The gradient of Gurken, a long-range morphogen, is directly regulated by Cbl-mediated endocytosis

Morphogens play pivotal roles during many important developmental stages to induce distinct cell fates by forming concentration gradients that provide positional information. The gradient formation is usually controlled by a restricted source, a limited half life and various mechanisms of trafficking through tissues (Entchev and Gonzalez-Gaitan, 2002; Tabata and Takei, 2004; Zhu and Scott, 2004). Cell surface molecules, such as morphogen receptors and heparan sulfate proteoglycans (HSPG)(Hacker et al., 2005), shape morphogen gradients by influencing their spreading. To understand the generation of positional information, it is important to assess how far a morphogen travels and how the morphogen gradient is maintained.

Gurken, a transforming growth factor α (TGFα) homolog, is a member of the Epidermal growth factor receptor (Egfr) ligand family. In Drosophila, this family includes a soluble secreted protein, Vein(Schnepp et al., 1996), and three transmembrane proteins, Keren (Krn), Spitz (Spi) and Gurken (Grk)(Neuman-Silberberg and Schupbach,1993; Reich and Shilo,2002; Schweitzer et al.,1995). The activation of Spi requires a proteolytical process involving Star and Rhomboid 1 (Rho). Star promotes the translocation of Spi from the endoplasmic reticulum to the Golgi apparatus where Rho, an intramembrane serine protease, cleaves Spi(Lee et al., 2001; Urban et al., 2001). Similarly, the transmembrane domain of Gurken is essential for its activity in the oocyte (Queenan et al.,1999), and Star and brother of rhomboid(brho; stet-FlyBase) are required for a cleavage process of Gurken to produce the secreted soluble active form of Gurken(Ghiglione et al., 2002).

Egfr signaling activated by Gurken during oogenesis is required for axes determination in the future embryo (Roth,2003). Gurken exhibits a striking asymmetrical localization at both RNA and protein level during middle stages of Drosophilaoogenesis. The 5′ and 3′ untranslated regions, as well as part of the coding region of gurken mRNA target its localization tightly to the future dorsal anterior corner (Thio et al., 2000; Van De Bor et al.,2005) and by efficient sorting through the transitional endoplasmic reticulum (tER)-Golgi units, Gurken protein is also directionally transported to this dorsal anterior region(Herpers and Rabouille, 2004). The spatial restriction of Gurken activates the Egfr in a restricted number of follicle cells, leading to the establishment of dorsal follicle cell fates(Neuman-Silberberg and Schupbach,1996). The further patterning of the dorsal egg shell structure is carried out by positive and negative feedback pathways of Egfr signaling(Freeman, 1998). Gurken activates the Egfr pathway to a maximum in the dorsal midline region where Rho expression is induced. Rho facilitates the activation of Spi that diffuses and activates Egfr in a border dorsal region. The active form of Spi is sequestered by the negative regulator Argos, which is expressed in the dorsal midline (Golembo et al., 1996; Klein et al., 2004). In addition, the negative regulator Pointed is also induced by high Egfr signaling at the dorsal midline (Deng and Bownes, 1997). As a consequence, the domain of high levels of Egfr signaling is split into two regions that promote dorsal appendage formation.

Whether Gurken travels to the ventral side of the egg chamber to influence the ventral follicle cell fate has never been completely resolved. Our previous examination of the expression of the Egfr target gene pipe,which is transcriptionally repressed by Egfr signaling, indicates that some level of Egfr activity is suppressed by Cbl in ventral follicle cells(Pai et al., 2000). In the absence of Cbl, pipe is repressed on the ventral side and this repression requires Gurken (Pai et al.,2000). Other studies suggested that Gurken indirectly regulates pipe expression through another diffusible morphogen induced by Notch signaling, which would be activated by interaction between mirror(mirr) and fringe (fng)(Jordan et al., 2000). However, further analyses of mirr and fng mutant follicle cell clones indicated that these two genes have no effects on pipeexpression. Moreover, the role of Spi, which is activated by rho, in regulating pipe expression has also been ruled out by examination of large rho mutant clones (Peri et al., 2002). These data suggest that the Pipe domain is directly defined by a long-range Gurken gradient. Based on indirect estimation through measuring the repression of pipe and mathematical modeling, the Gurken protein level at the ventral side of the egg chamber was estimated to be 10% of that at the dorsal side(Goentoro et al., 2006). However, Gurken was never directly detected at the ventral side of wild-type egg chambers, possibly owing to low levels and rapid turnover. We reasoned that if Gurken directly activates the Egfr in ventral follicle cells then,using better imaging approaches, Gurken should be detected in these follicle cells in which Cbl is required for Egfr signal attenuation.

In mammals, the Egfr is dimerized and autophosphorylated upon ligand activation, followed by recruitment of various proteins recognizing phospho-tyrosine to transduce or terminate Egfr signaling(Schlessinger, 2002). Cbl family proteins interact with the C-termini of activated Egfr molecules either indirectly through the adaptor protein Grb2 or directly through the tyrosine kinase binding (TKB) domain of Cbl(Schmidt and Dikic, 2005). The E3 ligase activity of Cbl promotes the ubiquitylation of activated Egfr while they are associated, and the Egfr-Cbl complex is then internalized by multiple routes (Schmidt and Dikic,2005). The ubiquitylated receptor binds to Epsin family proteins,and goes through a non-clathrin-dependent, lipid-raft-dependent route(Sigismund et al., 2005). In addition, Cbl serves as an adaptor for interaction with the CIN85/endophilin complex, leading to dynamin-dependent clathrin-mediated receptor endocytosis(Dikic, 2003; Petrelli et al., 2002; Soubeyran et al., 2002). Previously, we have demonstrated that both Cbl isoforms (CblS and CblL)downregulate Egfr signaling, and that the efficiency of Egfr signaling attenuation is controlled by the levels of the CblL isoform(Pai et al., 2006). The genetic interaction with shibire, a homologue of dynamin(Chen et al., 1991; van der Bliek and Meyerowitz,1991), further suggested that CblL may promote the internalization of activated Egfr to reduce signaling (Pai et al., 2006).

Here, we report the use of a HRP (horseradish peroxidase)-Grk fusion protein that allows us to directly trace the fate of Gurken protein in follicle cells during signaling. We found that the HRP-Grk fusion protein is internalized with the Egfr into follicle cells, a process that is mediated by Shibire. The Grk-Egfr complex trafficks through Rab5/7-associated endocytic pathways to the MVBs (multivesicular bodies) and the lysosome for signal termination. Gurken was visualized not only in the dorsal but also in ventral follicle cells. We document that Cbl facilitates the internalization of the Grk-Egfr complex into follicle cells using clonal analysis of loss of function mutations, as well as over-expression studies. Our results indicate that the distribution of the Gurken morphogen is controlled by the endocytosis of the Grk-Egfr complex through Cbl.

The strains used include: wild-type (Oregon-R, OreR); grk^HF (Schupbach,1987); top^IP02 and top^CO(Clifford and Schupbach, 1994); UAS-shi^TS1 (Kitamoto,2001); UAS-GFP-Rab5(Wucherpfennig et al., 2003); UAS-GFP-Rab7 (Entchev et al.,2000); UAS-nls-GFP(Shiga et al., 1996); EQ1 and GR1 Gal4 lines(Queenan et al., 1997); Cbl^F165 (Pai et al.,2000); UAS CblS-A1; UAS CblL-A18; and A12 (Pai et al.,2006). e22c-Gal4(Duffy et al., 1998) or GR1-Gal4 was used to drive expression of UAS-Flipase in follicle cells.

To generate HRP-grk, DNA encoding HRP(Connolly et al., 1994) was inserted into the gurken genomic fragment, which is a 5 kb EcoRV fragment derived from the original rescuing fragment(Neuman-Silberberg and Schupbach,1993) as shown in Fig. 2 of the publication by Thio et al.(Thio et al., 2000). This 5 kb fragment was subcloned into pBSKS as two separate plasmids, A5 (5′ EcoRV site at the beginning to EcoRV site in the third intron) and A6 (EcoRV site in the third intron to the Eagsite at the end). The Grk fragment in A5 was first cloned into pcDNA using the EcoRV site, and then DNA encoding HRP was inserted into pcDNA grk A5 using standard PCR-based cloning at the SacII site. This HRP-Grk A5 fragment was cloned into A6 at the EcoRV site. DNA encoding the HRP-Grk fusion protein was then cloned into pCaSpeR4 with the ApaI-EagI fragment and the construct was introduced in w¹¹¹⁸ hosts by P element-mediated transformation using standard methods.

The antibodies used for immunocytochemistry are the monoclonal anti-Gurken antibody 1D12 made in mouse (Queenan et al., 1999); the goat anti-HRP antibody (Jackson ImmunoResearch,1:250 dilution); the anti-CblL 8C4 ascites at 1:100 dilution(Pai et al., 2006); the anti-c-myc antibody 9E10 (Oncogene, at 1:50 dilution); a rabbit anti-Myc antibody (Santa Cruz Biotechnology); the rabbit anti-Egfr antibody, which was prepared according to the previously reported method(Jekely and Rorth, 2003) and used at 1:50; the Alexa Fluor 488 or 546 goat anti-mouse antibody (Molecular Probes); the Alexa Fluor 488 or 546 rabbit anti-goat antibody (Molecular Probes); and the TRITC-conjugated donkey anti-mouse antibody (Jackson ImmunoResearch). The antibodies used for immunoelectron microscopy were goat anti-HRP (Sigma; 1:1000 dilution) and rabbit anti-goat IgG (DakoCytomation Denmark, Glostrup, Denmark). The antibodies were detected with protein A-gold conjugated with 10 nm gold (Cell Microscopy Center, Utrecht, The Netherlands).

To induce Myc expression, e22cFLP/grk^HFHRP-grk;Cbl^F165FRT^80B/ πMFRT^80Badult females were subjected to heat shock for 1 hour at 37°C before dissection (Xu and Rubin,1993). Ovaries were dissected in PBS and fixed for 20 minutes in 200 μl 4% paraformaldehyde (PF) saturated with 600 μl heptane and 0.25%NP40. Staining procedures were performed as described(Neuman-Silberberg and Schupbach,1996) with minor modifications. To improve the penetration,ovaries were incubated with 1% TritonX-100 in PBST with 1% BSA before blocking, in which 10% normal rabbit serum or 1% BSA was used to reduce the background. To visualize actin, ovaries were incubated with 1 unit of phalloidin (Molecular Probes) for 30 minutes. For staining DNA, 0.5 μg/ml DAPI (Sigma) was added to the ovaries for 5 minutes. Egg chambers were further dissected and mounted in glycerol for examination under fluorescence confocal microscopy. The RNA hybridization procedure was carried out as described previously (Tautz and Pfeifle,1989). A digoxigenin (Dig)-labeled RNA probe was made using the DIG RNA labeling Kit (Roche). The probe was detected with alkaline phosphatase conjugated anti-DIG antibody (1:5000 dilution, Roche) and the NBP/BCIP developing kit (Promega).

Ovaries were dissected and fixed in a mixture of 0.2% glutaraldehyde and 2%paraformaldehyde in 0.1 M phosphate buffer for 3-4 hours. The individual egg chambers were embedded in 12% gelatin. The blocks were trimmed to size,immersed in PVP/sucrose (15%/1.7M) at 4°C overnight, mounted on aluminum pins in the proper orientation and stored in liquid nitrogen. Frozen sections were prepared by using an ultracryomicrotome (Reichert Ultracut S/Reichert FCS, Leica; Vienna, Austria). To obtain ultrathin sections that included the oocyte nucleus, 0.5 μm semithin sections were first cut, at-60°C, and examined with a light microscope after staining with 0.1% Toluidine Blue in 1%boric acid. When the nucleus was reached, the cutting temperature was lowered to-100°C and section thickness switched to 55 nm. Ultrathin sections were picked up with a drop of an equal volume mixture of methylcellulose(2%)/sucrose (2.3 M). The ultracryomicrotomy and protocol used for immunolabeling has been described in detail(Liou et al., 1996). Samples were viewed under a JOEL JEM-1230 transmission electron microscope. To compare the endocytic activity of follicle cells at different zones, the volume density of MVB/lysosomes per follicle cell profile was estimated by the point-counting method (Griffiths,1993). Only those cell profiles exhibiting MVB/lysosomes were graphed. Sampling was from 12 different grids (nine OreR and three grk^HF/grk^HF) from seven egg chambers (five OreR and two grk^HF/grk^HF). Cell profiles analyzed (pn) are indicated in the figure. Total areas quantified were 3227.23 μm² for OreR and 693.39 μm²for grk^HF/grk^HF.

The distribution of Gurken is dynamic in the oocyte during oogenesis. At stage 7, when the Gurken signal is required to induce a posterior cell fate in the adjacent follicle cells through Egfr activation, Gurken is mainly localized in the entire cytoplasm of the oocyte. A punctate distribution of Gurken in the apical side of posterior follicle cells can be easily detected with high magnification, as shown in Fig. 1A,B and previous reports(Ghiglione et al., 2002; Queenan et al., 1999). These punctate signals in the posterior follicle cells are very likely to be the Grk-Egfr complex in endocytic vesicles after the complex is internalized.

To test this possibility, we examined the effect of a dominant-negative form of Shibire [UAS-shi^ts(Kitamoto, 2001)] on the distribution of Gurken. Under a non-permissive temperature, these punctate signals in follicle cells were dramatically reduced(Fig. 1C,D), suggesting that Gurken accumulates in posterior follicle cells through a Dynamin-dependent process. The localization of Gurken in endocytic vesicles was further confirmed by the colocalization with a GR1-Gal4 driven expression of the early endosomal marker (GFP-Rab5)(Wucherpfennig et al., 2003)and a late endosomal marker (GFP-Rab7)(Entchev et al., 2000),respectively (Fig. 1E-H′). These data indicate that Gurken is internalized into posterior follicle cells through the Shibire and Rab5/7-associated endocytic pathway at the stage when the posterior follicle cell fate is defined.

After stage 8, the oocyte nucleus moves to the anterior of the egg chamber,and Gurken is localized around the oocyte nucleus. During this stage, the Gurken signal in follicle cells was only detected in 17% (n=24) of egg chambers, and the expression was restricted to a few follicle cells adjacent to the nucleus of oocyte (Fig. 1I-L). The punctate signals within individual follicle cells possibly correspond to the endocytic compartments, similar to that which was observed in the posterior follicle cells.

Previous studies have successfully used HRP-Boss(Sunio et al., 1999) and HRP-Wingless fusion proteins (Dubois et al., 2001) to study ligand distribution in endocytic trafficking. To visualize the endocytic trafficking of Gurken during stages 8 to 10, we made an HRP-Grk fusion to use as a tracer because HRP is relatively stable within the destructive environment of the endosome and lysosomal compartments. As 5′ and 3′ UTR, as well as part of the coding region of gurken, are important for the correct localization of gurkenRNA in Drosophila oogenesis (Thio et al., 2000; Van De Bor et al., 2005), HRP was inserted into exon 3 of the gurkengenomic sequence, between the signal peptide and the EGF repeat(Fig. 2A). We expected that this fusion protein, as driven by the endogenous gurken promoter,could mimic the endogenous expression and trafficking of the Gurken protein,and would still be detected after processing through endocytosis.

Several genetic and biological assays were performed to test whether the HRP-Grk fusion protein was correctly localized and could activate the Egfr pathway. HRP-Grk was homogenously expressed in the oocyte during early stages(Fig. 2C), and localized correctly to the dorsal-anterior corner of the oocyte during mid-stages(Fig. 2D,E). Similar to endogenous Gurken, HRP-Grk was not detected after stage 10B (data not shown). Therefore, the HRP insertion in the fusion protein did not alter the expression and localization of the Gurken protein. Furthermore, we tested the biological function of HRP-Grk in rescuing the phenotypes of grkmutants. The heterozygous null mutant females of grk^HFlaid mildly ventralized eggs (about 54%) and homozygous mutant females laid 100% eggs with strong ventralized egg shells. These phenotypes could be partially rescued by the HRP-grk transgene in seven independently derived lines, with the line HRP-grk-5 giving the highest rescuing ability as HRP-grk-5B almost completely reverted the heterozygous mutant phenotype to that of the wild-type (99%)(Table 1). In addition, the rescue ability showed dose dependence, two copies of this transgene rescued to a higher degree than one copy when tested in the homozygous grk^HF mutant. These results demonstrate that the HRP-Grk fusion protein can activate Egfr signaling similarly to endogenous Gurken, and that the expression level is not saturated.

As expected, the HRP-Grk provided higher detection sensitivity than did Gurken. The double staining with anti-Grk and anti-HRP antibodies on HRP-Grk expressing egg chamber showed that nearly all Grk signal in the oocyte or follicle cells overlapped with HRP signal, but some HRP signal in follicle cells was Grk negative (see Fig. S1 in the supplementary material). Furthermore, in ventral follicle cells, no signal was detected by the anti-Grk antibody, whereas punctate HRP signals were obvious. The data indicate that the difference between Gurken and HRP-Grk is in detection sensitivity rather than distribution. The frequencies of detecting HRP-Grk in follicle cells using an anti-HRP antibody during stages 8 to 9 were 88.6% (n=53 egg chambers) at the dorsal side and 69% (n=55) at the ventral side(Fig. 2F-I). As seen in the longitudinal section of confocal images of stage 8 to 9 egg chambers, HRP-Grk was not only detected in six to ten (the 1st to 10th main body follicle cell counting from anterior to posterior) follicle cells overlying the oocyte nucleus (Fig. 2F,G), but also in four to six follicle cells at the opposite ventral side(Fig. 2H,I). This was the first time that Gurken was detected in ventral follicle cells at a stage when Gurken is asymmetrically localized at the anterior corner of the oocyte. This is consistent with theoretical and genetic evidences provided previously(Goentoro et al., 2006; Pai et al., 2000; Peri et al., 2002), which indicated that some level of Egfr activity in ventral follicle cells is activated by Gurken.

To test whether the internalization of HRP-Grk in dorsal and ventral follicle cells is Dynamin dependent, HRP signal was examined in the temperature-sensitive and dominant-negative shibire mutant UAS-shi^ts. The punctate HRP signals were dramatically reduced both in dorsal and ventral follicle cells during stage eight to nine at non-permissive temperature (Fig. 2J-M), suggesting that Dynamin has a crucial role in the internalization process of Gurken. To follow the trafficking route of HRP-Grk,we marked early endosomes with UAS-GFP-Rab5 driven by GR1-Gal4, expressed in all follicle cells, starting from the germarium. Some HRP-Grk signals were colocalized with Rab5 both in dorsal and ventral follicle cells (59%, n=44, Fig. 2N-Q). Similarly,colocalization of HRP-Grk and GFP-Rab7 (36%, n=33), a late endosomal marker, was also detected (data not shown).

We then asked whether the internalization of Gurken is dependent on its interaction with Egfr. We generated an anti-Egfr antibody to monitor colocalization of HRP-Grk and Egfr. The specificity of this antibody was tested in mosaic egg chambers with Egfr mutant clones(Fig. 3A-C). Indeed, HRP-Grk and Egfr were colocalized (62%, n=39) in dorsal(Fig. 3D-F) and ventral (data not shown) follicle cells as revealed by double staining with anti-HRP and anti-Egfr antibodies. We then generated Egfr mutant clones in the follicular epithelium to test whether HRP signal in follicle cells is dependent on the presence of Egfr, and hence corresponds to the HRP-Grk-Egfr complex internalized into signal receiving cells. Two torpedo(top)/Egfr null mutant alleles, Egfr[CO] and Egfr[IP02],were tested. The molecular information of Egfr[IP02] suggests that this mutated Egfr has no cytoplasmic domain for signaling(Clifford and Schupbach, 1994). In the Egfr mutant follicle cells, marked by the absence of GFP signal, the HRP signal was significantly reduced or entirely absent(Fig. 3G-I). The fact that the HRP signal in follicle cells is Egfr-dependent suggests that Gurken binding triggers the entry of the ligand-receptor complex into follicle cells. We also found that, once inside the cells, Egfr localizes to Rab5 positive vesicles near the apical region and to Rab7-positive vesicles in the region between the apical membrane and the nucleus, respectively(Fig. 3J-O). Given that internalized Gurken in both dorsal and ventral follicle cells depends on a functional Egfr, these data indicate that Egfr is activated by Gurken not only in dorsal but also in ventral follicle cells. The results also show that the ligand-receptor complex is mainly internalized through the Dynamin- and Rab5/7-associated endocytic pathway at the stage when Gurken defines the dorsoventral cell fate.

To visualize the subcellular localization of HRP-Grk, we applied immuno-electron microscopy using an anti-HRP antibody. Herpers and Rabouille(Herpers and Rabouille, 2004)studied the exocytic pathway of Gurken in stage 9 Drosophila oocytes by immunoelectron microscopy using the Gurken antibody, and demonstrated that the Gurken protein is directionally transported through the tER-Golgi units at the dorsal-anterior corner of the oocyte. We first tested whether HRP-Grk trafficks through the same exocytic pathway. Consistent with their observation, we found that HRP-Grk positive tER-Golgi units were significantly concentrated in the D/A corner of the ooplasm (7.7∼16.9 gold/units),whereas the labeling density in tER-Golgi units located in the ventral or posterior of the ooplasm was much weaker (0.3∼0.8 gold/units)(Fig. 4C-E and data not show). Gurken protein gradually reached the highest peak at middle stage 9 in the tER-Golgi units at the dorsal anterior corner, and declined at late stage 9(Fig. 4E). This expression pattern corresponds well with the pattern detected by immunostaining with the Gurken antibody, in which no Gurken signal can be detected after stage 10B(Neuman-Silberberg and Schupbach,1996). We also checked the labeling in the space between the oocyte and the overlying follicle cells to examine the extracellular gradients of Gurken. However, a strong labeling even in wild-type egg chambers not expressing HRP-Grk was detected, which might result from trapped gold particles within this region (data not shown).

We then examined the fate of the Grk-Egfr complex in the follicle cells at stage 9. Gold particles in the follicle cells corresponding to Gurken in our cryosectioned fixed stage 9 egg chambers showed a distinct pattern. Except for some nonspecific binding to mitochondria, major gold particles were located in unidentified small vesicles, the MVB, and the lysosome in both dorsal and ventral follicle cells (Fig. 4F,G). We previously demonstrated that the internalization of the Grk-Egfr complex plays a major role in terminating Egfr signaling, as dorsalized egg shell phenotypes resulting from elevated Egfr signaling were observed when shibire activity was blocked(Pai et al., 2006). Taken together, our data indicate that the internalized Gurken travels through the early and late endosomes, and is targeted to degradation in the MVB/lysosome compartment. Consistent with our immunostaining result, this signaling attenuation process occurs both in the dorsal and ventral follicle cells,again indicating that Gurken has reached the ventral side of the egg chamber to activate the Egfr at ventral follicle cell plasma membrane. Interestingly,we noted that after early stage 9, the lateral follicle cells contained a larger lysosomal area than did the dorsal/ventral follicle cells(Fig. 4H-J). This difference was not observed in a gurken mutant background, indicating an early response to Egfr signaling in these cells(Fig. 4K).

Consistent with the role of Cbl as a negative regulator in Egfr signaling, overexpression of CblL results in phenotypes that are characteristic of reduction of Egfr signaling. Furthermore, overexpression of CblL suppressed the effect of reduced internalization of Egfr, suggesting that CblL may promote internalization of activated receptors(Pai et al., 2006). To further investigate the precise mechanisms by which Cbl regulates Egfr signaling, we examined the pattern of Gurken distribution in egg chambers with overexpression of CblL driven by GR1-Gal4. When analyzing Gurken by immunofluorescence, it is not possible to distinguish between Gurken protein that is still at the oocyte membrane versus Gurken protein that is extracellular or bound to the follicle cell surface. In egg chambers with over-expressed CblL, at this level of resolution, we found that Gurken was markedly reduced in late stage 9 egg chambers when Gurken should have reached its highest expression peak. This reduction was seen in more than 40%(n=190) of egg chambers with CblL overexpression(Fig. 5C,D). Some variation of Gurken immunostaining was seen in about 10% of control wild-type egg chambers(n=96) (Fig. 5D). The reduction occurred only at the protein level, while the mRNA was not changed,as indicated by an in situ hybridization assay (data not shown). By contrast,overexpression of CblS had much more subtle effects(Fig. 5B,D). Furthermore, in contrast to the reduction in the intercellular and membrane region, the internal punctate HRP signals were dramatically increased in follicle cells with medium overexpression of CblL, but were not detectable in cells with high expression levels of CblL (Fig. 5E-H). EQ1 and GR1-Gal4 driven expression in the follicle cells is somewhat patchy, resulting in follicle cells with both higher and somewhat lower expression levels, allowing a direct comparison of the effects of higher versus intermediate levels of CblL (see Fig. S2 in the supplementary material). This suggests that in cells with high levels of CblL the receptor-ligand complexes may be undergoing very rapid trafficking and degradation. In the CblL overexpressing follicle cells, colocalization of HRP-Grk punctate signals with Rab5 (41%, n=27, Fig. 5I), Rab7 (67%, n=21, Fig. 5J), and Egfr (51%, n=41, Fig. 5K) were detected. Taken together, these results suggest that more Grk-Egfr complexes are in the follicle cells with the help of CblL than that in wild-type cells without CblL overexpression. This leads to a reduction of Gurken outside the follicle cells and an increase of Gurken inside the follicle cells. As a consequence, the promotion of Grk-Egfr endocytosis by CblL can change the overall distribution of the Gurken morphogen. In summary,Cbl negatively regulates Egfr signaling by facilitating the internalization and degradation of Grk-Egfr complex and thus helps to shape the pattern of Gurken protein expression.

We next tested the effect of Cbl loss-of-function on Gurken expression and distribution. Unlike the significant changes caused by CblL overexpression, there was no globally detectable difference in the overall Gurken expression pattern in egg chambers containing Cbl mutant follicle cell clones (Fig. 6A,B). But a strong reduction of the internalized Grk-Egfr complex in the Cbl mutant follicle cells was observed in dorsal(Fig. 6F,G), ventral(Fig. 6H) and posterior regions(Fig. 6C,D) of egg chambers,indicating that the internalization of Grk-Egfr complexes autonomously requires functional Cbl. Interestingly, at a local level, high levels of the Grk-Egfr complex were detected in wild-type cells adjacent to Cbl mutant cells (Fig. 6F,G). Furthermore, the HRP-Grk signal was detected even in the 11th-12th main body follicle cells when the Cbl mutant clone was located in the dorsal-anterior corner, whereas normally no HRP-Grk signal was detected in follicle cells located more posteriorly than the 10th follicle cell (Fig. 6E). These results very strongly suggest that Gurken is readily diffusible and travels further towards the posterior region, when the Grk-Egfr complex is not efficiently internalized in anterior follicle cells.

In the Drosophila ovary, different levels of Egfr signaling are crucial for axis establishment of the egg and the future embryo. The asymmetrically localized Gurken has been suggested to provide different levels of ligand, leading to various levels of Egfr activation. Previous studies had suggested that Gurken acts as a free soluble ligand released by the oocyte after proteolytic processing in the Golgi compartment(Ghiglione et al., 2002). However, a direct examination of the Gurken gradient has never been possible,as it has been very difficult to distinguish the distribution of Gurken in the oocyte, in the extracellular space between the oocyte and the follicle cells or on the follicle cell membrane with the resolution of confocal microscopy in immunostaining assays. We tried to examine Gurken through TEM, but nonspecific binding of the anti-Gurken antibody in the space between the oocyte and the follicle cell made this impossible. Nevertheless, it is important to know how far Gurken can travel in the egg chamber to evaluate whether all ranges of Egfr activation are determined by one ligand, Gurken, or whether they might also be affected by other signals. We therefore took an alternate approach and examined the ligand-receptor complex in the signal-receiving cells.

The fate of Gurken after binding to its receptor has not been previously analyzed. We fused Gurken to HRP, which allowed a more sensitive detection for Gurken after its internalization into the follicle cells (see Fig. S1 in the supplementary material; Figs 1and 2). The presence of the cytosolic HRP positive particles depends on the presence and function of Egfr,indicating that Gurken is internalized with the Egfr as a complex. Our results with shibire mutants further demonstrate that Dynamin is required for this process (Fig. 2J-M and Fig. 3G-I). Interestingly, we never saw clathrin-coated pits with or without HRP in follicle cells, whereas in the oocyte, clathrin-coated pits were easily detected (data not shown). It is possible that clathrin-dependent internalization is not a major pathway for the Grk-Egfr complex. Both Gurken and Egfr are colocalized with Rab5 and Rab7(Fig. 2N-Q, Fig. 3J-O), suggesting that the Grk-Egfr complex takes this endocytic pathway to the MVB/lysosome compartment,where signaling is terminated and Gurken is degraded. Accumulating evidence suggests that receptor signaling can occur in different cellular compartments(Hoeller et al., 2005), and Egfr signaling in the late endosome has been shown to mediate through a p14/MP1 scaffold (Teis et al.,2002). However, in Drosophila follicle cells, the function of internalization of the Grk-Egfr complex is mainly to terminate signaling, as blocking endocytosis by interfering with Dynamin function results in an excess Egfr signaling phenotype(Pai et al., 2006). In other words, the Egfr signaling mainly occurs on the cytoplasmic membrane in follicle cells.

Our data showed that a significant amount of HRP-Grk was detected in ventral follicle cells by immunostaining and immunoelectron microscopy using an anti-HRP antibody. Direct secretion of Gurken from the ventral surface of the oocyte to the ventral follicle cells is very unlikely, given results from tER studies carried out by both Rabouille and colleagues, as well as by us. Furthermore, the autonomous effect of Cbl clones on the vesicular signal of HRP-Grk in dorsal and ventral follicle cells suggest that the major spreading of Gurken in the egg chamber was not achieved through transcytosis(Fig. 6C-H). Our results are much more compatible with a mechanism where Gurken diffuses as a free secreted form or in a complex with some extracellular molecule such as HSPG(Belenkaya et al., 2004; Hacker et al., 2005), and distributes to the dorsal and ventral side in the space between the oocyte and follicle cells. This interpretation is also consistent with a quantification model of the Gurken morphogen in which 10% of Gurken was predicted to appear on the ventral side of egg chambers compared with dorsal levels(Goentoro et al., 2006). Taken together, our data demonstrate that Gurken acts as a long-range morphogen,which directly determines the fates of ventral follicle cells.

Morphogen gradients are controlled by several mechanisms, such as the level of receptors and the rate of endocytosis(Lecuit and Cohen, 1998; Teleman et al., 2001). Goentoro and colleagues have shown that overexpression of Egfr in the follicle cells traps Gurken near the dorsal anterior corner and prevents its diffusion,leading to an expansion of Pipe expression(Goentoro et al., 2006). This finding suggests that in the wild-type situation, most Egfr is bound with Gurken in dorsal follicle cells, and excess free Gurken diffuses to the lateral/ventral side of egg chambers to form the gradient.

Here, we report that endocytosis also plays a crucial role in shaping the gradient of the Gurken protein, which is similar to observations for Wingless and Fgf8 (Dubois et al., 2001; Scholpp and Brand, 2004). Dubois and colleagues had shown that endocytosis regulates the asymmetry of the Wingless gradient using HRP-Wingless to monitor Wingless degradation. It has been well established that Cbl acts as an endocytic adaptor to mediate the internalization of Egfr in mammals (Rubin et al., 2005; Schmidt and Dikic, 2005). Our previous data showed that overexpression of CblL resulted in the reduction of Egfr signaling in eggshell patterning and repression of its target gene, argos(Pai et al., 2006). We demonstrate here that Cbl facilitates the endocytosis of the Grk-Egfr complex into follicle cells based on our clonal analysis of Cbl mutant follicle cells (Fig. 6C-H). Very strikingly, neighboring wild-type cells contained more punctate HRP-Grk signals (Fig. 6E,F), indicating that more Gurken was available to these wild-type cells. This result suggests that larger quantities of free Gurken can travel towards the posterior when the Grk-Egfr complex cannot efficiently enter the dorsal anterior cells in the absence of Cbl. However, this subtle change in Gurken distribution cannot be revealed through examining the overall Gurken gradient by immunostaining (Fig. 6A). Conversely, more Grk-Egfr complexes were removed from the membrane to the Rab5/7-associated endocytic pathway in follicle cells with over-expressed CblL when compared with those in wild-type cells, and under these conditions,Gurken diffusion was limited (Fig. 5C,H-K). This observation indicates that CblL serves as a rate-determining component of the endocytic pathway. Increased levels of CblL effectively promote the endocytosis of Egfr. Furthermore, the significant reduction of Gurken protein in egg chambers with CblL overexpression at stage 9 suggests that internalization of the Grk-Egfr complex may lead to a rapid recycling of Egfr on the cytoplasmic membrane of follicle cells and thus allow more binding to Gurken. In conclusion, we demonstrate here that increasing the rate of endocytosis by overexpression of CblL can modify the Gurken distribution.

Interestingly, we noticed that the area of the lysosome and MVB vesicle in the lateral follicle cells was bigger than that in the dorsal or ventral follicle cells in wild-type egg chambers, suggesting that the endocytic activity was higher in the lateral follicle cells(Fig. 5H-J). This difference was observed after early stage 9 and was Gurken dependent. The high endocytic activity in lateral follicle cells could set up a sharp boundary for the Gurken morphogen between the dorsal and ventral side of the egg chamber. This hypothesis is consistent with the quantification model(Goentoro et al., 2006), which predicted a steep reduction for the Gurken morphogen, from 63% to 22% at the lateral region. Thus, in response to early Gurken/Egfr activity, during stage 9 and later, once free Gurken binds to the Egfr in lateral follicle cells, the ligand-receptor complex may be internalized quickly as these lateral follicle cells may have a very active endocytic machinery. Because of a high activity of a degradation mechanism in the lateral follicle cells, only very low levels of Gurken would then be able to travel to the ventral region, leading to a low level of Egfr signaling in the ventral follicle cells, which would ensure the robustness of the Gurken gradient.

We thank Jean-Paul Vincent for the HRP plasmid, and Trudi Schüpbach for encouraging this project from the initial stage and for providing gurken genomic DNA, cDNA and fly stocks. We also thank Marcos González-Gaitán and Cheng-Ting Chien for providing fly stocks. Many thanks to the Microscope Facility at CGU for the excellent assistance with confocal microscopy. We are grateful to Trudi Schüpbach, Laura Nilson, Y-Henry Sun, Cheng-Ting Chien, Jui-Chou Hsu, Chien-Kuo Lee and Tzu-Chien V. Wang for critical reading of this manuscript. This work was supported by grants from the Chang-Gung Memorial Hospital (CMRPD140201-2 to L.P. and CMRPD340232 to W.L.), the National Science Council of Republic of China (NSC 95-2311-B-182-005 to L.P.) and the Ministry of Education, Taiwan,ROC (EMRPD170121 to L.P.).