grk-Egfr signaling, resulting at the end of this stage in two dorsolateral groups of follicle cells with high levels of BR-C flanking a dorsal midline region where BR-C levels are eliminated (Deng and Bownes,1997; Suzanne et al.,2001; Tzolovsky et al.,1999). This pattern of BR-C expression was also observed in germline mosaic egg chambers with a homozygous grk^2B6oocyte (Fig. 5), further indicating that synthesis of grk in the oocyte nucleus is not required to determine the DV axis of the egg chamber. To examine the BR-C pattern more closely, we counted the number of cells along the AP and DV dimensions of the dorsal midline domain. On average, this domain was approximately one cell shorter along both axes in germline mosaics with a mutant oocyte than in those with a wild-type oocyte (see Fig. 5). However, the ranges of these dimensions overlapped, and the dimensions of this domain for all mosaics fell within the range observed in non-mosaic heterozygous egg chambers at the same stage. Therefore, while the difference in BR-C expression between mosaics with a wild-type and mutant oocyte could represent a minor or late contribution of grk by the oocyte nucleus, the variability at this stage probably reflects the dynamic nature of the BR-C expression pattern and limits by definition the resolution of this analysis, precluding a definitive conclusion regarding possible subtle patterning differences.

Taken together, these results demonstrate that the nurse cells contribute sufficient grk to establish the DV axis of the egg chamber. While our data do not exclude the possibility that the oocyte also produces grk, these findings clearly indicate that DV patterning is largely normal even in the absence of a functional grk allele in the oocyte nucleus.

Grk-Egfr signaling is also required to guide the migration of a specialized subpopulation of follicle cells called border cells. This cluster of cells delaminates from the anterior follicular epithelium at stage 9 of oogenesis and migrates posteriorly, between the nurse cells, toward the oocyte. Upon reaching the anterior margin of the oocyte, the border cells migrate dorsally and assume a position adjacent to the oocyte nucleus(Spradling, 1993). Loss of Grk function in the germline or Egfr function in the border cells themselves leads to the failure of the dorsal phase of this migration, indicating that Grk provides a spatial cue that acts through Egfr to guide border cell migration(Duchek and Rorth, 2001).

To determine whether synthesis of grk in the oocyte nucleus is required for border cell guidance, we examined the position of the border cell cluster in germline mosaic egg chambers of the appropriate stage. Border cells were dorsally localized along the anterior margin of the oocyte in 12/14 germline mosaics with a wild-type oocyte, comparable to previous observations of wild-type egg chambers (Duchek and Rorth, 2001). In germline mosaics with a grk mutant oocyte, the border cell cluster was dorsally localized in 14/15 cases. These data indicate that synthesis of functional grk transcripts in the oocyte nucleus is not required to guide border cell migration.

The observation that AP and DV patterning occur normally in germline mosaics with a homozygous grk mutant oocyte suggests that the Grk signal is properly localized in these egg chambers. To test this prediction,we visualized the localization of the Grk protein in mosaic egg chambers using a monoclonal antibody. No specific immunoreactivity was observed in 14/14 homozygous mutant germline clones, confirming the specificity of the antibody(data not shown). In germline mosaics, the localization of Grk in those with a homozygous mutant oocyte (Fig. 6B,C) was indistinguishable from that of those with a wild-type oocyte (Fig. 6A). This observation indicates that normal Grk localization is achieved even when the transcript is derived exclusively from the nurse cells and indicates that production of wild-type grk transcripts in the oocyte is not required to generate a spatially restricted Grk signal.

Our analysis of genetic mosaics demonstrates that the presence of a functional copy of the grk gene in the oocyte nucleus is not required for grk function, indicating that the nurse cells provide grk to the oocyte. Germline mosaics are generated through mitotic recombination in heterozygous cells during the formation of the 16-cell germline cyst; each homozygous mutant cell is therefore ultimately derived from a heterozygous cystoblast. However, it is unlikely that sufficient grk from the cystoblast would persist through these mitoses, and the substantial subsequent increase in cell size, to account for the normal patterning observed in germline mosaics with a grk mutant oocyte. First, no grk mRNA or protein is present during the germline mitoses that generate each cyst, becoming detectable only later, as the complete 16-cell cyst becomes enveloped by the follicular epithelium(Neuman-Silberberg and Schüpbach,1993; Neuman-Silberberg and Schüpbach, 1996). Moreover, any sub-detectable levels of grk present earlier would not be predicted to be sufficient for patterning, as grk is weakly haploinsufficient: even grkheterozygotes, where grk is readily detectable, display some egg chamber patterning defects(Neuman-Silberberg and Schüpbach,1994). Therefore our data indicate that transcription of the grk gene occurs in the nurse cells and that this contribution can generate a properly localized Grk signal that is sufficient to mediate axial patterning of the oocyte.

Evidence for transcription of grk in the oocyte nucleus has been reported previously (Saunders and Cohen,1999). After treatment of females with colchicine to disrupt microtubule-based transport from the nurse cells to the oocyte, osktranscripts were retained in the nurse cells while grk transcripts were detected exclusively in the oocyte, consistent with grktranscription in the oocyte nucleus. However, it is also possible that there is a fundamental difference between grk and osk in the timing or mode of transport. For example, blocking transport of grkto the oocyte could require a level of microtubule disruption equivalent to that which would disrupt oocyte specification(Koch and Spitzer, 1983),potentially confounding analysis by this method. Alternatively, transport may occur via a microtubule-independent mechanism, or along microtubules that are stable and therefore insensitive to disruption by treatment with colchicine,which affects only dynamic microtubules. Such a population of stable microtubules, present at least in the early stages of oogenesis, has recently been described (Roper and Brown,2004).

In a complementary approach, transcripts from a reporter construct under the control of the grk promoter were observed to accumulate exclusively in the oocyte, consistent with activity of this promoter exclusively in the oocyte nucleus(Saunders and Cohen, 1999). However, subsequent mapping of grk transcripts suggested that the grk transcription start site is located farther upstream than previously recognized (Thio et al.,2000), indicating this construct probably contained elements present within the grk 5′ UTR. While there is no evidence that this sequence is relevant to mRNA transport or localization, strong conclusions about grk promoter activity cannot be drawn.

Our genetic approach demonstrates that the nurse cells provide grkto the oocyte. However, because it is difficult to detect subtle or late changes in patterning, our data do not exclude a contribution of grkby the oocyte nucleus. Given the lack of obvious defects in mosaics with a grk mutant oocyte, and the greater synthetic capacity of the polyploid nurse cells, it seems likely that any grk contribution by the oocyte nucleus would be minor. Nevertheless, the important conclusion of this work is the clear demonstration that the nurse cells transcribe grk, and that these transcripts are sufficient for normal Grk localization and the establishment of the oocyte AP and DV axes. These observations imply the existence of a mechanism for their transport from the nurse cells and subsequent dorsal anterior localization within the oocyte.

We have been unable to visualize the distribution of grk mRNA in germline mosaics, due to the incompatibility of GFP fluorescence with conditions required for in-situ hybridization and the inconsistent levels of signal obtained with available anti-GFP antibodies. However, it seems likely that the normal distribution of Grk in mosaics with a mutant oocyte reflects proper localization of grk mRNA. The alternative possibility, that grk mRNA is mislocalized in mosaics with a grk mutant oocyte but yields a properly localized protein, would imply that synthesis of wild-type grk transcripts in the oocyte nucleus is required to localize nurse-cell-derived transcripts. Although there is evidence for translational regulation of unlocalized grk, which would account for the normal Grk distribution observed in such mosaics(Norvell et al., 1999), it is unclear how localization of grk transcripts from the nurse cells would depend on additional grk production in the oocyte nucleus. Although we cannot exclude such a model, due to its complexity it seems less likely.

The site of grk transcription has important implications for the consideration of potential localization mechanisms. While the localization of the grk mRNA between the oocyte nucleus and adjacent cortex has led to the proposal that this distribution arises from grk transcription in the oocyte nucleus and local anchoring of grk transcripts(Goodrich et al., 2004; Norvell et al., 1999; Palacios and St Johnston,2001; Saunders and Cohen,1999), such a model would not address the proper localization and patterning function of grk contributed by the nurse cells. As grk encodes a secreted protein, it has also been suggested that a dorsal anterior concentration of exocytic pathway components within the oocyte could contribute to its localization. A careful analysis of transitional endoplasmic reticulum and Golgi compartments, however, reveals a uniform distribution and indicates that polarized Grk distribution is driven by the localization of its mRNA (Herpers and Rabouille, 2004).

The mechanism of transport of grk transcripts from the nurse cells to the oocyte is unknown. However, within the oocyte, proper dorsal anterior localization of grk mRNA requires the heterogeneous nuclear ribonuclear protein (hnRNP) proteins Sqd (also known as Hrp40) and Hrb27C(also known as Hrp48), which bind to the 3′ UTR of the grk mRNA(Goodrich et al., 2004; Neuman-Silberberg and Schüpbach,1993; Norvell et al.,1999). These hnRNPs form a complex with the nascent grktranscript, then recruit cytoplasmic proteins to the grk RNP complex upon its export from the nucleus to regulate grk localization and translation in the cytoplasm (Goodrich et al., 2004; Norvell et al.,1999). Although this model proposes that the interaction of Sqd and Hrb27C with the grk transcript occurs in the oocyte nucleus(Goodrich et al., 2004; Norvell et al., 1999), it seems plausible to suggest that these complexes assemble in the nurse cells to regulate grk localization within the oocyte. Localization of grk mRNA within the oocyte also requires transport on microtubules,because disruption of the minus-end-directed microtubule motor cytoplasmic Dynein results in defects in grk localization(Brendza et al., 2002; Duncan and Warrior, 2002; Januschke et al., 2002; MacDougall et al., 2003). It is unclear, however, whether these factors are required for transport of grk into the oocyte, as the most obvious consequence of loss of their function is mislocalization of grk mRNA to the oocyte anterior margin.

Taken together with previous work, our data favor a model in which grk transcripts are assembled in the nurse cell nuclei into hnRNP particles containing Sqd and Hrb27C, followed by recruitment upon nuclear export of cytoplasmic factors regulating localization and translation. These grk-containing complexes would ultimately associate with microtubule motors, resulting in minus-end-directed transport along microtubules emanating from the dorsal anterior region; indeed there is evidence for a scaffold of microtubules around the oocyte nucleus(Clark et al., 1997; MacDougall et al., 2003; Theurkauf et al., 1992). Remaining to be resolved, however, is a mechanism that would distinguish microtubule-based localization of grk from that of anteriorly localized messages that are not dorsally restricted. Although grkcould interact with specific trans-acting factors in a distinct hnRNP particle, a potential sorting mechanism would nevertheless require differences in the dorsally oriented microtubules to allow recognition by motors carrying grk-containing particles. While modifications of tubulin itself or the association with distinct microtubule-associating proteins could distinguish microtubule networks(Westermann and Weber, 2003),the mechanism underlying the sorting of grk from other anterior transcripts remains to be determined.

We thank Stefan Luschnig, Donald Morisato, Trudi Schüpbach and Julio Vazquez for fly strains, Greg Guild for the anti-BR-C antibody, Cheryl Van Buskirk and K. Nicole Clouse for advice on immunohistochemistry, and Jennifer Thompson for technical assistance. We are also grateful to Trudi Schüpbach, Paul Lasko and David Dansereau for helpful comments on the manuscript. The monoclonal antibodies 1D12 and 40-1a, developed by Trudi Schüpbach and Joshua Sanes, respectively, were obtained from the Developmental Studies Hybridoma Bank developed under the auspices of the NICHD and maintained by The University of Iowa, Department of Biological Sciences,Iowa City, IA. L.C. and L.N. were supported by the Canada Research Chairs Program(www.chairs.gc.ca).