Drosophila Activin-β and the Activin-like product Dawdle function redundantly to regulate proliferation in the larval brain

The TGF-β family of growth and differentiation factors controls a wide range of developmental processes including early axis specification, cell proliferation, apoptosis, differentiation and stem cell propagation(Massague et al., 2000). These factors bind to a heteromeric receptor complex containing type I and II transmembrane serine/threonine kinases. Upon ligand binding, the type I receptor kinase is activated by the type II receptor and phosphorylates members of the R-Smad family of transcription factors. TGF-β signaling pathways are broadly divided into either the BMP or Activin/TGF-βbranches, based on which Smad transcription factors they phosphorylate(Feng and Derynck, 2005). The BMP family of ligands signal through R-Smads1/5/8 whereas the Activin/TGF-β branch utilizes R-Smads2/3. Phosphorylation of R-Smads facilitates their association with a common Co-Smad and retention of the R-Smad-Co-Smad complex in the nucleus, where it regulates transcription of target genes in association with a wide variety of co-factors(Miyazono et al., 2006).

Drosophila employs both BMP and Activin-like signaling pathways to regulate numerous developmental processes. Among the seven TGF-β type ligands, Decapentaplegic (Dpp), Screw (Scw), and Glass bottom boat (Gbb) are members of the BMP family, whereas Activin-β (Actβ) and Dawdle (Daw)(Parker et al., 2004; Serpe and O'Connor, 2006),fall within the Activin/TGF-β branch. Two novel ligands, Maverick (Mav)and Myoglianin, are sufficiently diverged to preclude clear assignment to a particular ligand subfamily (Parker et al., 2004). Both the BMP and Activin/TGF-β pathways employ Punt as a common type II receptor, and pathway specificity is provided by the type I receptors (Brummel et al.,1999; Brummel et al.,1994; Das et al.,1999; Letsou et al.,1995; Ruberte et al.,1995; Serpe and O'Connor,2006; Zheng et al.,2003). Tkv and Sax bind BMP-type ligands and phosphorylate Mad,whereas Babo binds Activin-like ligands and signals through Smad2 (also known as Smox - FlyBase) (Das et al.,1999; Serpe and O'Connor,2006; Shimmi et al.,2005a; Shimmi et al.,2005b; Zheng et al.,2003).

Parsing out the functional relationship between individual TGF-βligands and receptors and particular developmental processes is a difficult process, especially in vertebrates, in which the numbers of ligands and receptors are large. In general, this process involves biochemically matching a particular ligand with one or more signaling receptors and comparing the phenotypes produced by knockdown of the different signaling components. In Drosophila, the contributions of BMP pathway components to numerous developmental processes, including cell-fate specification, imaginal-disc growth and patterning and synapse development, have been well studied(reviewed by Parker et al.,2004). By contrast, the Activin pathway is less well characterized. Unlike mutations in the BMP pathway, mutations in Activin signaling components have not been shown to directly affect differentiation but instead appear to primarily regulate neuronal wiring and proliferation. For example, clonal analysis of either babo or Smad2 mutants has shown that this pathway regulates mushroom body remodeling during metamorphosis (Zheng et al.,2003), morphogenesis of ellipsoid body neurons in the adult(Zheng et al., 2006) and motor axon guidance in the embryo (Parker et al., 2006; Serpe and O'Connor,2006). Regarding proliferation, loss of both maternal and zygotic babo leads to small brains and small but properly patterned wings(Brummel et al., 1999),indicating a potential role in regulating proliferation of larval mitotic tissue.

Analysis of Drosophila Actβ and Daw ligand loss-of-function phenotypes indicates that these two ligands probably regulate separate aspects of neuronal wiring, as dominant-negative and RNAi constructs that reduce the activity of actβ phenocopy the mushroom body remodeling defects seen in babo and Smad2 mutants(Zheng et al., 2003), whereas null mutants of daw phenocopy the babo and Smad2mutant motor axon guidance defects (Parker et al., 2006; Serpe and O'Connor, 2006). In neither case, however, was an obvious proliferation defect reported for loss of either ligand.

In this paper, we investigate the role of the DrosophilaBabo/Smad2 pathway in larval brain development. We show that mutations in babo and Smad2 result in small brains with altered innervation of photoreceptor axons within the lamina and medulla. In contrast to the wiring defects described above, however, we demonstrate that these aberrations are not caused by defects in photoreceptor innervation or changes in cell fate of target neurons within the brain. Instead, they result primarily from reduced proliferation within the optic lobe and central brain leading to a reduction in the size of the photoreceptor target field. We further demonstrate that the Babo receptor is required in neuroblasts and that the ligands Actβ and Daw function redundantly to control proliferation in the brain. These results suggest that while the two DrosophilaActivin-like ligands have at least partially independent roles in regulating neuronal wiring, they have largely redundant roles in regulating proliferation within the brain.

The strong loss-of-function babo alleles, babo³²,babo⁵² and babo²⁶, have been described previously (Brummel et al.,1999), as has the strong Smad2^mb388 allele(Zheng et al., 2003). Previously, we reported that babo³² homozygotes have very low levels of BrdU incorporation in third-instar brains(Brummel et al., 1999). However, we have since found that the babo³² chromosome contains a second hit, which when homozygous leads to developmental delay and severely reduced cell proliferation (M.B.O., unpublished). Therefore, only heterozygous combinations of babo alelles are used in these studies. The daw^ex32 and daw^ex11 are strong and moderate loss-of-function alleles, respectively(Serpe and O'Connor, 2006),whereas daw⁴ is a presumed null mutation(Parker et al., 2006). The Cyclin A^c8LR1 allele is a strong amorph(Sigrist and Lehner, 1997) and was obtained from the Bloomington Stock Center. For genetic rescue experiments, babo mutant lines carrying various Gal4 drivers were crossed to different babo mutant alleles carrying UAS transgenes. The 1407>Gal4 driver (Luo et al.,1994) was obtained from K.-F. Fischbach and the worniu>Gal4 has been previously described(Albertson et al., 2004). The da>Gal4, elav>Gal4, and ey>Gal4 lines were obtained from the Bloomington Stock Center. Mutant babo alleles with different Gal4 drivers and UAS transgenes were balanced by the Cy^O::Tm6,Tb compound chromosome. Null babophotoreceptor clones were generated by crossing FRTG13babo⁵²/Cy^O::Tm6,Tb flies to FRTG13uasCD8GFP; eyGal4-uasFlp/Cy^O::Tm6,Tb flies. GFP-positive MARCM clones were induced by heat shock at 36°C for 1 hour of early second-instar larvae from the cross of elavGal4hsFlp;FRTG13tubGal80/Cy^O and FRTG13babo^w224UASmCD8GFP/Cy^O. To identify daw, actβ double mutants, non-GFP, y larvae were obtained from the stock daw/Cy^O, actin-GFP; actβ^ed80/Dp1:4 (y+, spa). All crosses were done at 25°C.

The daw promoter-Gal4 line was generated by cloning a 9 kb PCR fragment containing the first intron and upstream sequences using the following primer pair 5′-CTGAGCCCCTACGTCTGTATGATATG-3′ and antisense 5′-GATCTTCTGGATCGCCTTTGGTTTCA-3′ into the pPelicanGal4 plasmid, which is derived from pPelican(Barolo et al., 2000). Transgenic flies were generated by standard injection procedure.

Freshly hatched larvae were collected for 5 hours on apple-agar plates and staged to white prepupal stages (120 hours for yw and 144 hours for babo mutants).

Larval brain lobes, together with imaginal discs, were dissected in PBS and fixed in 3.7% formaldehyde in PBS for 1 hour at room temperature, and were washed by PBS plus 0.1% Triton X-100 (PBT). Antibodies purchased from the Developmental Studies Hybridoma Bank (DSHB) include mouse 24B10 (1/100 dilution), rat anti-Elav (7E8A10, 1/400), mouse anti-Dachshund (mAbdac2-3,1/50), mouse anti-Robo 13C9, (1/50), mouse anti-Repo 8D12, (1/50 dilution),mouse anti-BrdU G3G4 (1/100), anti-Arm 7A1 (1/500), mouse anti-Pros MR1A(1/100) and mouse anti-Discs large (1/1000). Rabbit anti-phosphorylated histone H3 (pSer¹⁰, H-0412) was from Sigma (1/400). Rabbit anti-Cyclin A antibody was used at 1/500 dilution(Whitfield et al., 1990; Nakato et al., 2002). Rabbit anti-Scrib was diluted at 1:2500 (Albertson et al., 2004), and guinea-pig anti-Mira at 1:500(Lee et al., 2006). Anti-active Caspase-3 antibody was a gift from Idun Pharmaceuticals, and used at 1/2000 dilution. Rat monoclonal N-Cadherin antibody, DUex8, was a gift from L. Zipursky and was used at a 1/50 dilution. Fluorescent conjugated secondary antibodies (Molecular Probes) were used at 1/200 dilution. All primary antibodies were diluted in PBT and incubated with tissue samples at 4°C overnight. Secondary antibodies were typically incubated with tissue samples for 2 hours at room temperature. Confocal images were taken with a Zeiss Axio confocal microscope or a Leica TCS SP2.

Larval brain lobes and attached eye discs were dissected in PBS and transferred to M3 complete medium containing 0.4 mg/ml BrdU (Roche) and incubated at 25°C for 30 minutes before fixation. Fixed tissue was treated in 2 N of HCl at room temperature for 30 minutes before the addition of anti-BrdU monoclonal antibody.

Previous work has implicated the Babo/Smad2 pathway in proper morphogenesis of several larval and adult brain neurons(Zheng et al., 2003; Zheng et al., 2006). We noticed that mutant third-instar larvae also showed variable defects in photoreceptor innervation of the lamina and medulla(Fig. 1A vs B). Differentiated photoreceptor cells are normally organized as ommatidial units, each consisting of eight photoreceptor cells, R1-R8, which send their axons through the optic stalk into developing optic lobes. The axons of R1-R6 terminate and form a neural plexus at the lamina, whereas the axons of R7 and R8 project to the medulla (Clandinin and Zipursky,2002) (Fig. 1A). In babo mutant brains, photoreceptors R1-6 formed a lamina plexus that was very reduced in size (Fig. 1B). In addition, the R7 and R8 axon projections to the medulla were highly perturbed, and their growth cones formed bundles instead of a regular lattice network typical of controls(Fig. 1A vs 1B insets). This defect was not corrected at later stages, as axon targeting was still highly perturbed in 3-day-old babo mutant pupae, and rotation of the medulla and lamina relative to one another did not take place(Fig. 1D vs E). Smad2mutants exhibited similar photoreceptor axon targeting defects, indicating that the babo defect is caused by the loss of the canonical Activin signaling pathway (Fig. 1C).

To further characterize the babo phenotype, we also immunostained babo mutant brain lobes with anti-Dachshund (Dac) antibody to highlight lamina neuron precursor cells, and with anti-Elav to visualize differentiated neurons. In the most severe cases, babo mutants exhibited reduced numbers of lamina cap neurons(Fig. 2A vs B) and loss of Dac-positive cartridge neurons. In larvae with less severe phenotypes, the number of Dac-positive cells in each cartridge was not greatly different from that seen yw controls, but the total number of lamina cartridges in babo mutants is smaller than that in wild type (see Fig. S1C in the supplementary material).

The severe aberrations of R7 and 8 termini in the medulla neuropil, which is composed in part by axons projecting from the lobula cortex, suggested that organization of this target field is probably disrupted. Consistent with this view, we find that babo mutants show a disorganized and smaller medulla neuropil compared with wild type, as revealed by anti-Robo(Fig. 2C vs D) which concentrates in the medulla axons derived from the lobular cortex neurons(Tayler et al., 2004). In addition, anti-N-Cadherin, which stains both photoreceptors and the medulla neuropil (Lee et al., 2001),also reveals a disruption in the medulla neuropil formation(Fig. 2G,I vs H-J).

Glial cells have been implicated as intermediate targets for photoreceptor axon targeting (Clandinin and Zipursky,2002). Therefore, we examined babo and Smad2mutants with the glia-specific Repo antibody and compared them to ywcontrols. White prepupae of wild-type larvae showed three layers of well-aligned glial cells at the lamina and evenly distributed glia at the medulla cortex (Fig. 2E). By contrast, babo and Smad2 mutants exhibited a reduced number of glia, with some misalignment of the glial cells within both the lamina and medulla (Fig. 1C and Fig. 2F).

Although based on only a limited number of markers, our data suggest that differentiation of neurons and glia is not grossly disrupted in the optic lobe of babo and Smad2 mutants, raising the possibility that the observed retinal axon targeting defects may result from the production of fewer lamina and medulla progenitor cells within the optic lobe (see below).

As Drosophila retinal axons have been shown to regulate proliferation in the lamina via delivery of Hedgehog and the EGF-like ligand Spitz to the target field (Huang and Kunes, 1996; Huang and Kunes,1998; Huang et al.,1998), and because Drosophila Activin is expressed in photoreceptors (see Fig. S1A in the supplementary material), we wished to address whether babo function is required in photoreceptor axons or the brain for proper photoreceptor axon targeting and optic lobe development.

The babo gene is alternatively spliced, producing two protein isoforms, Babo_a and Babo_b, which differ in their extracellular ligand-binding domain (Wrana et al., 1994). The babo_a isoform has been specifically implicated in mushroom body remodeling(Zheng et al., 2003), while ectopic expression of either the Babo_a or Babo_b isoform rescues dorsal neuron morphogenesis defects of babo mutants. To examine the isoform requirements for proper optic lobe development and axon targeting, we expressed babo_a and babo_b individually or in combination using the ubiquitous daughterless Gal4 driver (da>Gal4). Expression of the babo_a isoform alone rescued neither the photoreceptor axon targeting nor brain size defects (Fig. 3B). By contrast, ubiquitous expression of babo_b alone (Fig. 3A) or babo_a and babo_btogether (data not shown) not only rescued photoreceptor axon targeting and brain lobe defects, but also restored viability. These rescue data suggest that the babo_b isoform alone, at least when overexpressed,is sufficient to provide proper signaling activity for normal optic lobe development, axon targeting and viability.

To address tissue- and cell-type-specific requirements for babofunction, a number of different Gal4 drivers were employed. Using the eye-specific driver eyeless Gal4 (ey>Gal4), we found that expression of neither babo_a, babo_b nor a combination of both in eye discs was able to rescue the babo mutant phenotype (Fig. 3C and data not shown). This suggested that babo function in the photoreceptors is not sufficient for directing normal photoreceptor axon targeting. We also used the FLP/FRT system to generate babo mutant clones in an otherwise babo heterozygous background. Large babo-mutant photoreceptor cell clones were found to project their axons normally into babo heterozygous brain lobes(Fig. 3E), providing further support for the argument that babo function in photoreceptor cells is dispensable for proper axon targeting and optic lobe development.

As glial cells provide both guidance cues and trophic factors for optic lobe development (Hidalgo et al.,2006), and because their numbers are reduced in babomutants, we asked if selective expression of babo in glial cells was able to suppress any aspect of the babo mutant phenotype. As shown in Fig. 3D, expression of babo_a and babo_b in glial cells using the glia-specific driver repo>Gal4 did not alter the babomutant phenotype. By contrast, expression of babo_a and babo_b using the 1407>Gal4 driver(Luo et al., 1994), which is expressed in both neuroblasts and many differentiating neurons of the brain but not in the eye disc (Fig. 3F), was able to rescue both brain size and the axon-targeting defect (Fig. 4G). In addition,the worniu>Gal4 driver that is expressed primarily in neuroblasts(Albertson et al., 2004) and ganglion mother cells (GMCs), but not the eye disc(Fig. 3H and see Fig. S3 in the supplementary material), was also able to rescue the babo mutant optic lobe phenotype (Fig. 3I),and some of these mutant animals survived to adults. Lastly, we found that overexpression of the EcR-1B receptor, the only known target of Actβ/Babo_a signaling in the brain lobes, could not rescue the axon-targeting defects (data not shown) in contrast to its ability to suppress neuronal remodeling defects (Zheng et al.,2003; Zheng et al.,2006).

Taken together, these results suggest that, unlike the case for mushroom body remodeling and dorsal neuron morphogenesis or motor axon guidance, the failure of photoreceptors to properly innervate the lamina and medulla is not caused by an inability of these neurons to receive Babo/Smad2 signals or target cells to differentiate properly. Instead, the defect in innervation may arise from the production of too few target field cells, perhaps as a result of proliferation defects within the optic lobe.

babo mutant brain lobes are 25-40% smaller than baboheterozygous brain lobes throughout third-instar life(Fig. 4A). Previously, we have shown that babo mutant brain lobes do not exhibit an increase in apoptosis, suggesting that the small size of babo brains probably results from reduced proliferation and not cell death(Brummel et al., 1999). However, it is not clear how proliferation is affected in babo mutant brains. Because the optic lobes comprise the majority of brain lobe tissue,and have been well characterized (Egger et al., 2007), we focus on the growth and proliferation of the optic lobe.

In wild-type white prepupae, optic lobes make up more than half of the brain lobe. At late stages, the optic lobe epithelium produces many medulla neuroblasts and laminar precursor cells (LPCs). These neuroepithelial cells of the outer proliferation center (OPC) and inner proliferation center (IPC), as well as medulla neuroblasts and LPCs, undergo rapid proliferation to produce the cells of the lamina and medulla cortices(Fig. 4B)(Egger et al., 2007). As noted above, in babo mutants the optic lobe is much smaller than in wild-type controls. We find that the overall organization of the OPC and IPC epithelia is normal and that they retain their ability to produce neuroblasts and LPCs. However, there is approximately a 50% decrease in the number of medulla neuroblasts (Fig. 4D),a decrease in the number of ganglion mother cells produced from these medulla neuroblasts, and a subsequent decrease in the number of Elav⁺maturing neurons within the medulla cortex(Fig. 4B,C). Furthermore, there is a decrease in the number of LPCs and thus a decrease in the number of lamina cartridges and laminar neurons (Fig. 2B and Fig. 4C, and see Fig. S2B,C in the supplementary material). Similar results are observed for central brain neuroblasts, where we observed a slight reduction in neuroblast numbers (see Fig. S2A in the supplementary material). We observed no evidence for an increase in apoptosis, as revealed by staining for Caspase-3 (also known as Decay - FlyBase)(Fig. 5N-P), confirming previous results (Brummel et al.,1999). We conclude that babo function is required for generating the normal number of medulla neuroblasts and laminar precursor cells in the optic lobe.

To further address whether proliferation rates are altered in babomutant optic lobes, we generated GFP-positive marked baboloss-of-function clones using the MARCM system(Lee and Luo, 2001). As there are different patterns of mitotic clones within the developing optic lobes(data not shown), we sought to compare only the babo mutant and wild-type clones at similar positions within the optic lobe (example clones in Fig. 4E vs F) and central brain(Fig. 4H vs I). This analysis revealed that babo mutant clones in the optic centers and central brain contained 30-50% fewer cells than did wild-type control clones, which agrees with the overall reduction of brain size and lamina cartridge size. We conclude that babo mutants have reduced proliferation of optic lobe and central brain neuroblasts.

We next investigated whether proliferation was affected by using BrdU and phospho-histone H3 (P-H3) labeling to identify cells in the S and M phase of the cell cycle, respectively. As shown in Fig. 5, wandering third-instar larvae of babo^32/52 and Smad2^mb388mutants exhibited many S-phase cells in both the OPC and IPC, as revealed by BrdU incorporation, although the sizes of these zones were reduced compared with wild type (Fig. 5A vs B,C). By contrast, the number of cells in M phase was significantly reduced compared with wild type(Fig. 5D vs E,F), resulting in an alteration of the M:S ratio in babo mutant brains(Fig. 5G). This suggests that loss of babo/Smad2 signaling delays the transition from S-to-M phase of the cell cycle within the optic lobe. By contrast, babo mutant eye discs did not show a significant decrease in the size of the disc or number of cells in M phase, although we did observe ectopic M-phase cells within the morphogenetic furrow (see Fig. S1A vs B in the supplementary material), similar to the previously reported dppmutant phenotype (Horsfield et al.,1998; Penton et al.,1997).

We next examined the levels of different Cyclins in babo mutant clones and fully mutant brains. We found no difference in the levels of Cyclin B and E (data not shown); however, Cyclin A levels were enhanced in both babo clones (Fig. 5J-L) and fully mutant brains(Fig. 5H,I). Ectopic expression of Cyclin A has been previously shown to accelerate the G1/S transition as well as delay cells from exiting M phase(Lehner and O'Farrell, 1989; Lehner et al., 1991; Sprenger et al., 1997). Thus,elevated Cyclin A in the babo mutants could lead to the observed decrease in M-phase cells. Consistent with this model, heterozygosity for the Cyclin A^c8RL1 mutation substantially suppressed the babo mutant phenotype (Fig. 5M). We conclude that loss of Babo/Smad2 signaling leads to elevated Cyclin A levels, which contribute to the optic lobe proliferation defects seen in babo mutants.

Previous work has demonstrated that both Drosophila Activin-βand the Activin-like protein Dawdle signal through Activin type I receptor Babo to Smad2 (Zheng et al.,2003; Parker et al.,2006; Serpe and O'Connor,2006), suggesting that one or both are likely candidates for regulating optic lobe development. In situ hybridization studies revealed that both are expressed in the developing optic lobes(Fig. 6A,B). In addition, daw is also expressed in many glial cells, including surface glia within the optic lobe, and Actβ is expressed in mushroom body neurons(Fig. 6C)(Serpe and O'Connor, 2006)(and data not shown).

To functionally characterize the role of the actβ and daw genes in larval optic lobe development, an EMS-induced actβ^ed80 mutant was identified in a screen for fourth chromosome lethal mutations (S.H., L.P. and J.L., unpublished). This mutant line carries a stop codon (W 711 stop) within the prodomain N-terminal before the maturation cleavage site and is a presumed null allele. A large fraction of the actβ^ed80 homozygous mutants survived to the pharate adult stage and a small proportion (<2%)eclosed. The escaper flies could not move their wings, walked very slowly and did not establish a breeding stock. As the mutation is rescued by expression of a UAS-actβ transgene, we conclude that there are no other lethal mutations on the actβ^ed80 chromosome. As shown in Fig. 6F, these mutants exhibited normal optic lobe size and photoreceptor axon projections.

We also examined several daw mutant lines for optic lobe morphology. We observed no photoreceptor axon-targeting defects in daw³/daw⁴ heteroallelic mutant larvae or the daw^ex11 homozygous larvae (data not shown). However, in a low percentage (4%, n=50) of daw^ex32 homozygous deletion mutants(Fig. 6D), we did observe photoreceptor axon targeting and optic lobe defects in white prepupae that were reminiscent of those seen in babo mutants(Fig. 6E). The low penetrance of the babo-like phenotypes in daw mutants, and the overlapping expression of actβ and daw in the optic lobes, suggested that these two genes might act redundantly to control growth in the optic lobes. To investigate this issue, we examined the phenotypes of various daw, actβ double mutant combinations. In the daw^ex11/daw^ex11,actβ^ed80/actβ^ed80double mutants overall brain morphology was relatively normal. However,approximately 20% (n=30) of the mutant larvae exhibited collapsed and bundled R7 and 8 growth cones that were not seen in either single mutant(Fig. 6G). In the stronger daw^ex32/daw^ex32actβ^ed80/actβ^ed80allelic combination, the penetrance of the severe small brain and axon-targeting phenotype seen in daw^ex32 mutants alone was increased from 4 to 50% (n=30). In the daw⁴/daw^ex32,actβ^ed80/actβ^ed80combination, larvae died during late second- or early third-instar stages,precluding us from examining the brain phenotype. We conclude that Actβand Daw function redundantly to control proliferation in the optic centers of the Drosophila larval brain.

In this report, we have demonstrated that an Activin-like signal is essential for proper development of the Drosophila brain. Specifically, this pathway is important for establishing the correct number of central brain and optic lobe neuroblasts and stimulating their proliferation. This reflects a direct requirement for reception of Activin signals in neuroblasts, as re-expression of the Activin receptor specifically in neuroblast lineages of babo mutant animals largely rescues the proliferation defects of the entire brain.

It is not entirely clear how Activin signaling in neuroblast lineages maintains the wild-type number of brain neuroblasts. Approximately 100 neuroblasts per brain lobe are formed during embryogenesis, and most go quiescent at the embryonic/larval transition. From L1-L3 they progressively re-enter the cell cycle to resume their cell lineages. Perhaps the slower cell cycle of Activin pathway mutants inhibits exit from quiescence and promotes premature differentiation. Mutations in trol, which encodes a heparin sulfate proteoglycan Perlecan, prevent neuroblast reactivation and lead to a severe reduction in neuroblast numbers and brain size(Datta, 1995; Park et al., 2003). Many TGF-β ligands bind to heparin sulfate proteoglycans, and thus part of the effect of Activin signaling on brain size might be mediated by Perlecan or other proteoglycans such as the glyipican Dally (see below).

The small brain size is not just caused by a reduced number of neuroblasts,however; babo mutant clones that contained a single neuroblast produced fewer daughter cells in a given time window than did wild-type neuroblasts, presumably due to the increased expression of Cyclin A and the delay in metaphase exit. One additional possibility is that the Activin signal may affect neuroblast temporal identity progression in larval neuroblast lineages, similar to the effect of temporal identity mutations on embryonic neuroblast lineages (Isshiki et al.,2001), leading to the failure to produce early, mid or late subsets of larval lineages. Testing this hypothesis awaits the development of markers for different neurons within optic lobe neuroblast lineages.

The importance of Activin in regulating neuroblast proliferation is reminiscent of the positive role that Activin/Nodal signaling plays in regulating the cell cycle of mouse and human ES cells(James et al., 2005; Ogawa et al., 2006). In those cells, as in Drosophila neuroblasts, Activin/Nodal signaling enhances, but is not absolutely required for, cell proliferation(Ogawa et al., 2006). Another point of potential similarity is that the mES cells endogenously produce an Activin/Nodal signal leading to an autocrine/paracrine regulation of proliferation. While our in situ data are not of sufficient resolution to unambiguously assign expression of actβ and daw to particular cell types, both are expressed in the optic proliferation zones where neuroblasts are highly concentrated. It is also possible that some ligand may be supplied by the innervating photoreceptors. Activin is strongly expressed in R7 and 8 (Ting et al.,2007) and like hh and spitz may provide a tropic signal that simulates proliferation in the target tissue(Huang and Kunes, 1996; Huang et al., 1998).

Lastly, it is interesting to note that Activins are not the only TGF-β-like factors required for proliferation of Drosophilaneuroblasts. The BMP family member Dpp is expressed in four regions in each brain lobe (Kaphingst and Kunes,1994; Yoshida et al.,2005). Two lie in the dorsal and ventral margins of the posterior optic zone neuroepithelium near what has been termed the lamina glial precursor region (Yoshida et al.,2005), whereas the other two smaller zones are more interior at the base of the inner proliferation zone. In the brain, the dpploss-of-function phenotype is remarkably similar to that seen in babomutants (Kaphingst and Kunes,1994; Yoshida et al.,2005). A potential trivial explanation for the similarity in phenotypes might be that Activin signaling is required for dppexpression, or vice versa. However, dpp is still expressed in babo mutants (see Fig. S1C,D in the supplementary material) and daw and actβ are both still expressed in dppmutants, although it is difficult to know in each case whether the levels are equivalent. Thus, both Dpp and Activin signaling appear to be required to stimulate brain neuroblast proliferation.

In addition to regulating proliferation in the brain, Dpp signaling plays a major role in regulating proliferation in other tissues, including the imaginal discs (Burke and Basler,1996; Rogulja and Irvine,2005). Once again, Activins may collaborate with BMPs in regulating proliferation in this tissue. In particular, we note that babo mutants show ectopic P-H3 staining within the morphogenetic furrow of the eye disc, which is also observed in loss-of-function mutants in the Dpp receptor Tkv (Horsfield et al.,1998). Furthermore, babo mutant wing disc clones can grow large in contrast to clones mutant in dpp signaling components(Burke and Basler, 1996),although the overall sizes of babo mutant discs are not affected proportionally as much as is the brain (C.C.Z. and M.B.O., unpublished). Therefore, the way in which BMP and Activin inputs regulate the cell cycle might be different in discs versus the brain, or the two tissues might exhibit different sensitivities to common inputs.

How Activins and BMPs affect the cell cycle is not entirely clear. In the wing disc, Dpp signaling through Tkv/Mad has been shown to promote the G1-S transition (Martin-Castellanos and Edgar,2002). In the brain, babo mutants exhibit a decrease in the M/S ratio, which could be due to a decrease in cells at the G2/M phase of the cell cycle. Consistent with this view, we find that Cyclin A levels are enhanced in babo mutants and that heterozygosity for a Cyclin A mutation suppresses the babo phenotype. This is very similar to that seen in dally mutants, which also affect brain development by causing a delay in the G2-M transition within the outer proliferation centers. Just as we have found for babo mutants, heterozygosity for Cyclin A suppresses the dally cell cycle defect(Nakato et al., 2002).

Given the results described above, one attractive model for how both BMPs and Activins contribute to cell cycle progression is that they regulate the cycle at different points: Activins at G2-M and BMPs at G1-S. Alternatively,as previous work has suggested that Cyclin A probably has roles in regulating both G2-M and G1-S transitions in Drosophila(Lehner and O'Farrell, 1989; Lehner et al., 1991; Sprenger et al., 1997) and as Smads can form heterotrimers (Chacko et al., 2004), it may be that a composite signal composed of a Smad2/Mad/Medea heterotrimer acts at several points in the cell cycle. Interestingly, several potential target genes that are regulated by both Activin and BMP signals in the larval brain have been identified by microarray studies using activated receptors (Yang et al., 2004), but no obvious candidates for genes that might influence proliferation are evident within the list.

Lastly, we note that daw has been demonstrated previously to play a role in motoneuron axon guidance in the embryo, while actβ has been implicated in mushroom body remodeling(Zheng et al., 2003) and more recently in regulating the terminal steps in photoreceptor R8 targeting during pupal stages (Ting et al.,2007). As both ligands are expressed in each of these tissues(P.A.J. and M.B.O., unpublished), it is possible that there may be functional redundancy that limits the severity of the previously observed phenotypes. Consistent with this view, we have recently found that both actβand daw modulate neurotransmission at the neuromuscular junction and that the double mutant phenotypes are more severe than those seen in the single mutants (Yi Ren and M.B.O., unpublished), similar to their redundant function in regulating neuroblast proliferation described here. In conclusion,all available data suggest that Activin signaling plays at least two important roles in Drosophila nervous system development. First, it ensures that the proper numbers of cells are produced in the CNS; and second, it helps establish correct functional connections between neurons and their synaptic partners.

We thank Thomas Neufeld, and Hiroshi Nakato for anti-Cyclin E and anti-Cyclin A antibodies, and also for comments on the manuscript. Anti-N-Cadherin was generously provided by Larry Zipursky. We thank Karl-Friedrich Fischbach and Vanessa J. Auld for fly lines, Mary Jane Shimell for actβ and daw in situ hybridization, and Ying Li for examining dpp mRNA expression in babo mutants. J.Q.B. was supported by a NSF IGERT pre-doctoral training grant. M.B.O. and C.Q.D. are Investigators of the Howard Hughes Medical Institute.