Positioning sensory terminals in the olfactory lobe of Drosophila by Robo signaling

The olfactory lobe of Drosophila shows a remarkable level of precision in its organization. Sensory neurons from the antenna project to the lobe and synapse with second-order interneurons within morphologically distinct structures called glomeruli(Stocker, 1994). The specificity of synapse formation is revealed by the targeting of neurons expressing individual odorant receptor genes to corresponding glomeruli within the lobe (Vosshall et al.,2000; Gao et al.,2000). Each glomerulus is identifiable by its size, shape and position, making developmental and functional analysis easier than in vertebrates (Laissue et al.,1999; Buck, 2000). The projection neurons, which are equivalents of vertebrate mitral/tufted cells, project from individual glomeruli to higher centers in the calyx of the mushroom bodies and lateral horn (Stocker et al., 1990; Jefferis et al.,2001; Jefferis et al.,2002; Marin et al.,2002; Wong et al.,2002). These neurons, together with local interneurons that connect between glomeruli, provide circuitry for information processing,involving integration of sensory input with possible feedback from more central pathways (Fiala et al., 2000; Ng et al., 2002; Wang et al.,2003). An understanding of how this highly stereotyped network arises during development will help elucidate general rules governing circuit design in phylogenetically distinct organisms.

The development of pattern within brain structures can be addressed at two levels. First, what are the mechanisms by which individual sensory neurons and their target interneurons find each other? Second, how are these synapses located reproducibly within the three-dimensional architecture of the neuropil? Studies in the Drosophila embryo have provided valuable insights about how axonal scaffolds and synapses are organized within the axis of the embryonic midline (Dickson,2001; Grunvald and Klein,2002; Ziatic et al.,2003). This instruction involves a combinatorial expression of receptors on the axonal surface, and attractive and repulsive ligands secreted at the midline. It is becoming increasingly clear that similar principles could apply in patterning more complex three-dimensional contours in adult brains (Richards, 2002; Bagri et al., 2002; Hutson and Chien, 2002; Plump et al., 2002).

In the olfactory system of both vertebrates and insects, neurons expressing a given odorant receptor gene project with remarkable precision to specific glomerular sites within the olfactory lobe. In rodents, the specificity of axonal convergence is mediated by the odorant receptors(Wang et al., 1998) in collaboration with cell-surface molecules – the Ephrin A proteins(Cutforth et al., 2003). In Drosophila, the SH2/SH3 adaptor, Dreadlocks (Dock) and serine/threonine kinase Pak, form part of a signaling module that is necessary for the precise guidance of olfactory neurons to their glomerular targets(Ang et al., 2003). Dock and Pak are known to be downstream of receptors that play key roles in axon guidance, such as the Down Syndrome Cell Adhesion Molecule (Dscam)(Schmucker et al., 2000) and the Roundabout (Robo) family (Fan et al.,2003)

Loss of Dscam function affects the development of olfactory projections; neurons enter the lobe but terminate inappropriately at ectopic sites (Hummel et al., 2003). As Dscam can theoretically encode 38,000 isoforms by alternative splicing, this locus provides an intriguing means by which axons can recognize their postsynaptic partners. Interestingly, 68 combinations of transcripts have been detected in olfactory neurons. The dendritic arborization of projection neurons within specific glomeruli has been shown to occur prior to the arrival of the sensory neurons into the lobe(Jefferis et al., 2003). The mechanism of projection neuron patterning is therefore likely to be independent of sensory neurons and has been shown to require the POU domain transcription factors Drifter and Acj6(Komiyama et al., 2003). How are sensory neuron arbors correctly positioned to allow interaction with appropriate neural partners?

In this paper, we investigate the role played by Robo receptors as possible guidance cues during formation of the olfactory map. A precedent for the role of Robo/Slit mediated signaling in determining topographic projections exists in the vomeronasal system of the rodent(Knoll et al., 2003). We demonstrate that a combinatorial code, defined by the domains and levels of Robo receptors, patterns not only the location of glomeruli but also the formation of the commissures which connect the two lobes. Our results from loss-of-function and gain-of-function analysis suggest that Robo receptors together with their ligand Slit are involved in positioning the arbors of the olfactory neurons at defined sites within the lobe.

SG18.1-Gal4 (Shymala and Chopra, 1999), lz-Gal4(Lebetsky et al., 2000), ato-Gal4 (Hassan et al.,2000), Or22a-Gal4 and Or47b-Gal4(Vosshall et al., 2000)strains were used to analyze olfactory neurons. For ectopic expression of Robo receptors, we used UAS-robo(2X), UAS-robo3 and EP2582(UAS-robo2) obtained from the Dickson laboratory (Rajgopalan et al.,2000a; Rajgopalan et al., 2000b); and UAS-robo from the Goodman laboratory (Simpson et al.,2000a; Simpson et al.,2000b). UAS-comm and UAS-slit were kindly provided by Guy Tear (Kidd et al.,1999); UAS-N-sybGFP (UNG12) by Mani Ramaswami; loco-Gal4, slit-Gal4 and slit-lacZ by Gerd Technau; GH146-Gal4 by Reinhard Stocker; and ato¹/TM3,Df(3R)p¹³/TM3 and UAS-ato by Andrew Jarman. Mutant alleles y,w; robo2¹/CyO wglacZand y,w; robo3¹/CyO wg lacZ were obtained from B. Dickson. The deficiency uncovering the slit locus Df(2R)WMG, wg^Sp–1 Bl¹ lt¹/In(2LR)Gla, wg^Gla–1(52A9-10;52D9-15); the FRT stocks yw; P{ry+t7.2=neoFRT}40A/CyO; Tub-Gal80 FRT40A; ey-FLP; UAS Cdc42.^V12 and balancer stocks were obtained from the Drosophila Stock Center at Bloomington,Indiana. EP31010 (UAS-abl) was obtained from the DrosophilaStock Center at Szeged.

All fly stocks were grown on standard cornmeal medium at 25°C. White prepupae (0 hours after puparium formation; APF) were collected and allowed to develop on moist filter paper. This stage lasts 1 hour, hence the error in staging is ±30 minutes. Wild-type pupae take about 100 hours to eclose when grown at 25°C in our laboratory.

Clones of robo mutants were generated using the mosaic analysis with repressible cell marker (MARCM) method described by Lee and Luo(Lee and Luo, 1999). Stocks of FRT40A robo2¹/CyO FRT40A and robo3¹/CyO were generated using standard recombination. Males of genotype Or-22a-Gal4;robo2¹ (or robo3¹) FRT40A/CyO;ey-Flp/+ were crossed to Tub-Gal80 FRT40A/CyO;UNG12/TM6-Tb females. Non-CyO, non-Tb flies were dissected and stained with antibodies against GFP and mAbnc82 (see below). GFP expressing neurons are homozygous for the mutation. To ascertain the efficiency of eye-Flp,we crossed eye-Flp/TM6-Tb flies with Act5C>CD2>Gal4, UAS-nlsGFP. All non-Tb flies showed GFP expression in the entire eye antennal disc. The brains of these animals did not show any clones in the cells around the olfactory lobes,establishing that eye-Flp generates clones only in the sensory neurons of the olfactory system.

Dissection and antibody staining of pupal and adult brain whole mounts were carried out as described previously(Jhaveri et al., 2000). Where mAbnc82 was to be used, the protocol of Laissue et al.(Laissue et al., 1999) was followed. The primary antibodies used were anti-Robo (1:10 from Corey Goodman), anti-Robo2 (1:200 from Corey Goodman), anti-Robo3 (1:100 Developmental Studies Hybridoma Bank at University of Iowa), mouse anti-Slit(1:50 from Spryos Artavonis-Tsakonis), mAbnc82 (1:10 from E. Buchner), rabbit anti-GFP (1:10,000 from Molecular Probes) and rabbit anti-Repo (1:5,000 from S. Susinder). Secondary antibodies used were Alexa 488- and Alexa 568-coupled goat anti-mouse and anti-rabbit IgG (1:200 Molecular Probes). Labeled samples were mounted in anti-fading agent, Vectashield (Vector Laboratories) imaged on BioRad Radiance 2000 at 1 μm intervals; data were processed using Confocal Assistant 4.2 and Adobe Photoshop 5.5.

In order to quantitate staining intensity, the regions of interest on 0.7 or 1 μm sections were demarcated and pixel intensity was estimated using Image J software. The cumulative intensity over the volume of the glomerulus(for Robo experiments) or the entire lobe (for Slit experiments) was estimated.

The Drosophila olfactory lobe comprises about 50 glomeruli located at fixed positions within the mediolateral, anteroposterior and dorsoventral axis (Fig. 1A). Sensory neurons expressing a given candidate odorant receptor (Or) target to the same glomeruli (Fig. 1B) and also send projections to the contralateral lobe(Fig. 1A,B, arrow). Previous work (Jhaveri et al., 2000)has shown that adult olfactory neurons differentiate within the first one-third of pupal life, radiate over the lobe anlage and transit across the midline (Fig. 1C). Sensory neurons invade the lobe during the next one-third of pupation and form distinct glomeruli (Fig. 1D).

We have used antibodies against the three Robo receptors to examine their localization in olfactory neurons during pupal life. The patterns of Robo,Robo2 (Lea – FlyBase) and Robo3 are rather dynamic and appear markedly different when examined early during lobe development(Fig. 2A-D), when compared with later after glomeruli are formed (Fig. 2G-I). During the first ∼20 hours after puparium formation(APF), when the olfactory neurons are on the surface but have not yet invaded the lobe, Robo is expressed uniformly on all afferent axons (arrowhead in Fig. 2A,D). Robo2 is present at low levels in all neurons but is enriched in regions lateral to the commissure(red arrows in Fig. 2B). A careful examination of confocal sections through a number of pupal lobes stained with anti-Robo2 suggests that immunoreactivity is lower as axons transit the midline (white arrow in Fig. 2B,D) than just prior to/after crossover. Expression of Robo2 declines in older pupae and is no longer detectable by ∼40 hours APF (not shown). Axons that express high Robo3, lie at more medial positions in the outer nerve layer (red arrows in Fig. 2C,D). The analysis of patterns of expression indicates that populations of neurons possess unique combinations of Robo, Robo2 and Robo3 that change during development.

Ziatic et al. (Ziatic et al.,2003) showed that robo3 expression in the embryonic peripheral nervous system is regulated by the proneural gene atonal(ato). In the adult olfactory system, ato specifies a subset of neurons that are the first to develop and appear to guide the rest of the axons into the lobe (Jhaveri and Rodrigues, 2002). In ato¹/Df(3R)p¹³animals, these `pioneers' fail to form and the rest of the neurons stall at the entry to the olfactory lobe. A subset of the Ato-independent neurons express Robo3 (arrows in Fig. 2E). Furthermore, only a subset of the Ato-dependent neurons visualized by Ato::GFP express Robo3 (Fig. 2F). As expected, these occupy medial positions in the outer nerve layer (arrows in Fig. 2F). These data together suggest that Robo3 is not expressed in genetically defined subset of neurons in the pupal olfactory system.

Sensory neurons begin to invade the lobe from about 25 hours APF and the first signs of glomerular organization become apparent by around 36 hours APF(Jhaveri et al., 2000; Jefferis et al., 2003). Glomerular formation occurs sequentially and by 60 hours APF most of the glomeruli have formed. The entry of glial cell processes and concomitant increase in lobe volume, results in some re-organization of glomerular position and the adult pattern (Fig. 2J) can only be discerned by about 80 hours APF. Robo and Robo3 are enriched in subsets of sensory neurons as they terminate within the lobe(Fig. 2G-I). Robo is detected in most axons, although at differing levels(Fig. 2G,I), while Robo3 is strongly enriched in terminals within a smaller number of glomeruli(Fig. 2H,I). A comparison of stained 60 hour APF lobes (Fig. 2H,I) with the adult glomerular map(Fig. 2J) suggests that Robo3-expressing neurons tend to preferentially target more dorsomedial locations. An estimation of Robo and Robo3 immunoreactivity in identified glomeruli supports the idea of a combinatorial code in determining sensory neuron position (Table 1).

We stained brains of different pupal ages with antibodies against the secreted ligand Slit (Rothberg et al.,1990; Kidd et al.,1999). A sheet of cells in the midline of the sub-esophageal ganglion expresses high levels of Slit (arrowheads in Fig. 2K). Immunoreactivity declines in later pupae (after 60 hours APF) and is absent in the adult. The midline cells do not express the glial marker Reverse Polarity (Repo) (red in Fig. 2K). Other regions in the midbrain closely associated with groups of Repo-positive glial cells were also labeled by anti-Slit (arrows in Fig. 2K). The diffuse nature of the staining makes it difficult to ascertain whether the glia are the source of secreted Slit in the midbrain. At 20 hours APF, the boundaries of the olfactory lobes are clearly demarcated by the presence of surrounding glial cells (broken lines in Fig. 2K). Slit levels within the lobe neuropil is significantly higher than that of the background(Fig. 2L). Expression can be detected from 14 hour APF and begins to decline by 60 hours APF (not shown).

The MARCM method (Lee and Luo,1999) combined with ey-FLP(Newsome et al., 2000)generates large patches of homozygous tissue in the eye-antennal disc(Ang et al., 2003; Jefferis et al., 2003). As flip-out occurs early, phenotypes generated in mature neurons result from a lack of gene function from the beginning of differentiation. We generated clones of robo2¹ and robo3¹ and examined targeting of a small number of sensory neurons marked by the Or22a-Gal4 transgene. Sensory neurons expressing Or22a normally project to glomerulus designated DM2 (Fig. 3A) (Vosshall et al.,2000) and cross-over to the contralateral lobe in the inter-antennal commissure (arrow in Fig. 3A).

Neurons lacking Robo2 function (robo2¹ clones) fail to cross over to the contralateral lobe and terminate at the midline forming small `glomerular-like' structures (arrowheads in Fig. 3B,C). The terminals showed immunoreactivity against the synaptic marker nc82 (arrow in Fig. 3B). Targeting to DM2 occurred normally although in many (13 out of 16) cases the glomeruli appeared less intensely innervated by GFP-expressing neurons. Loss of Robo3 function(robo3<sup>1</sup> clones), however, affected targeting of axons rather dramatically (Fig. 3D-F). In all cases, some mutant neurons did project correctly to DM2 (Fig. 3D,F) although a subset of axons strayed to ectopic sites (arrows in Fig. 3D,E). Commissure formation was unaffected (not shown). The erroneously placed terminals formed`glomerular-like' organizations as revealed by staining with mAbnc82, but these did not correspond in shape or position to those previously identified by Laissue et al. (Laissue et al.,1999). A large irregular shaped `glomerulus' located ventrally in the posterior region of the lobe was most frequently observed(Fig. 3E,F). In about half the preparations, an additional site was observed in a dorsolateral location(Fig. 3D,F). We ascertained that such ectopic targets were never found in control animals carrying the or22a-Gal4 (14.6) transgene(Bhalerao et al., 2003).

As Robo is expressed rather generally in olfactory neurons, we decided to study loss-of-function by targeted misexpression of antagonists of signaling(Dickson, 2001), rather than in clones. SG18.1-Gal4 expresses in a large fraction of olfactory neurons thus revealing most of the glomeruli(Fig. 3G, asterisks) as well as the antennal commissure (arrow in Fig. 3J). Ectopic expression of commissureless (comm)(Kidd et al., 1998) using SG18.1-Gal4 resulted in disorganization of glomerular patterning (compare Fig. 3H with 3G) with a weak effect on the commissure (arrow in Fig. 3H). Comm has been shown to downregulate Robo, although its effect on Robo2 and Robo3 is less well understood (Rajagopalan et al.,2000a; Rajagopalan et al.,2000b). The phenotype of Comm ectopic expression suggests that Robo is necessary for determining sensory neuron position within the lobe. Abelson kinase (Abl) phosphorylates the CC0 and CC1 domains of Robo, thus antagonizing signaling (Bashaw et al.,2000). Ectopic expression of either Abl or a constitutively active Dcdc42^v12 completely abolishes glomerular formation(Fig. 3I,K). Sensory neurons expressing Dcdc42^v12 (SG18.1::Dcdc42^v12) show an attraction for the midline and terminate there forming `glomerular-like'structures at the midline (Fig. 3L, asterisks). Results from loss-of-function clones predict such a phenotype for robo2 nulls. Constitutive activation of Dcdc42 is known to affect cytosketal dynamics generally, and could phenocopy a loss-of-function of all Robo receptors(Fritz and VanBerkum,2002).

We ectopically expressed Robo in sensory neurons to test whether the domains and levels of receptors are instructive in positioning of sensory terminals within the lobe. SG18.1::GFP was used to drive Robo in olfactory neurons; the positions and morphology of glomeruli could be visualized by GFP. Robo is expressed endogenously in all olfactory neurons and the small increase in level caused by driving a single copy of the UAS-robo transgene did not significantly alter lobe morphology (not shown). Higher levels achieved by driving three copies of the transgene abrogated glomerular formation (Fig. 4A). Changing the nature of the Robo code by misexpressing Robo3,however, resulted in a dorsomedial shift of projections(Fig. 4B). The commissure forms normally when either Robo or Robo3 are misexpressed (red arrows in Fig. 4A,B). Ectopic expression of Robo2, however, completely abolishes commissure formation with a less severe effect on glomerular morphology(Fig. 4C).

Next, we tested whether the genetic elements participating with Robo signaling in other well-studied systems also operated in the Drosophila adult olfactory system (Grunwald and Klein, 2002). We crossed a deficiency for the Slit region into an SG18.1 UAS-GFP UAS-robo2 recombinant. In this situation, where endogenous levels of the ligand were reduced by 50%, commissure formation, which is disrupted by the ectopic expression of Robo2, was restored and glomerular morphology also returned to normal (Fig. 4D). Targeted down-regulation of Robo signaling by misexpression of Comm(Fig. 4E) or activated Dcdc42^v12 (Fig. 4F),respectively, also suppress the phenotype caused by elevated Robo2.

Results presented above argue that sensory neuron positioning within the lobe is determined by signaling through the Robo receptors. Reduction of Slit levels suppress the effect of receptor overexpression, demonstrating that the phenotypes are mediated through endogenous ligand. In this case, alteration of the geometry of the Slit gradient by misexpression would be expected to alter terminal positioning of sensory neurons. We drove high Slit expression in glial cells around and within the lobe using loco-Gal4 (data not shown) (Jhaveri et al.,2000). Staining of the adult lobes in these animals with an antibody against the synaptic marker mAbnc82 revealed the presence of ectopic glomeruli outside the normal lobe circumference (broken lines in Fig. 4H,I). Increasing Slit levels further using multiple copies of the transgene led to more severe effects.

Our model proposes that olfactory neurons traveling in the outer nerve layer possess a different combination of Robo receptors that respond to Slit by branching into the lobe and arborizing at specific positions. In order to understand this positional code, we selected a Gal4 line that would allow us to drive expression in a set of neurons projecting to identified glomeruli from early during development. lz-Gal4;UAS-GFP labels two glomeruli – DM6 and DL3 – during development and in the adult brain (Fig. 5A-D), thus providing a means to examine the location of selective sensory neurons when the combinations of Robo are altered. We found that a change in the levels of any of the three Robo receptors, caused by misexpression using the lz-Gal4 driver, altered the positions of these identified terminals(Fig. 5E-P). The phenotypes showed variable expressivity; however, we were able to categorize preferred positions for the terminals in each treatment(Fig. 5D,H,L,P; Table 2).

Elevated Robo levels shift DL3/DM6 neurons to more central locations(arrowheads in Fig. 5E,G; 1 in H). Robo3 misexpression shifted the positions of the arbors most frequently to a mediodorsal axis (Fig. 5M-P). Large irregular-shaped glomeruli (arrowhead in Fig. 5M, 1′ in P) were frequently observed (Table 2). The changes in neuronal positions observed by Robo2 misexpression(Fig. 5I-L) were somewhat surprising given our hypothesis that Robo2 is involved largely in commissure formation. We suggest that high levels of Robo2 induced by lz-Gal4could interfere with the function of endogenous receptors. Robo2 misexpression most frequently produced cases where projections were seen terminating within a single lobe (arrowheads in Fig. 5I).

The ectopic `glomeruli' produced by alterations in the Robo code showed a normal organization of cellular elements(Fig. 5Q-V). In the wild type,terminal branches of sensory neurons remain at the periphery of each glomerulus (Fig. 5Q,S). Dendritic arbors of the lobe interneurons, filled the entire glomerulus as seen by GFP driven by GH146-Gal4(Jefferis et al., 2001)(Fig. 5R, green) or the synapse specific marker mAbnc82 (Fig. 5Q′,R′,S red). Glomeruli produced by misexpression of any of the Robo receptors also showed a similar organization as evidenced by mAbnc82 staining (Fig. 5T-V).

Our expression and genetic data allows us to propose the model summarized in Fig. 6. Neurons arriving at the olfactory lobe in the antennal nerve express Robo, and those expressing high levels of Robo3 additionally decussate onto the medial side of the outer nerve layer (Fig. 6A). The position of an axon in the nerve layer is influenced by Slit levels, although the identity of the cells that contribute Slit still needs to be elucidated. Several regions of Slit expression have been detected in the brain, although the cells at the midline express highest levels. Robo2, which is expressed at very low levels in all sensory neurons, is elevated after they cross the midline thereby preventing re-crossing. Later during pupation(Fig. 6B), sensory axons branch into the lobe and terminate at distinctive positions regulated by their unique Robo code in response to Slit levels. This allows short-range interactions with the dendritic arbors of projection neurons leading to formation of glomeruli.

Our results, taken together with findings from a wide variety of tissues and organisms, suggest the following steps by which precise patterning of brain regions take place during development. Regional specification of ectodermal tissues selects neuronal lineages in which identities of groups of neurons are specified (Jan and Jan,1994). Mechanisms to regulate receptors for long-range repulsive and attractive cues are among the properties that each differentiated neuron acquires.

The mechanism by which attractive and repulsive cues act to shape neural architecture is most elegantly demonstrated in the midline of the Drosophila embryo. A combinatorial expression of Robo receptors respond to the diffusible ligand Slit to dictate positioning of axons within longitudinal tracts (Rajgopalan et al., 2000a; Rajgopalan et al., 2000b; Simpson et al., 2000a; Simpson et al., 2000b). At the midline, commissural neurons downregulate receptor expression by Comm-mediated protein degradation (Rosenzweig and Garrity, 2002). Similar principles guide the selection and shaping of axon fascicles in the retinotectal system of the zebrafish, as well as several major pathways in mammalian brains(Hutson and Chien, 2002; Plump et al., 2002; Bagri et al., 2002). The general principle of growth cone repulsion by Slit-mediated Robo signaling appears to be a conserved theme in circuit design(Richards, 2002).

Can a similar mechanism be exploited to determine the location of synapses in three-dimensional space? Ziatic et al.(Ziatic et al., 2003) have chosen projections from the chordotonal organs specified by Ato to demonstrate how Robo3 can specify location of sensory arbors in the central nervous system. The terminals act as substrate upon which synaptic interactions with second-order neurons are built leading to formation of connectivity. A similar process also appears to operate in giant fiber system of adult Drosophila, where Robo receptors have been shown to play key roles in synapse formation (Godenschwege et al.,2002). We envisage a comparable developmental strategy in the olfactory lobe of the Drosophila adult. The correct spatial and temporal regulation of Robo receptors and their response to Slit levels determines the positioning of neurons and branching of terminal arbors at specified positions within the lobe. This facilitates short-range interactions with appropriate projection neurons which are already present within the developing olfactory lobe (Jefferis et al., 2003).

Although the source and nature of the Slit gradient within the developing olfactory system remains obscure, we have demonstrated that an alteration in levels by ectopic expression leads to aberrant lobe patterning. Slit exerts growth cone repulsion at the Drosophila embryonic midline through its action on Robo receptors, as well as by silencing the attractive cues of Netrins by action on the DCC receptor(Stein and Tessier-Lavigne,2001; Giger and Kolodkin,2001). In the cortex of the mammalian brain, Slit is widely and dynamically expressed and exerts a major role on dendritic growth and branching (Zinn and Sun, 1999; Whitford et al., 2002). Studies on cortical cultures demonstrated that concentrations of Slit strongly affect dendritic length as well as number of branch-points per cell. Hence,the complex geometry of a Slit gradient during development and its ability to induce attraction, repulsion, as well as branching, could provide a means to sculpt axon pathways in three-dimensional space.

The stereotypic location of each glomerulus within the lobe, and the connectivity within and between glomeruli, are important aspects of information processing in these functional units(Rodrigues, 1988; Galazia and Menzel, 2000; Fiala et al., 2002; Ng et al., 2002; Wang et al., 2003). The relatively conserved pattern of olfactory glomeruli within different individuals suggests a robust developmental program for their formation.

The effects of Robo loss of function are incompletely penetrant and suggest that multiple signals must exist for targeting of neurons. We propose that the Robo receptors act in concert with DSCAM isoforms to guide sensory neurons to the ∼50 unique positions within the three dimensional architecture of the lobe (Hummel et al., 2003). DSCAM and also Robo signaling could converge onto the Dock/Pak module,mutations in which also result in targeting defects(Ang et al., 2003; Fan et al., 2003). An accurate spatiotemporal estimation of receptors, ligands and signaling cascades across the developing olfactory lobe, together with mathematical modeling could provide valuable insights about the determination of glomerular patterns. Such knowledge is likely to be more generally applicable in the analysis of more complex brain structures including those of vertebrates.

We are grateful to Marina Voloshina and Niranjan Gaikwad for generating recombinant strains. We are indebted to Guy Tear, Erich Buchner, Gerd Technau,Leslie Vosshall, Andrew Jarman, Barry Dickson and Corey Goodman who providing us with many fly stocks and antibodies. We gratefully acknowledge K. VijayRaghavan, Satyajit Mayor and Michael Bate for their interest and valuable discussions; and K. VijayRaghavan and Mani Ramaswami for comments on the manuscript. This work was funded by grants from HFSP (RG0134), the Kanwal Rekhi Fellowship (to D.J. and A.S.) and core funding from TIFR.