Focal adhesion kinase controls morphogenesis of the Drosophilaoptic stalk

Glial cells play important roles during the formation of the nervous system. For instance, they provide trophic support to neurons, modulate axon pathfinding and guide nerve fasciculation(Booth et al., 2000; Leiserson et al., 2000; Lemke, 2001; Gilmour et al., 2002; Chotard and Salecker, 2004). At present the molecular and cellular mechanisms that regulate glial cell behavior, such as migration and morphogenesis, are largely unknown. The main classes of central nervous system glia in Drosophila exhibit many morphological and functional similarities to their mammalian counterparts(Freeman and Doherty, 2006);therefore they represent a good model for studying general mechanisms underlying glial development and behavior.

The Drosophila visual system is one of the most extensively studied nervous systems (Huang and Kunes,1996; Kunes, 2000; Clandinin and Zipursky, 2002; Chotard and Salecker, 2004). It consists of the compound eyes and the optic ganglia. The compound eye consists of approximately 750 ommatidial units, each of which contains eight photoreceptor cells. During the development of the visual system,photoreceptor cells send their axons (R axons) from the eye primordium (eye disc) to their targets in the optic lobe of the brain through a structure called the optic stalk. All R axons from an eye disc come together and make a single thick bundle in the optic stalk(Fig. 1A). The optic stalk also contains two kinds of glial cells, the wrapping glial (WG) cells and surface glial (SG) cells (Meinertzhagen and Hanson, 1993; Perez and Steller, 1996; Hummel et al.,2002). The WG cells are intermingled with R axons and extend long processes that wrap around several R axons, so that the entire R axon bundle is subdivided into groups of smaller bundles. The WG processes form an inner sheath to segregate these bundles from each other during the pupal stage. By contrast, SG cells form an outer sheath that surrounds the entire bundle of R axons. It is clear that the optic stalk glia play roles in R axon innervation and ensheathment (Rangarajan et al.,1999; Hummel et al.,2002); however, the precise structure and development of the optic stalk remain largely unknown. Hence, we have focused our efforts to elucidate SG cell development and optic stalk morphogenesis.

We found that SG cells in the optic stalk have characteristic bipolar morphology and form a single-cell monolayer covered by a basement membrane(BM). Optic stalk expansion occurs before R axon innervation. Moreover, even in those mutants with no R axon innervation, SG cells form a single-cell-layer tube that develops normally. Therefore, we conclude that SG cells autonomously form the optic stalk. We then sought to elucidate the molecular mechanisms underlying optic stalk formation. In screening for mutants that exhibit defects in R axon innervation or for genes that are specifically expressed in the optic stalk, we identified the Fak56D and CdGAPr genes as encoding important components required for optic stalk formation. Fak56D (Fox et al.,1999; Fujimoto et al.,1999; Palmer et al.,1999) is a Drosophila homolog of mammalian focal adhesion kinase (FAK; also known as Ptk2), which is known to be a main regulatory component of focal contacts (Parsons,2003; Mitra et al.,2005). Focal contacts are large integrin complexes that anchor the cytoskeleton of cells to the extracellular matrix(Critchley, 2000). While focal contacts generate traction forces during migration, their disassembly is also crucial to the control of cell migration(Lauffenburger and Horwitz,1996; Webb et al.,2002; Webb et al.,2004). FAK is involved in cell migration through disassembly of focal contacts (Ilic et al.,1995; Gilmore and Romer,1996; Parsons et al.,2000), or through regulation of cytoskeletal rearrangement via Rho GTPases (Hildebrand et al.,1996; Zhai et al.,2003). In Drosophila, several studies indicate that Fak56D also acts in focal contacts in a similar way to that of mammalian FAK. Fak56D is tyrosine-phosphorylated and localized to focal contacts upon the plating of embryonic cells onto extracellular matrix components(Palmer et al., 1999). However, as no Fak56D loss-of-function phenotype has been reported,the in vivo function of Fak56D remains elusive.

In Fak56D mutant animals, SG cells were not arranged into a tubular structure, although cell proliferation and differentiation were normal. Clone labeling analysis suggested that SG cell distribution along the anteroposterior (AP) axis was defective in Fak56D mutants. We also identified CdGAPr as a possible functional partner of Fak56D. CdGAPr has a GTPase activating domain that is homologous to that of mammalian CdGAP (Sagnier et al.,2000). It has been shown that mammalian CdGAP regulates cell cytoskeletal rearrangement through Rac or Cdc42(Lamarche-Vane and Hall,1998); however, in vivo function of neither CdGAP nor CdGAPr has been described. We found that loss-of-function alleles of CdGAPrexhibited a Fak56D-like phenotype. Moreover, a strong genetic interaction was observed between Fak56D and CdGAPr,indicating that these genes act together to regulate SG cell behavior. Recently, mammalian CdGAP was shown to localize to focal contacts, and to be required for regulation of cell morphology and motility(LaLonde et al., 2006). Our results provide strong evidence that optic stalk shape is controlled by mechanisms acting autonomously in SG cells, in which Fak56D and CdGAPr play crucial roles.

The following mutants and transgenic strains were used: yw flies were used as wild-type controls. optomotor-blind^P1(omb^P1) is a reporter construct that contains a P-element carrying lacW upstream of the omb (bifid - FlyBase)gene (Tsai et al., 1997). GMR-Gal4 was described previously (Hay et al., 1997). UAS-GFP encodes a cytoplasmic form of GFP and UAS-mCD8::GFP encodes a membrane targeted form of GFP(Lee and Luo, 2001). Act>CD2>Gal4 was described previously(Pignoni and Zipursky, 1997). Df(2R)BSC26 (Parks et al.,2004), EY13451 (Bellen et al.,2004), ED(2L)1303 (Ryder et al., 2004) and CdGAPr-6.1(Billuart et al., 2001) were obtained from the Bloomington Stock Center. Fak56D^CG1 was described previously (Grabbe et al.,2004), as was mys¹ and mys^XG43 (Bunch et al.,1992), UAS-Fak56D(Palmer et al., 1999), sine oculis¹ (Kunes et al., 1993) and ptcGal4 (G559.1)(Shyamala and Bhat, 2002). NP4702-Gal4, NP2109-Gal4 and NP3053-Gal4 were isolated from a screening of enhancer trap lines (NP lines) (Hayashi et al., 2002) for SG cell-specific expression.

Immunohistochemistry was performed as described previously(Huang and Kunes, 1996). Rabbit anti-Perlecan (also known as Trol - FlyBase)(Friedrich et al., 2000) and rabbit anti-LamininA (Gutzeit et al.,1991) were used at a dilution of 1:3000. Mouse anti-Repo, mouse anti-Dac and mouse anti-Caoptin (referred to as mAb24B10 throughout the text)were obtained from the Developmental Studies Hybridoma Bank and were used at a 1:10 dilution. Rabbit anti-phospho-Histone H3 (Upstate Biotechnology) was used at a 1:200 dilution. Mouse anti-α-Tubulin (Sigma) was used at a dilution of 1:500. Rabbit anti-β-galactosidase (β-gal) (Cappel) was used at a dilution of 1:1000. Goat Cy3 anti-HRP (Accurate Chemical and Scientific) was used at a dilution of 1:200. Following secondary antibodies (Jackson) were used at 1:200 dilutions: anti-mouse Cy3, anti-mouse Cy5, anti-mouse FITC,anti-rabbit Cy3, anti-rabbit Cy5. Specimens were viewed on a Zeiss LSM510 confocal microscope.

For analysis of the optic stalk growth, the number of pixels within an area of the optic stalk cross section encircled by the BM, which is visualized with anti-Perlecan, was calculated with Adobe Photoshop. Statistical analysis was performed using Microsoft Excel. Statistical differences were compared by Student's t-test.

In situ hybridization was performed as described previously(Nagaso et al., 2001). Sense and antisense RNA probes were synthesized using the DIG RNA labeling kit(Roche).

Total RNA was isolated from late third instar larval brains of ywor CdGAPr^NP3053 homozygotes. Reverse transcription was performed with 3 μg total RNA and Superscript II reverse transcriptase(Invitrogen). The resulting cDNA was amplified by SYBR Premix Ex Taq (Takara),and products were detected on an ABI PRISM 7000 (Applied Biosystems). The fold decrease of CdGAPr transcripts was calculated relative to Actin 5C transcripts.

The primers used for real-time PCR were: CdGAPr(5′-AATCGCCCACTTTCAGTGTC-3′ and 5′-GGTACGCTCAGTTCGTTAGG-3′) (123 bp); Actin 5C(5′-AAGTGCGAGTGGTGGAAGT-3′ and 5′-CATGCGCCCAAAACGATGA-3′) (125 bp).

For labeling a subset of SG cells, males with the genotype of Actin>CD2>Gal4 were mated to hsflp¹²²; UAS-CD8::GFP females. For labeling a subset of SG cells in Fak56D mutants, males with the genotype of Actin>CD2>Gal4;Fak56D^CG1/CyoAct were mated to hsflp¹²²; Fak56D^CG1/CyoAct; UAS-CD8::GFP females. The progeny at 60-84 hours after egg laying (AEL) were heat shocked (37°C for 5 minutes). Dissections were carried out at late third instar larval stage.

During development of the Drosophila adult visual system, R axons extend toward the brain through the optic stalk(Fig. 1A). In the optic stalk,two types of glial cells express the glia-specific transcription factor Repo(Xiong et al., 1994; Halter et al., 1995)(Fig. 1B). These cells are WG and SG, as classified according to their position, morphology and marker expression (Fig. 1I). We found that ptcGal4 and NP4702-Gal4 specifically labeled WG and SG,respectively, at late third instar larval stage. Using green fluorescent protein (GFP), we closely observed the morphology of WG and SG cells. As described previously (Hummel et al.,2002), WG cell bodies are located in the eye disc or inside the lumen of the optic stalk and are intermingled with R axons. They extend long processes along R axons and establish axonal ensheathment(Fig. 1C). Each WG cell process wraps several ommatidial R axons into a bundle within the optic stalk. However, NP4702-expressing SG cell bodies are located in the optic stalk closely associated with the R axon bundle, but never intermingled with R axons(Fig. 1D)(Hummel et al., 2002). We found that SG cells have long processes extending along the AP axis(Fig. 1E,E′). These SG cells form a monolayered tube that ensheathes the entire R axon bundle(Fig. 1F). We measured the size of the optic stalk and found that it was invariable, suggesting that size of the optic stalk structure is precisely regulated(Fig. 1G,H). We also found that the optic stalk is covered by the BM. Major components of the BM, LamininA and Perlecan, were localized on the surface of the SG cells(Fig. 1F,G and Fig. 2). Together, SG cells form a monolayered tubular structure that is covered by the BM(Fig. 1J), resulting in the formation of the outer sheath of R axons. This SG cell tubular structure corresponds to the optic stalk (Steller et al., 1987; Kunes and Steller,1993).

The Drosophila larva has a visual system much simpler than that of the adult. This structure is known as Bolwig's organ, consists of 12 photoreceptors and develops in the embryonic stage. In disconnectedmutant larvae, Bolwig's nerves fail to project properly, which leads to failure of optic stalk formation and R axon projection(Steller et al., 1987). These results indicate that the optic stalk is formed along Bolwig's nerves and later serves as the R axon pathway. However, development of the optic stalk at a later stage is largely unknown. We examined optic stalk development and asked if R axons are required for development of the optic stalk. At second instar larval stage before photoreceptor cell differentiation, glial cells form a tubular structure in which Bolwig's nerves extend(Fig. 2A,A′,C). The SG cell marker (NP4702-Gal4) is expressed in these glial cells(Fig. 2C). At early third instar stage the R axons have yet to extend; however, we found that the optic stalk expands and glial cells increase in number(Fig. 2B,D). The mean cross-sectional area of the early third instar optic stalk was about five times larger than that of the late second instar(Fig. 3E). In addition, on average 38 glial cells were observed (n=25) in the early third instar optic stalk, while only 15 glial cells on average were observed in late second instar larvae (n=13). These data indicate that the optic stalk undergoes expansion at larval stages before R axon innervation.

We further examined how the optic stalk expands during the early third instar larval stage. While the size of SG cells did not seem to change much during this optic stalk expansion (compare Fig. 2C with D), mitotic cells were observed in the optic stalk glia (Fig. 2E,E′). Although glial cells could migrate into the optic stalk during the expansion, virtually no GFP labeled clones generated on the optic lobe surface (Fig. 2G-G″) entered into the optic stalk. Therefore a boundary may exist between the optic stalk and optic lobe. In addition, virtually no ptclacZ-positive WG cells were seen in the optic stalk at early third instar stage, except for an occasional appearance in the basal region of the eye disc (n=12) (Fig. 2F,F′). These data suggest that optic stalk expansion is primarily or wholly due to SG cell proliferation.

We next examined if optic stalk formation is affected in the absence of R axons. In sine oculis (so) mutants eye formation is defective, and thus no R axons project(Kunes et al., 1993; Perez and Steller, 1996). However, we found that in these mutants the optic stalk grew as normally as the wild type, although it failed to maintain a round-shaped cross section(Fig. 3A-F). These data indicate that the proliferation and arrangement of SG cells are autonomously regulated and independent of R axons.

The above results clearly show that SG cells autonomously shape the optic stalk. We then explored the molecular mechanisms underlying SG cell development. As the optic stalk is formed before R axon innervation(Fig. 2), disruption of this structure may affect some aspects of the projection pattern of R axons, as has been observed previously in disconnected mutants(Steller et al., 1987). We then searched mutants for abnormal R axon projection and identified Fak56D. A null mutant allele for Fak56D^CG1(Grabbe et al., 2004) showed a defect in R axon bundle formation at the optic stalk(Fig. 4A,B). We found that the optic stalk was also severely disrupted in the Fak56D mutant(Fig. 4C). As the severity of the optic stalk phenotype was variable, we classified the phenotype into five groups (Fig. 4C). Class 0 represents the wild-type phenotype, in which the length is at least two times longer than the diameter. In class 1, length is shortened or the diameter is broadened but length is still longer than diameter. In class 2, length is equal to diameter. In class 3, length is shorter than diameter. In class 4,length is much shorter than diameter and the optic stalk is only barely seen. In order to avoid ambiguity, we considered classes 2-4 to be optic stalk defective. According to this classification, about 80% of Fak56D^CG1 animals exhibited significant defects in optic stalk morphogenesis (Fig. 4D). The phenotype of Fak56D^CG1 in trans to a deficient chromosome that uncovers the Fak56D locus was indistinguishable from that of Fak56D^CG1 homozygotes(Fig. 4D). We never observed a shortened optic stalk with a diameter larger than its length in late second instar Fak56D mutants, although slightly truncated optic stalks were occasionally observed (Fig. 4E). Defects in optic stalk morphology became clearer and more severe as development proceeded to the third instar stage(Fig. 4E). Therefore, these data indicate that Fak56D is not required for the initial establishment of the optic stalk but is required during optic stalk expansion.

The above results indicate that Fak56D is required for optic stalk morphogenesis and R axon bundle formation. The Fak56D mRNA is expressed in the third instar larvae (Fig. 4F). To address whether Fak56D acts in SG cells or R axons, we expressed exogenous Fak56D in Fak56D mutants using the UAS/Gal4 system (Brand and Perrimon,1993). We expressed Fak56D using NP2109-Gal4, which is specifically expressed in SG cells and WG cells but not in photoreceptor cells(Fig. 5A). Alternatively we used NP4702-Gal4, which is expressed in SG cells but not in photoreceptor cells (Fig. 5B). In both cases the optic stalk defect was significantly rescued(Fig. 5D,E,G). In contrast,expression of Fak56D only in photoreceptor cells using GMR-Gal4(Fig. 5C) did not rescue the optic stalk defect (Fig. 5F,G). These data provide strong evidence that Fak56D is autonomously required in SG cells for optic stalk formation but is not required in R axons. We also found that overexpressed Fak56D in a wild-type background resulted in a thinner optic stalk (Fig. 5H), further supporting a crucial role for Fak56D in controlling optic stalk morphology.

We next investigated which process of SG cell development is affected in Fak56D mutants. Mammalian FAK is reported to function in the regulation of cell proliferation (Gilmore and Romer, 1996). However, during the early larval stages in which SG cells proliferate, no significant difference in SG cell numbers was observed between wild type and Fak56D mutants. Therefore, it is unlikely that Fak56D plays a role in regulation of SG cell proliferation during optic stalk expansion. We then examined the morphogenesis of the optic stalk. In wild type, strong expression of α-Tubulin was observed in SG cells within the optic stalk(Fig. 6A). In Fak56Dmutants, the intensity of α-Tubulin was preserved. However, in contrast to wild type, the SG cells failed to form a tubular structure but were instead spread over the optic lobe surface (Fig. 6B). This abnormal distribution of SG cells was also confirmed using omb^P1 as an alternative SG cell marker (data not shown). We also found that this defect in SG cell localization occurs before R axon innervation (Fig. 6C-D′), thus supporting the hypothesis that the primary defect of the Fak56D mutant is in SG cells.

To gain more insight into the Fak56D phenotype, we examined how SG cells are distributed during optic stalk formation. We found that clonally labeled SG cells with GFP are always distributed along the AP axis in a chain-like shape (Fig. 6E). We almost never observed clones that spread along the dorsoventral axis(n=58). This suggests that SG cells are actively arranged during optic stalk expansion rather than simply distributed as a result of cell division. In Fak56D mutants, SG cells exhibited the same elongated morphology as wild type (Fig. 6F). However, we occasionally observed SG cells that failed to correctly distribute along the AP axis(Fig. 6G). In order to precisely quantify the SG cell distribution, we generated clones consisting of a few cells in wild type or Fak56D mutants and examined the positions of SG cells. We found that SG cells in Fak56D mutants dispersed along the AP axis to a lesser extent than wild-type cells examined under the same conditions (Fig. 6H). Mean distance between SG cells within clones was shorter (P<0.001) in Fak56D mutants (18.6±1.8 μm; n=46,) than in wild type (29.4±2.1 μm; n=47).

To gain additional insights into the molecular mechanisms underlying SG cell morphogenesis, we screened 400 Gal4 enhancer trap lines for SG-cell-specific expression and identified line NP3053(Fig. 7A), which has an insertion in the first intron of the CdGAPr gene(Fig. 7B). NP3053-Gal4 is expressed restrictively in SG cells and WG cells, as well as the glial cells that surround the optic lobe (Fig. 7A). We found that the optic stalk was severely disrupted in animals homozygous for CdGAPr^NP3053(Fig. 7C,D). Real-time PCR demonstrated that CdGAPr transcript level was decreased in CdGAPr^NP3053 homozygotes by approximately six-fold compared with wild type, thus suggesting that CdGAPr^NP3053is a loss-of-function mutation. CdGAPr^NP3053 in trans to another P insertion allele CdGAPr^EY13451, or in trans to the deficiency chromosome Df(2L)ED1303, which uncovers the CdGAPr locus, also exhibited an optic stalk defect similar to that of CdGAPr^NP3053 homozygotes(Fig. 7E). Moreover, expression of CdGAPr dsRNA (Billuart et al.,2001) with NP3053-Gal4 resulted in this similar optic stalk defect(Fig. 7F). Therefore, these genetic data suggest that CdGAPr is required for optic stalk formation.

We also found that CdGAPr and Fak56D genetically interact. Trans-heterozygous mutants for CdGAPr and Fak56Dexhibited the same defect as Fak56D homozygotes in optic stalk formation (Fig. 7G). In comparison, heterozygous mutations in either gene alone exhibited only slight defects in optic stalk formation (Fig. 7G). These data suggest that CdGAPr acts together with Fak56D within a common genetic pathway.

Mammalian FAK and CdGAP are known to act in integrin signaling. We next tested whether integrins and Fak56D act within the same signaling cascade. Integrins function as heterodimers of one α and one β subunit. Two genes encode integrin β subunits in the fly, βPS and βν. We found that null mutants for myospheroid, which encodes βPS,showed significant genetic interaction with Fak56D and animals trans-heterozygous for myospheroid and Fak56D exhibited a similar phenotype to Fak56D homozygotes(Fig. 7G). This suggests thatβ PS acts together with Fak56D in optic stalk morphogenesis.

SG cells are regularly arranged and form the optic stalk, a monolayered tube that surrounds the entire R axon bundle. During larval development, the optic stalk becomes larger as SG cells proliferate, and this seems to adjust the size of the stalk so that it can wrap around the R axon bundle. The proliferation and morphogenesis of SG cells could depend on signals from the incoming R axons, because various cellular events of glial cells, such as differentiation, migration and proliferation, have been shown to require interaction with neuronal axons. For example, in the developing rodent optic nerve, Sonic hedgehog (Shh) from retinal ganglion cells (RGCs) promotes the proliferation of astrocytes (Wallace and Raff, 1999; Dakubo et al.,2003). Moreover, initial formation of the optic stalk during embryogenesis was shown to be dependent on the axons from larval photoreceptor cells (Steller et al., 1987). However, our data demonstrate that SG cells in Drosophila form the optic stalk independently of R axons. First, SG cells proliferate before R axon innervation, which leads to the expansion of the optic stalk. Second, in so¹ mutants the optic stalk develops normally despite complete lack of R axon innervation. In addition, specific expression of a Fak56D transgene in SG but not photoreceptor cells rescued a defect in optic stalk morphology, hence indicating the presence of an intrinsic mechanism in SG cells to regulate optic stalk morphogenesis. Some kinds of glial cells are reported to develop independently of axon innervation in a similar way to SG cells. For instance, during Drosophila visual system development,retinal basal glia (RBG) are derived from the optic stalk and migrate into the eye disc. This process was shown to be independent of R axons(Rangarajan et al., 1999). In these cases, glial cells must behave by unknown mechanisms independent of axon cues. As our results demonstrated that SG cell development is highly independent of R axons, this system can provide an excellent model to elucidate mechanisms underlying axon-independent glial development and behavior.

In this study, we have elucidated cellular mechanisms underlying optic stalk morphogenesis in Drosophila. Optic stalk morphology and development are precisely regulated, and the optic stalk expands throughout the larval stages via cell proliferation (Figs 1, 2). We also found that SG cells were distributed along the AP axis during optic stalk formation. As cells in a clone are often associated with each other, SG cells may migrate along the nearby SG cells that have extended processes along the AP axis. Such homotypic migration is reported for neuronal precursors in the adult mammalian subventricular zone (SVZ). Neuronal precursors migrate to the olfactory bulb in chains, by sliding along each other without the assistance of other cell types (Luskin, 1993; Lois and Alvarez-Buylla, 1994; Wichterle et al., 1997). Such directed distribution of SG cells is likely to be necessary for keeping the optic stalk thin and long. The mechanism regulating tube morphogenesis is largely unknown. Our finding that SG cells undergo cell arrangement during optic stalk expansion provides an interesting insight into the morphogenesis of a tubular structure.

It is known that the optic stalk disappears during the pupal stage and that the ommatidia are set much closer to the optic lobe. We found that SG cells begin to locate at the surface of the optic lobe during early pupal stages(data not shown). This is similar to what we observed in Fak56Dmutants during larval stages. It seems that the optic stalk is degenerating during the pupal stage through relocation of SG cells to the optic lobe, and this process might be regulated by FAK activity.

In an attempt to identify the molecular mechanisms underlying optic stalk formation, we found that the optic stalk was disrupted in Fak56Dmutants. Expression of a Fak56D transgene in SG cells, but not photoreceptor cells, rescued the defect. This indicates that Fak56Dis autonomously required in SG cells. Fak56D is apparently required during the expansion of the optic stalk. As we did not detect any differences in number of SG cells between wild type and Fak56D mutants, it is unlikely that Fak56D regulates the proliferation of SG cells. We found that SG cells were mis-localized on the surface of the optic lobe in Fak56D mutants instead of forming a tubular structure. Visualization of newly divided cells by clonal labeling suggests that SG cells tend to migrate along the AP axis, and this could lead to formation of a longitudinal tube structure (Fig. 8). In Fak56D mutants SG cells partially lose tendency to migrate along the AP axis; this could lead to enlargement of the diameter as opposed to the length. Ectopic localization of SG cells on the optic lobe observed in Fak56D mutants is likely to be the result of the optic stalk widening. Mammalian FAK plays a central role in cell migration(Mitra et al., 2005); hence in Drosophila Fak56D may regulate optic stalk formation via regulation of cell migration.

Fak56D might also be required for adhesion between SG cells. siRNA-mediated mammalian FAK knockdown results in loss of N-cadherin-based cell-cell adhesion in Hela cells (Yano et al., 2004). In addition, it has been shown that overexpression of a kinase-defective mutant of FAK in cultured cells blocks the hyperosmolarity-induced E-cadherin accumulation at the cell periphery(Quadri et al., 2003). As SG cells are closely attached to each other, adhesion between SG cells may be important for keeping optic stalk tubular structure. SG cells are also attached to the BM throughout optic stalk formation. Because Fak56D was implied to regulate integrin adhesion(Grabbe et al., 2004), it is possible that Fak56D is required for SG cells to form proper adhesion to the BM.

In addition to Fak56D, we also identified CdGAPr as a regulator of optic stalk morphology. Fak56D and CdGAPrexhibited a strong genetic interaction. As both mammalian FAK and CdGAPr are known to act at focal contacts, this raised the possibility that focal contacts are important for optic stalk morphogenesis. Main components of focal contacts, such as integrins or Paxillin, are conserved between vertebrates and invertebrates (Wilcox et al.,1989; Bokel and Brown,2002; Chen et al.,2005). In Drosophila, Fak56D is implicated in integrin-involving molecular mechanisms(Palmer et al., 1999; Grabbe et al., 2004). We found a genetic interaction between Fak56D and myospheroid(mys), which encodes a β-subunit of integrin, βPS. This suggests that Fak56D and CdGAPr act together with integrins in focal contacts to regulate SG cell behavior. Because the interaction between Fak56D^CG1 and mys¹ (a null allele) is much weaker than the interaction between Fak56D^CG1 and CdGAPr hypomorphic alleles, it is possible that another βsubunit, βν, compensates for the loss of βPS. βPS andβν were shown to act together to regulate midgut cell migration(Devenport and Brown, 2004). It is also possible that Fak56D function is regulated via other receptors,such as G-protein coupled receptors (GPCRs). Mammalian Pyk2, another FAK family member, is known to be activated via GPCRs(Dikic et al., 1996; Yu et al., 1996).

CdGAPr encodes a GAP domain that regulates the activity of Rho-family GTPases. The mammalian CdGAP(Lamarche-Vane and Hall, 1998; LaLonde et al., 2006) was shown to regulate Rac and Cdc42, which are known regulators of the actin cytoskeleton (Nobes and Hall,1995), and was reported to be required for cell morphology and motility (Etienne-Manneville and Hall,2002). This raises the possibility that FAK regulates cytoskeletal rearrangement via activation of CdGAP, although exact functional interaction between the two proteins remains elusive. In fact, the mammalian FAK interacts directly with guanine nucleotide exchange factors or GTPase-activating proteins in the regulation of cytoskeleton(Hildebrand et al., 1996; Liu et al., 2002; Medley et al., 2003; Zhai et al., 2003). More biochemical studies are required for elucidating the molecular mechanisms that underlie cell behaviors regulated by focal adhesion signaling.

We thank Adrian Moore for his critical reading of the manuscript. We thank Y. Watanabe for access to ABI PRISM 7000 and A. Kagami for assistance with real-time PCR assay. We are grateful to all the members of Tabata laboratory for helpful discussions during the course of the work. We also thank S. Baumgartner for providing antibodies, and T. A. Bunch and R. H. Palmer for fly strains. We thank the Bloomington Stock Center, Exelixis, the National Institute of Genetics, the Szeged Drosophila Stock Center and the Kyoto Stock Center for fly stocks. We would also like to thank the Genetic Studies Hybridoma Bank for antibodies. This work was supported by Grants-in-Aid from the Ministry of Education, Science and Culture of Japan and by the Toray Science Foundation to T.T. Support for S.M. was provided by a Predoctoral Fellowship from the Japan Society for the Promotion of Science for Japanese Junior Scientists.