DPP signaling controls development of the lamina glia required for retinal axon targeting in the visual system of Drosophila

The mechanisms by which each axon is navigated to the correct destination and establishes synapses with its correct partner is a central issue for the formation of a functional nervous system. Glial cells present in the nervous system support the growth and survival of the neurons, and, in some cases, the glial cells act as a platform for the migration of neuronal cell precursors(Rakic, 1988). In addition to their well-known supportive role for neuronal cells, glial cells may also play active roles in axon navigation and synapse formation (reviewed by Chotard and Salecker, 2004). In the embryonic central nervous system of Drosophila melanogaster,midline glia have been shown to control the guidance of commissural axons, a process mediated by roundabout (robo) receptors expressed in neurons, as well as slit in the midline glia and commissureless, which regulates ROBO(Tear et al., 1996; Brose et al., 1999; Kidd et al., 1998; Kidd et al., 1999). In addition, in mammals it has recently been shown that thrombospondins secreted from astrocytes promote synaptogenesis in cultured retinal ganglion cells(Christopherson et al.,2005).

The visual system of Drosophila provides a powerful genetic tool with which to address the molecular and cellular mechanisms underlying these processes. The visual system comprises a pair of compound eyes and optic ganglia, which are the visual processing centers of the brain (reviewed by Meinertzhagen and Hanson,1993; Clandinin and Zipursky,2002; Tayler and Garrity,2003). A repeated unit of the compound eye, the ommatidium,contains eight photoreceptor neurons (R neurons). Axons from photoreceptor R7 and R8 connect to a deeper target site known as the medulla. By contrast,R1-R6 neurons stop between two layers of lamina glial cells at the bottom of the ganglion layer known as the lamina(Winberg et al., 1992). R axons provide signals, Hedgehog (Hh) and Spitz, to the lamina neuron precursors (LPCs) and induce LPC differentiation into lamina neurons, the future synaptic partners of the R axons(Huang and Kunes, 1996; Huang and Kunes, 1998; Huang et al., 1998). The lamina glia play a crucial role in the guidance of R1-R6 growth cones, by serving as an intermediate target of R1-R6 axons(Poeck et al., 2001). It has also been shown that R axons play an important role in the migration of lamina glia from the progenitor domain (Suh et al., 2002). These glia arise from dorsal and ventral marginal progenitor zones and migrate into the lamina target region.

Although both the importance of lamina glia for R axon targeting and the mechanism for their migration have been analysed(Dearborn and Kunes, 2004),little is known about the signaling mechanisms that regulate the development of these glial cells. We show that decapentaplegic (dpp)plays a key role in the process. DPP is a homolog of BMP and is a member of the TGFβ superfamily, which is widely conserved among species and which plays central roles in diverse cellular and molecular processes (reviewed by Hogan, 1996; Massague and Chen, 2000). In Drosophila, DPP provides positional information by acting as a morphogen (Lecuit et al.,1996; Nellen et al.,1996) (reviewed by Tabata and Takei, 2004). In the vertebrate nervous system, signaling mediated by BMP family proteins operates in diverse processes during development. For example, BMP promotes the differentiation of astroctytes from multipotent progenitor cells, but inhibits the differentiation of neuronal and oligodendroglial cell types (reviewed by Mehler et al., 1997).

In the Drosophila visual system, dpp is expressed in dorsal and ventral marginal zones via induction by wingless(wg) (Kaphingst and Kunes,1994). An optic lobe mutant for dpp shows defects in the neuropile in the medulla cortex. These defects are presumably caused by a failure in neuronal cell fate specification of cells in the outer proliferation center (OPC), i.e. in the neuroblast populations(Kaphingst and Kunes, 1994). However, brains from animals with the same dpp mutant allele also lack retinal axon input. In turn, this affects multiple steps of visual system development and makes it difficult to address the other roles of DPP signaling in the visual system.

We examined the role of DPP signaling in the visual system using clonal analysis and the targeted expression of inhibitors specific to signaling. We show that dpp is required for the establishment of correct R-axon projection patterns via the control of lamina glia development. The expression of glial cells missing/glial cells deficient (gcm;also known as glide) is the earliest marker for lamina glia differentiation (Dearborn and Kunes,2004). gcm-positive cells, as well as mature glial cells,are greatly reduced in dpp mutant animals and in animals with clones mutant for the intracellular components of DPP signaling. Conversely,overactivation of DPP signaling induces ectopic expression of gcm and the production of mature glial cells. Taken together, these data suggest that DPP signaling is required for R axon projection, as DPP signaling mediates the production of lamina glia, the intermediate targets of the R axons, via the control of gcm expression.

The following mutant and transgenic strains were used in this study. y w flies were used as wild-type controls. dpp-lacZ is an enhancer-trap allele of dpp(Tabata and Kornberg, 1994). optomotor-blind (omb)-Gal4 contains a P element carrying Gal4 in the promoter region of the omb gene(Lecuit et al., 1996). The NP6099-Gal4 driver is expressed in the lamina precursor cells and developing lamina neurons; expression is hardly detected in the mature lamina neurons (Hayashi et al., 2002)(D.U., S.M. and T.T., unpublished). UAS-GFP encodes a cytoplasmic form of GFP and UAS-GFPnls encodes a nuclear-tagged form of GFP(Lee and Luo, 1999). UAS-Daughters against dpp (Dad) has been used in a variety of tissues to specifically block processes mediated by DPP signaling(Tsuneizumi et al., 1997; McCabe et al., 2004; Kawase et al., 2004). UAS-thickveins (tkv)^Q253D has an amino acid replacement in tkv and acts as a constitutive-active form of the receptor (Wiersdorff et al.,1996). UAS-glide^DN and UAS-glide^N7-4DN have been described by Soustelle et al.(Soustelle et al., 2004). gcm-lacZ^r87 is a reporter that has a P element insertion containing lacZ in the promoter of the gene(Jones et al., 1995; Vincent et al., 1996; Badenhorst, 2001). sine oculis¹ (so¹) is a spontaneous allele that lacks eye formation almost completely(Kunes et al., 1993). hh^bar3 is an eye-specific regulatory allele(Lee et al., 1992; Heberlein et al., 1993; Huang and Kunes, 1996; Huang and Kunes, 1998). The dpp^d6 and dpp^d12 alleles are described as disc-specific regulatory alleles, but defects in the brain have also been reported (Kaphingst and Kunes,1994). Medea⁸ is identified as a null allele and Medea⁴ as a strong hypomorph(Wisotzkey et al., 1998); they showed an indistinguishable phenotype in the experiments described here.

Medea⁴ and Medea⁸ mutant clones were generated by the FLP-FRT-mediated mitotic recombination system,as previously described (Xu and Rubin,1993), in the Minute mutant background; y w; Medea⁴ FRT82B/TM6B or y w; Medea⁸FRT82B/TM6B flies were crossed to y w hsflp¹²²; FRT82B Ubi-GFP M (3)^w124. The progeny were heat shocked at the first instar at 37°C for 60 minutes, and dissected in the late third instar. Overexpression using the GAL4-UAS system was performed at 29°C except for for UAS-glide^DN, which was performed at 25°C.

Immunohistochemistry was performed as described previously(Huang and Kunes, 1996; Takei et al., 2004). Rat anti-DAC and rat anti-β-Galactosidase (β-GAL) were raised against synthetic peptides (Hokudo) and diluted 1:100 and 1:4000, respectively. Anti-Cyclin A was a gift from S. B. Selleck and was diluted 1:200(Nakato et al., 2002). Mouse monoclonal anti-β-GAL (Promega) was diluted 1:250. The mouse monoclonal antibodies anti-Caoptin (referred to as mAb24B10 throughout the text),anti-CUT and anti-REPO were provided from the Developmental Studies Hybridoma Bank and were used at a 1:10 dilution. Specimens were mounted with Vectashield Mounting Media (Vector Laboratories) and viewed on a Zeiss LSM510 confocal microscope.

dpp is expressed in the dorsal and ventral margins of the posterior region of the optic lobe, adjacent to the cells expressing wg, which induces dpp expression(Kaphingst and Kunes, 1994; Song et al., 2000)(Fig. 1B). Glial cells in the lamina target region arise from these regions and migrate into the lamina target region as they contact R axons(Perez and Steller, 1996; Dearborn and Kunes, 2004)(Fig. 1B,C). Axons from R1-R6 neurons stop between two rows of glial cell layers, the epithelial and marginal layers, and form the lamina plexus. The third row of glial cells, the medulla glia, is located just beneath the marginal glia(Fig. 1C). The homeodomain protein REPO is expressed in these glial cells(Xiong et al., 1994; Halter et al., 1995).

We compared the expression pattern of dpp-lacZ, an enhancer-trap allele of dpp (Tabata and Kornberg, 1994), with the expression pattern of REPO(Fig. 1D-F). At a stage prior to glia differentiation and migration, expression of the dpp reporter is detected in the dorsal and ventral margins of the lamina target region(Fig. 1D). dppcontinues to be expressed at the margins of the lamina target region throughout the third larval instar (Fig. 1E,F).

The role of DPP signaling in optic lobe development was examined by generating clones mutant for Medea, which encodes a transcription factor downstream of TGFβ-mediated signaling(Das et al., 1998; Wisotzkey et al., 1998; Hudson et al., 1998; Muller et al., 2003). We observed defects in R axon projections and lamina morphology in two independent alleles of Medea mutants, both of which are considered to be strong alleles (Wisotzkey et al.,1998) (Fig. 2B-E; n=8). Such defects were observed only when large clones were generated in the posterior-dorsal or ventral domains, which presumably include glial cell progenitors (Fig. 2B,D).

When clones were induced in other regions (i.e. the OPC, lamina or medulla), R axon targeting defects were not observed. In addition, neither CUT expressed in the medulla (Fig. 2G,H; n=19) nor DAC expressed in the lamina(Fig. 2I,J; n>20)were affected in these clones, suggesting that DPP signaling is not required for the development of these cell types. Furthermore, Cyclin A accumulation in G2 to early M phase of the cells in the outer proliferation center (OPC) and the first row of the lamina neurons(Nakato et al., 2002) was not affected (Fig. 2L,M; n=16), suggesting that DPP signaling is not required for cell cycle progression in the OPC and lamina neuron precursors.

These results suggest that Medea activity is required in the posterior marginal cells for correct R axon targeting. Because this region includes glial cell progenitors and lamina glia have been shown to play an important role in R axon targeting, we assume that these defects could be due to a failure in some aspect of lamina glia development.

To further address the direct role for DPP signaling in optic lobe development, we analysed the antagonizing effects of DPP signaling by targeted expression of the negative regulator Daughters against dpp(Dad) (Tsuneizumi et al.,1997). We overexpressed Dad in various domains within the optic lobe using several Gal4 drivers. Defects in the pattern of R axon projection were observed when optomotor-blind (omb; bifid - FlyBase)-Gal4 was used as a driver(Fig. 3B,F). The penetrance of the phenotype was about 40% (n=41). The expression pattern of Dachshund (DAC), a marker for the differentiating lamina neurons, was also impaired, and the lamina failed to form the regular crescent shape(Fig. 3C,G). omb is expressed in large posterior domains, including in glial cell progenitors and in mature glia in the target region (Huang and Kunes, 1998; Dearborn and Kunes, 2004) (Fig. 3A). Together, these data suggest that DPP signaling is required in the glial cell lineage for the differentiation/migration of lamina glial cells, and DPP loss causes the defect of R axon targeting.

Although omb is also expressed in the retinal basal glia, and in the dorsal and ventral margins of the eye disc, we did not observe significant defects in R axon projections when DPP signaling was blocked in these cells. Lamina glia serve as intermediate targets of R1-R6 axons and thus are required for the establishment of the correct R axon projection pattern(Poeck et al., 2001). Therefore, we examined the structure of the lamina glia layer and the morphology of the cells to investigate whether they are affected by Dad overexpression. We found that the lamina glia layers were indeed defective in these brains. Epithelial, marginal and medulla glia layers were not clearly distinguishable, cells were not regularly spaced(Fig. 3K) and there was a defect in the cell shape. Although epithelial and marginal glia have a cuboidal shape with many small processes(Poeck et al., 2001)(Fig. 3J,J′), the glial cells failed to develop their characteristic cuboidal shape and look irregular in these mutant brains (Fig. 3K,K′). Taken together, these results suggest that DPP signaling controls the development and formation of the regular array of lamina glial cells.

The above results show that blocking DPP signaling can disrupt the formation of the lamina glia layers. Defects in either differentiation or cell migration of the glial cells can cause these phenotypes. It has been reported that when migration of the lamina glia is defective, REPO-positive glia get stuck in the progenitor domain (Poeck et al., 2001; Suh et al.,2002). However, we did not observe such `stuck' cells when DPP signaling was blocked (data not shown). These results led us assume that DPP signaling is required for differentiation rather than for migration of the glial cells.

To further distinguish between these possibilities, first we followed the process by which the progenitor cells differentiate into lamina glia. The gcm gene is expressed in the glial cell precursors and acts as a genetic switch to promote glial cell fate by directly activating the expression of glia-specific genes in the embryo(Hosoya et al., 1995; Jones et al., 1995; Vincent et al., 1996; Akiyama et al., 1996; Egger et al., 2002) (reviewed by Granderath and Klambt,1999; Van De Bor and Giangrande, 2002). In the developing optic lobe, the onset of gcm expression marks the onset of lamina glia differentiation(Dearborn and Kunes, 2004). During migration towards the neuropile, gcm-positive glia start to express REPO, which is known to be downstream of gcm(Dearborn and Kunes, 2004).

gcm expression was analysed using an enhancer-trap line, gcm-lacZ^rA87, which has been shown to faithfully reflect the expression pattern of endogenous gcm in the embryo(Jones et al., 1995; Vincent et al., 1996).β-GAL signal was observed in the glia lineage, including in precursors at the margin of lamina target region, which is virtually the same pattern as that described by Dearborn and Kunes(Dearborn and Kunes, 2004) when using an anti-GCM antibody (Fig. 4A,A′). β-GAL signal was detected in mature epithelial and marginal glia (Fig. 4C),which were also positive for omb-Gal4(Fig. 4B). When cells migrate into the targeting region, REPO expression is observed(Fig. 4A,A′). In addition, gcm expression is seen in the developing lamina neurons(Fig. 4D). With these markers,lamina glia differentiation can be monitored right from the stage when cell fate is specified.

To address whether gcm expression depends on R axon input or factors in the optic lobe, we analysed gcm expression in brains mutant for sine oculis (so). In this mutant, eye formation is defective (indeed, the eyes are almost completely absent) and, thus, there is no R axon projection into the optic lobe(Kunes et al., 1993; Pellez and Steller, 1996). gcm expression was observed together with REPO in this mutant background (Fig. 4E,F). In addition, glial cell migration occurred to some extent,although these glial cells fail to form the normal three-layer structures(Fig. 4G). This indicates that R axon innervations are necessary for the lamina glia to efficiently migrate and form regular three-layer structures, a finding that is consistent with previous reports (Perez and Steller,1996; Huang and Kunes,1998; Suh et al.,2002; Dearborn and Kunes,2004). In addition, glia differentiation and migration also depends upon a factor within the brain.

To address whether this factor in the brain is indeed dpp, we analysed the brains of dpp loss-of-function mutants(Kaphingst and Kunes, 1994)(dpp^d6/dpp^d12 carrying gcm-lacZ). A severe reduction in the number of gcm-lacZexpressing cells was observed in these brains(Fig. 4I). These cells co-express REPO, indicating that they are glial cell populations(Fig. 4J). There is no R axon input in brains with this allelic combination of dpp, as this combination disrupts the normal contribution of dpp to the eye formation (Wiersdorff et al.,1996; Pignoni and Zipursky,1997; Curtiss and Mlodzik,2000). However, the decrease of gcm- and REPO-expressing cells was more prominent than that in the so¹ mutant,which also lacks R axon innervations (Fig. 4E,F,I,J). Together, these data suggest that the further reduction in the number of gcm-expressing cells is due to a reduction of dpp activity. In addition, gcm expression in these mutant brains was not completely lost, possibly because they are disc-specific regulatory mutants that might still have residual dpp activity.

As previously mentioned, Hh induces the differentiation of lamina neurons. DAC-positive lamina neurons fail to differentiate in hh^bar3, the eye-specific allele of hh(Heberlein et al., 1993; Huang and Kunes, 1998)(Fig. 4K,L). We observed that gcm expression and REPO-positive glia differentiation and migration were still observed in hh^bar3 mutant brains(Fig, 4K,M), which is consistent with results reported by Huang and Kunes(Huang and Kunes, 1998), who observed REPO expression in hh^bar3 mutant brains. Conversely, gcm expression in the lamina neuron appeared to be significantly reduced (Fig. 4M). This result suggests that gcm expression in the lamina depends on Hh, and that its expression is controlled independently in these two cell populations.

We also observed that there are gcm-positive cells that are neither REPO- nor DAC-negative, which are not present in the target region of the wild-type brain. These cells could be glia-type cells that remain in an immature state, as it has been shown that R axon innervation plays a role in the terminal differentiation of glial cells(Perez and Steller, 1996). It is also possible that these are fated to the lamina but fail to differentiate because of a lack of R axon innervations.

To further address the requirement of DPP signaling in lamina glia development, gcm expression in Medea mutant clones was examined. When large clones were generated in the posterior domain, defects in gcm-lacZ expression were observed(Fig. 5, n=10). In the specimen shown in Fig. 5A-C,defects in the expression of gcm in the lamina were observed as well(Fig. 5B). In such cases, DAC expression was also defective and appeared truncated near the edge of the mutant clone (Fig. 5C). Conversely, Medea mutant clones generated in the OPC and lamina do not affect DAC expression (Fig. 2I,J) or gcm-lacZ expression (not shown). Therefore, we conclude that it is likely that this defect in the lamina is indirectly caused by defects in the development of lamina glial cells, rather than by a direct effect on lamina development. In the specimen shown in Fig. 5D,E, only a small number of wild-type cells expressed gcm-lacZ in the migratory route from the presumptive progenitor zone, and no mutant cells did. These results suggest that DPP signaling is required autonomously in these cells for the expression of gcm.

When small clones were generated inside the lamina target region, the mutant cells expressed gcm-lacZ(Fig. 5F). This suggests that once the glial cells migrate into the lamina target region, the DPP signal is no longer required.

The above results suggest that dpp is required for gcmexpression and for differentiation of the lamina glia lineage. We next investigated whether ectopic activation of DPP signaling induces glia differentiation. To this end, we expressed UAS-tkv^Q253D, a constitutively active form of tkv(Wiersdorff et al., 1996),with an omb-Gal4 driver. An ectopic gcm expression domain was induced in these brains (Fig. 6F; n=15), and REPO-positive cells were induced inside of the ectopic gcm domain (Fig. 6H), indicating that overactivation of DPP signaling can induce ectopic glia differentiation. Thus, cells in the omb domain are able to express gcm in response to the dpp signal, and subsequently develop into glial cells when they receive higher-than-endogenous levels of the signal.

By contrast, when tkv^Q253D was expressed in the OPC,neither gcm nor REPO expression was induced (data not shown). This suggests that only cells in the posterior domain of the optic lobe are competent to become glia in response to DPP signaling.

From the above results, we conclude that dpp controls gcmexpression in lamina glia progenitors. Given the role of gcm in the embryo for the determination of glial cell fate, it is likely that gcm controls the differentiation of lamina glial cells as well. The gcm gene exists as a cluster with the homologous gene gcm2,which may act redundantly in some contexts(Kammerer and Giangrande,2001). We made use of the UAS-glide^DNconstruct, which encodes a chimeric protein that combines (1) a DNA-binding domain that is highly conserved between gcm and gcm2, and(2) the repressor domain of Engrailed, which renders the chimeric protein capable of blocking the function of both gcm and gcm2(Soustelle et al., 2004). When glide^DN was induced via omb-Gal4, R axon projection defects associated with defects in lamina morphology were observed(Fig. 7F,G; n>20). In these brains, very few REPO-positive cells can be seen in the presumptive lamina target region (Fig. 7H),indicating that the differentiation of the lamina glia is compromised. This defect was not observed with UAS-glide^N7-4DN, which carries a point mutation in the DNA-binding domain that makes the protein unable to bind DNA (Soustelle et al.,2004). These data confirm that the phenotype observed with the UAS-glide^DN is the result of the specific blocking of gcm function rather than a non-specific effect of the chimeric protein. Taken together, the results suggest that gcm is indeed required for lamina glia differentiation.

As mentioned, gcm is also expressed in the lamina precursors/neurons. To investigate the function of gcm in the lamina,we expressed UAS-glide^DN in the lamina using NP6099-Gal4 (Hayashi et al.,2002) (D.U., S.M. and T.T., unpublished)(Fig. 7I-K). A severe reduction in the number of DAC-positive lamina cells was observed in this situation(Fig. 7L,M; n=12). In such specimens, R-axon projection to the presumptive lamina target field is still observed, although the pattern is not completely normal(Fig. 7M). glide^N7-4DN did not cause such defects, again confirming the specificity of the phenotype. These data suggest that gcmactivity is also required for lamina neuron differentiation.

The establishment of the correct R axon projection patterns is crucial for the construction of the visual system in Drosophila. The complex set of events that controls axon projection includes interplay between axons and their targets: the R axons need to be positioned for establishing contacts with their synaptic targets and, conversely, the lamina relies on the R axons as a source of factors required for their own development(Huang and Kunes, 1996; Huang et al., 1998). The lamina glia play a crucial role in this process by serving as intermediate targets for R1-R6 axons (Poeck et al.,2001). Here, we have shown that dpp, which is expressed in the dorsal and ventral marginal regions in developing lamina target region,plays a key role in the development of the optic lobe by inducing the expression of gcm to initiate a genetic program for lamina glia differentiation (summarized in Fig. 8).

wg at the posterior-most domain induces the expression of dpp and omb (Huang and Kunes, 1998; Song et al.,2000). Some wg-expressing cells extend projections towards the lamina target region. These cells extend scaffold axons along which the lamina glia migrate (Dearborn and Kunes, 2004). Thus, it was possible that the wg signal is involved in the migration and/or differentiation of lamina glia. However,partial elimination of Wg activity with a wg^ts allele does not cause a specific defect in glia migration(Dearborn and Kunes, 2004). Therefore, wg may play a role in organizing domains in the visual cortex by activating/repressing various genes, rather than contributing to the generation of specific cell types (Song et al., 2000).

We have shown that Medea is required for lamina glia development. Medea encodes a co-SMAD and mediates a range of DPP/BMP/TGFβsignaling events (reviewed by Mehler et al., 1997). In addition to dpp, four related genes -glass bottom boat (gbb), screw, activin and activin2 - have been identified in Drosophila. GBB signals through TKV/Saxophone (SAX) and Wishful Thinking (WIT) type I and type II receptors, respectively (MacCabe et al., 2003). Activin uses Baboon as a type I receptor, and Punt and WIT as type II receptors(Zheng et al., 2003). We examined brains mutant for gbb and wit, but failed to observe any defects in lamina glia development (data not shown). Together with the results presented here, then, we conclude that it is highly likely that dpp is the ligand responsible for lamina glia development. However,we cannot exclude the possibility that one or more of the DPP-related ligands acts redundantly in this process.

In the embryo, gcm initiates the specification of glial cells from neural cells of various lineages. gcm expression is strictly controlled to ensure the correct separation of glial versus neuronal cell fate. Analysis of the cis-regulatory elements of gcmsuggests that gcm expression depends on multiple regulatory elements to allow the control of lineage-specific transcription and autoregulation(reviewed by Jones, 2004). Our analysis suggests that a different situation exists in the optic lobe; gcm is expressed in the glia and the lamina neuronal cells, and is required for the differentiation of these cell types. In addition,differentiation is controlled differently in the lamina and in the glia. In the lamina, gcm expression seems to be controlled by hh, and in the glia, by dpp. These results suggest that gcm is controlled and functioning in a different manner in the optic lobe. Uncovering the mechanisms of the control and function of gcm would probably prove an intriguing focus for future research.

DPP and its vertebrate homolog BMP play crucial roles in many aspects of development by controlling patterning, cell growth and differentiation. Our analysis reveals a role for DPP signaling in lamina glia differentiation in the Drosophila visual system. DPP has also been reported to function in several aspects of visual center development; for instance, DPP signaling has been shown to be involved in the proliferation and migration of the subretinal glia in eye disc development, which plays an important role in the R axon navigation (Rangarajan et al.,2001; Hummel et al.,2002). In addition, Kaphingst and Kunes(Kaphingst and Kunes, 1994)reported defects in the medulla neuropile in dpp mutant animals,suggesting a role for dpp in neuronal fate specification. Furthermore, tkv is expressed in lamina precursor cells just ahead of the lamina furrow, where these cells meet R axons and start to differentiate(H. Tanimoto and T.T., unpublished). Although this possibility is one of the things that prompted us to look for a role of DPP signaling in lamina development, we failed to uncover any defects when Mad or Medea clones were generated in the OPC or the lamina. Moreover, dpp appears to be expressed in the inner proliferation center (IPC),which will form the lobula, in addition to its expression in the dorsal and ventral marginal domains. Thus, dpp may be required for some aspects of lobula development. Unfortunately, this cannot be easily addressed at this moment because of a lack of appropriate markers. Further study of the requirements for dpp in the lamina, the medulla, the lobula and other cell types could lead to a more comprehensive understanding of how DPP signaling controls differentiation and other events during development of the visual system.

We thank the Bloomington Stock Center and the Kyoto Stock Center for fly stocks. We would also like to thank S. B. Selleck for Cyclin A antibodies, and the Genetic Studies Hybridoma Bank for antibodies. We are grateful to all the members of Tabata laboratory for helpful discussions during the course of the work and, in particular, we thank Yuki Takei for valuable comments on the manuscript. We thank Yuko Maeyama for her excellent technical help. This work was supported by Grants-in-Aid from the Ministry of Education, Science and Culture of Japan, and by Toray Science Foundation.