Dock and Pak regulate olfactory axon pathfinding inDrosophila

The survival and reproduction of many animal species are critically dependent on their keen sense of smell. In mammals, the detection of diverse odors in the environment is mediated by ∼10³ different seven transmembrane odorant receptors, expressed by ∼10⁶ olfactory neurons (ONs) arrayed in the olfactory epithelium (reviewed byFirestein, 2001). The perception of odor requires that the odor information be encoded in a spatial map by the precise targeting of olfactory axons in the brain (reviewed byMombaerts, 2001).

The exquisite connectivity of the odor map raises questions about its development. What are the molecular mechanisms that allow olfactory axons to connect precisely with their targets? A number of guidance molecules have been found to be expressed in the developing olfactory system in vertebrates(reviewed by Key and St John,2002). Targeted disruptions of the neuropilin and ephrin A5 genes in mouse show that they direct the projections of olfactory axons to broad zones in the olfactory bulbs (Cloutier et al., 2002; Knoll et al.,2001; Schwarting et al.,2000). Furthermore, mutations of the odorant receptor genes reveal that they control the convergence of olfactory axons on specific glomeruli(Mombaerts et al., 1996;Wang et al., 1998). Little is known about the molecular mechanisms by which these proteins coordinately guide olfactory axons to their correct glomerular targets(Key and St John, 2002).

Drosophila provides an excellent system with which to unravel odor map development at the molecular and cellular levels. The anatomy and development of the fly odor map not only bear close resemblance to those of mammals, it is also simpler, containing only ∼60 Or genes(Clyne et al., 1999;Gao and Chess, 1999;Scott et al., 2001;Vosshall et al., 1999) and∼40 uniquely identifiable glomeruli(Laissue et al., 1999). InDrosophila, ONs differentiate in the antennal disc and send their axons to the antennal lobe (AL), the fly equivalent of the mammalian olfactory bulb (Jhaveri et al., 2000;Stocker et al., 1990). Within the AL, olfactory axons synapse on the dendrites of projection neurons (PNs)in distinct glomeruli. As in mammals, each glomerulus is innervated by olfactory axons expressing the same Or gene(Gao et al., 2000;Scott et al., 2001;Vosshall et al., 2000). Remarkably, each glomerulus is also innervated by PNs born at a specific time in development (Jefferis et al.,2001). Thus, in Drosophila, there is a precise pairing of ONs expressing a given Or gene with PNs of a specific identity.

Previous studies have identified Dock and Pak (p21-activated Kinase) as components of a signaling cascade that regulates the projection of photoreceptor axons (Garrity et al.,1996; Hing et al.,1999; Rao and Zipursky,1998). Dock is the fly homolog of the human Nck protein,consisting entirely of three SH3 domains and a single C-terminal SH2 domain(Garrity et al., 1996). The domain structure of Dock suggests that it acts as an adapter, linking receptors to downstream signaling proteins(Li et al., 2001). Pak is a serine/threonine kinase that is also highly conserved in evolution. Kinases of this family are defined by the ability to bind, and be activated by Cdc42 and Rac, key regulators of the actin cytoskeleton(Bagrodia and Cerione, 1999;Daniels and Bokoch, 1999). Experiments, in both Drosophila and mammalian cells, show that Dock and Pak function in a signaling cascade to regulate cell motility(Hing et al., 1999;Sells et al., 1999).

In this study, we show that Drosophila olfactory axons take stereotyped pathways in an outer nerve layer to find and innervate their cognate glomeruli. We found that dock and Pak are necessary for the precise wiring of the olfactory map. Using cell-type specific cDNA rescue, antibody localizations and clonal analyses, we observed thatdock and Pak function autonomously in olfactory axons to control various choice points in their trajectories to the cognate glomeruli. Finally, structure/function studies show that Dock and Pak function in a signaling pathway in ON axon guidance. Thus, Dock and Pak may form a core signaling cascade employed by different ON subclasses to pathfind to their respective glomeruli.

All fly lines were obtained from the Drosophila stock center except for the following. 981-Gal4 and SG18.1-Gal4(Jhaveri et al., 2000) were kindly provided by V. Rodrigues. Or22a-Gal4, Or47a-Gal4 andOr47b-Gal4 (Vosshall et al.,2000) were gifts from L. Vosshall. GH146-Gal4(Stocker et al., 1997) was from R. Stocker, UAS-nsyb::GFP(Estes et al., 2000) was from M. Ramaswami and ey-FLP; FRT82 Pak¹⁶/TM6B(Newsome et al., 2000b) was from B. Dickson. UAS-dock (Rao and Zipursky, 1998) was provided by Y. Rao, while UAS-Pakwas constructed by ligating a Myc-tagged Pak cDNA into pUASTand then transforming into the fly using standard germline transformation technique.

MARCM (Lee and Luo, 1999)was carried out by heat-shocking third-instar larvae of the following genotypes: hs-FLP/+; dock Or47a-Gal4/dock (or +); FRT82 Gal80/FRT82 UAS-mCD8::GFP and hs-FLP/+; dock Or47b-Gal4/dock (or+); FRT82 Gal80/FRT82 UAS-mCD8::GFP at 37°C for 40 minutes. Adult brains were dissected and processed as described below.

Adult brains (from 1- to 2-day-old animals), pupal antennae and larval imaginal discs were dissected in phosphate buffered saline (PBS). Tissues were fixed in PLP (2% paraformaldehyde, 0.25% sodium periodate, 75 mM lysine-HCl and 37 mM sodium phosphate pH 7.4), washed with PBST (PBS with 0.5% Triton X-100) and subjected to antibody staining. nc82 mAb (1:20) (A. Hofbauer, PhD thesis, University of Wurzburg, 1991) was a gift from A. Hofbauer. Rabbit anti-Dock 1:500 (Clemens et al.,1996) was a gift from J. Dixon. Mouse 22C10 mAb (1:20) and rat anti-ELAV, 7E8A10 (1:20) were from Developmental Studies Hybridoma Bank. Rabbit anti-GFP polyclonal antibody (1:100) was from Clontech, and rat anti-CD8 α subunit mAb (1:100) was from Caltag. The secondary antibodies, FITC-conjugated goat anti-rabbit, Cy3-conjugated goat anti-mouse and FITC-conjugated goat anti-rat, were purchased from Jackson Laboratories and used at 1:200 dilutions.

Adult antennae from animals expressingOr-Gal4/UAS-lacZ^nuclear were fixed in 25% gluteraldehyde for 1 hour, washed in PBST and stained in 0.2% X-Gal solution(Ashburner, 1989). The antennae were then cleared in Faure's mountant (34% v/v chloral hydrate, 13% glycerol,20 mg/ml gum Arabic) and photographed with the SPOT-RT cooled CCD camera. Sensilla on the third antennal segment were counted by projecting images on a video monitor.

We first noticed that the prominent paired protuberances at the anterior of the brain, the ALs, are severely reduced in dock and Pakhomozygous escapers compared with wild type. To probe the structure of the ALs, we stained the dock^P1/dock^P1 andPak⁶/Pak¹¹ mutant brains (seeTable 1 for descriptions of the mutant alleles) with the nc82 mAb, which stains the AL neuropil (A. Hofbauer,PhD thesis, University of Wurzburg, 1991). In wild type, ALs are rounded(∼90 μm diameters) and exhibit distinct anatomical subdivisions: the glomeruli (Fig. 1A). Indock and Pak mutants however, ALs are smaller (50 to 60μm), mis-shapen and have an amorphous neuropil, suggesting that glomeruli development is defective (Fig. 1B,C). The loss of well-defined glomeruli in dock andPak mutants suggests that guidance and/or synaptogenesis of ON axons are disrupted.

To ascertain the fate of the individual glomeruli, we expressed the synaptobrevin::GFP fusion protein (encoded by the UAS-nsyb::GFP gene)(Estes et al., 2000) under the control of specific odorant receptor promoters (Or-Gal4 drivers)(Vosshall et al., 2000). The presynaptic varicosities of the DM2, DM3 and VA11m glomeruli(Laissue et al., 1999), which are located at various positions in the AL, were surveyed using theOr22a-Gal4, Or47a-Gal4 and Or47b-Gal4 drivers, respectively. In wild-type controls, DM2 and DM3 are spheroidal neuropils of ∼10 μm diameters and are located in the mediodorsal surface of the lobe(Fig. 1D,G). Va11m is crescent in shape, ∼40 μm long and 10 μm wide, and is located on the anterolateral surface, close to the nerve entry point(Fig. 1J). In bothdock^P1/dock^P1 andPak⁶/Pak¹¹ mutants, DM2, DM3 and VA11m are severely mis-shapen or split into smaller structures that scattered randomly in the AL (Fig. 1). Indock, 42% of DM2 (n=26) and 75% of DM3 (n=36) are ectopically located or splinter into smaller structures, scattered in the neuropil (Fig. 1E,H). Although the integrity and position of VA11m remain relatively unchanged in bothdock and Pak mutants, it is enlarged and appears to extend into the domains of surrounding glomeruli. In 95% of dock(n=20) and 95% of Pak (n=18) ALs it is seen to engulf the adjacent VA1d glomerulus completely(Fig. 1K,L). There does not appear to be any consistent pattern in the distribution of the ectopic glomeruli, which are scattered randomly and differ from one AL to another. As assessed by the randomness in the positions of DM2 and DM3, and the aberrant structure of VA1lm, the phenotype of Pak is indistinguishable from that of dock, although defects occur at a slightly lower penetrance(for example, 66% of DM3, n=18;Fig. 1I).

The severe defects in glomerular development of dock andPak mutants lead to the critical question of the focus ofdock and Pak actions in the developing AL. The precise formation of the olfactory map requires multiple interactions between ingrowing olfactory fibers and major cell types of the AL, such as the PNs and AL glia (Oland and Tolbert,1996). In theory, dock and Pak may act in any of these cells to direct the wiring of the AL.

Staining with the anti-Dock antibody showed that Dock is highly expressed in the dendrites and axons of ONs during the period of axon pathfinding to the AL (see supplemental figures). Staining with the anti-Pak antibody showed that Pak is similarly localized to ON axons (data not shown), although the lack of a Pak protein null mutation prevents us from assessing the specificity of the staining. The presence of Dock and Pak proteins in ON axons is consistent with a role in ON axon guidance. To test the notion that the proteins are needed in ONs, we asked if removal of dock and Pak functions specifically from ONs would disrupt their glomerular targeting. We found that theey-FLP/FRT method of site-specific recombination(Newsome et al., 2000a)induced the formation of large patches of homozygous tissues in the eye-antennal disc but not in the brain (see supplemental figures). We used the ey-FLP/FRT system to test whether dock andPak are required in the antenna for the precise targeting of ON axons. Clones of wild-type, dock^P1 andPak¹⁶ mutant tissues were induced in the antennae of animals carrying either Or47a-Gal4 or Or47b-Gal4 transgenes(see Fig. 2 legend for the respective genotypes). The mosaic animals were allowed to develop to adulthood and their brains were stained with nc82 mAb and anti-GFP to probe glomerular development. In animals with wild-type antennae, the AL neuropil is subdivided into well-defined glomeruli and ON axons converged precisely on their target glomeruli (Fig. 2A). By contrast, in animals with dock^P1/dock^P1 orPak¹⁶/Pak¹⁶ antennae, the ALs are severely mis-shapen and aglomerular, and ON axons terminated in ectopic locations(Fig. 2B,C). We conclude from these observations that dock and Pak function within the antennae to regulate the precise targeting of ON axons.

The mosaic experiments do not preclude an additional requirement fordock and Pak in the brain or in nonneuronal cells of the antenna. To test the hypothesis that dock and Pak function specifically in ONs for AL development, we asked if the expression of wild-type dock and Pak cDNAs in ONs would rescue thedock and Pak mutant phenotypes. The enhancer trap lineSG18.1-Gal4 (Jhaveri et al.,2000) was chosen to drive the expression of the cDNAs because we observed that it is preferentially expressed in a subpopulation of ONs with little expression in the brain. Examination of antennae fromSG18.1-Gal4/UAS-mCD8::GFP animals at 30 hours after puparium formation (hAPF) showed that GFP is expressed only in cells with a bipolar morphology characteristic of ONs (Fig. 2D). Examination of the brain in pupae at 30 and 55 hAPF, and in the adult, showed that GFP is specifically localized to the nerve layer of the AL and in numerous glomeruli, with little staining elsewhere(Fig. 2E,Fig. 3A). Removal of both the antennae and maxillary palps abolished GFP expression in the ALs, showing that the AL staining is contributed by olfactory afferents from these two organs(Fig. 2F). Hence,SG18.1-Gal4 is expressed preferentially in ONs of the antenna and the maxillary palp. Expression of dock and Pak cDNAs under the control of SG18.1-Gal4 strongly rescued the glomerular structures of the AL (Fig. 2G-J). We monitored four landmark glomeruli (VA1d, VA6, DA4 and DM6) as an indicator of normal AL development (Fig. 2K). Expression of cDNAs under the control of SG18.1-Gal4increased the average frequency of distinctly identifiable glomeruli from 14%to 85% in dock^P1/dock^P1 ALs and 35% to 94% inPak⁶/Pak¹¹ ALs. Although the rescue was strong,we noticed that not all glomeruli were restored. The incomplete rescue is probably due to the lack of SG18.1-Gal4 expression in some glomeruli(Fig. 2E). Nonetheless, the ability of SG18.1-Gal4 to support the substantial rescue of the mutant phenotypes indicates that it is suitably active during olfactory axon pathfinding. We conclude from these experiments that dock andPak genes are specifically required in the ONs to direct the guidance of the ON axons.

We infer from the ectopic glomeruli that ON axons fail to project to their correct destinations in dock and Pak mutants. To assess whether ON axons are misprojecting in dock and Pak mutants,we probed fiber pathways from the antenna to the AL. During development, ON axons leave the antenna and travel to the AL between 20 and ∼50 hAPF(Jhaveri et al., 2000). Axon pathways in the pupal antenna were visualized by staining 30 hAPF antennae with 22C10, a mAb against Futsch (see Fig. 6A), while pathways in the adult AL were labeled with either the mCD8::GFP or GAP::GFP fusion proteins (encoded by UAS-mCD8::GFP orUAS-GAP::GFP) expressed under the control of various Gal4drivers. Axon trajectories were examined with standard fluorescence microscopy or after three-dimensional (3D) reconstructions of 2D confocal sections.Fig. 3A shows a representative reconstruction of a wild-type adult AL labeled by the SG18.1-Gal4driver, which is expressed in coeloconic and trichoid sensilla(Jhaveri and Rodrigues, 2002). The overall pattern of ON projection in the AL is bilaterally symmetric and relatively invariant between individuals (seeFig. 3B1 for an interpretive drawing). GFP-expressing fibers extend through the antennal nerve and enter the AL at its anterolateral point (close to VA1lm). Thereafter, the axons radiate over the AL surface in characteristic tracks within a fibrous nerve layer, to target their cognate glomeruli. Contralateral axons course through a prominent commissure that connects the opposite ALs. That the individual ON classes project in stereotyped pathways was confirmed by labeling ONs with the Or22a-Gal4, Or47a-Gal4 and Or47b-Gal4drivers. In confocal reconstructions, Or47a axons can be seen to take relatively straight paths from the nerve entry point to DM3, their target glomerulus (Fig. 3E,F1). Examination of single Or47a axons (n=83) using the MARCM technique (Lee and Luo, 1999)showed that each axon projects determinately to DM3, where it splits into an ipsilateral branch, that terminates in DM3, and a contralateral branch, that projects across the commissure (Fig. 4A,B). Axons appear to extend directly to DM3 without making substantial changes in direction. Many fibers appear to project individually,although a number of fibers also merge into fascicles as they converge upon their targets (Fig. 3E).Or47a axons always remain in the nerve layer throughout the entire trajectory until in the vicinity of their target, whereupon the converging fibers enter the lobe and terminate precisely on DM3. Or47b axons,however, terminate immediately upon disembarking from the nerve, establishing the crescent shape VA1lm (Fig. 3I,J1). A distinct fascicle issues from the medial edge of VA1lm carrying collaterals between the glomerulus and the commissure. Single-axon analysis (n=38) showed that the contralateral axons project in relatively straight paths, seldom straying beyond the confines of a narrow zone, between VA1lm and the commissure(Fig. 4E,F). Our analyses show that in the wild type, olfactory axons make bilaterally symmetric and stereotyped patterns of projections to their glomeruli in the AL.

In antennae of dock^P1/dock^P1 andPak⁴/Pak⁶ mutants at 30 hAPF, cells exhibiting a bipolar morphology can be seen, projecting their axons out of the antenna in distinct fascicles, a pattern indistinguishable from the wild type(Fig. 6A-C). However, once in the AL, the pathways are clearly abnormal. In both dock andPak mutants, instead of forming characteristic tracks,SG18.1-Gal4-expressing fibers interweave to form a dense mat(Fig. 3B2-D). In 22%(n=32) of the dock^P1/dock^P1 brains, ON axons extend aberrantly to dorsal brain regions(Fig. 3C, right AL). Disruption in the overall projection pattern is reflected in the trajectories of the individual ON subtypes. In 3D confocal reconstruction, Or47a axons can be seen to deviate from their stereotyped pathways from the outset,veering to distant part of the AL, including even the core, in chaotic,meandering trajectories (Fig. 3F2). Visualization of single Or47a axons indock^P1/dock^P1 mutants showed that misrouting affect both the ipsilateral and contralateral branches of an axon(Fig. 4C,D). In 21% of the cases (n=45), both branches remain in the ipsilateral AL. Most of the branches (64%, n=45) ultimately fail to reach their destination,forming mis-shapen glomeruli in ectopic locations. Interestingly, in 9%(n=45) of the cases, three axon branches were observed. However, the extra branch is usually short and not associated with ectopic glomeruli. Although Or47b axons terminate in a single large glomerulus as in wild type, it is strongly mis-shapen in dock and Pak mutants(Fig. 3J2). Furthermore, the contralaterally projecting Or47b axons are severely defasciculated, projecting through a wide area of the AL surface. Examination of single Or47b neurons showed that whereas ipsilateral axons terminate normally in VA1lm, contralateral axons are frequently misrouted,either projecting to dorsal brain regions or stopping short of the VA1lm glomerulus (Fig. 4G,H). Our analyses of the dock and Pak mutant phenotypes therefore show that dock and Pak are not necessary either for the outgrowth of olfactory axons or for their projection through the antennal nerve. Instead, dock and Pak function primarily in the guidance of ON axon within the AL to steer the sensory fibers precisely to their cognate glomeruli.

The similarity in dock and Pak olfactory connectivity phenotypes suggests that these genes might function in a signaling cascade to regulate the targeting of ON axons. It has previously been shown that Dock and Pak interacts through the N-terminal PXXP motif of Pak and the second SH3 domain of Dock (Hing et al.,1999). We now observed that the P9L mutation(Pak⁴), which affects the N-terminal PXXP motif of Pak and abolishes its ability to bind to Dock, strongly disrupts ON axon pathfinding(Fig. 3H). To evaluate the hypothesis that Dock-Pak interaction is critical for ON axon guidance, we examined the requirements of the individual Dock domains for AL development. Expression of wild-type UAS-dock cDNA(Rao and Zipursky, 1998) under the control of SG18.1-Gal4 significantly rescued thedock^P1/dock^P1 aglomerular phenotype (compareFig. 5A with 5B). Although mutations that disrupt the first SH3 domain, the third SH3 domain or the SH2 domain do not affect the ability of dock to rescue the mutant phenotype (Fig. 5C,E,F), a mutation that disrupts the second SH3 domain completely abolisheddock activity (Fig. 5D). The requirement of both the N-terminal PXXP motif of Pak and the second SH3 domain of Dock suggests that physical interaction between Dock and Pak is necessary for proper AL development. To obtain evidence of a functional relationship between dock and Pak we tested for genetic interactions between the genes. First, we asked if simultaneous reduction of dock and Pak functions by half would disrupt AL development. The ALs of dock^P1/+;Pak⁴/+ animals exhibit normal axon trajectories and glomerular subdivisions similar to those of wild-type or heterozygous animals (see supplemental figures). However, the absence of genetic interaction in the compound heterozygotes should not be taken as evidence against dock and Pakfunctioning in the same signaling cascade. It is still possible that the decreased levels of dock and Pak are sufficient for normal functioning of the signaling pathway. We also asked if expression of the constitutively active Pak^myr(Hing et al., 1999) in thedock mutant would correct the mutant phenotype. However, expression of UAS-Pak^myr with SG18.1-Gal4 in wild type strongly disrupted the AL structure (L.-H. A and H. H., unpublished). The strong gain-of-function phenotype makes its use in genetic epistasis studies unfeasible.

As dock and Pak are autonomously required in ONs, one explanation for defective targeting of their axons is that cell fate specification of the neurons is disrupted. To explore this possibility, we examined in detail the critical steps of ON differentiation in both mutants. By monitoring ON development at various stages, we demonstrate that ONs differentiate normally in dock^P1/dock^P1 andPak⁶/Pak¹¹ homozygotes. First, antennal discs of dock and Pak mutants show lozenge gene expression in three semi-elliptical domains similar to wild type (see supplemental figures). Second, in nascent antennae derived from 30 hAPF mutant pupae, ONs express both the ELAV and Futsch antigens (Fig. 6A-C). Furthermore, axons extend out of the antenna normally. Third, in the mutant antennae, Or22a, Or47a and Or47b genes are expressed by characteristic numbers of cells that are scattered randomly within circumscribed domains as in wild type(Table 2;Fig. 6G-I; see supplemental figures)(Vosshall et al., 1999;Vosshall et al., 2000). Fourth, the three olfactory sensilla types (basiconic, trichoid and coeloconic) are found on the mutant antennae and have similar morphology,numbers and distribution as in wild type(Table 2,Fig. 6D-F; see supplemental figures)(Shanbhag et al., 1999). Examination of indigenous cells of the AL, such as the PNs and the AL glia(see supplemental figures),showed that these cells also differentiate normally in the dock andPak mutants. However, in the mutants, the dendritic arborization of PNs are more diffused and the number of glia processes are somewhat reduced. Thus, morphogenetic changes of the cells that accompany normal glomerular development fail to occur properly when the dock or Pakgenes are disrupted.

The connectivity of the olfactory map, and thus the logic of olfactory coding, is a direct consequence of the precise targeting of olfactory axons to their cognate glomeruli during development. The cellular and molecular mechanisms that underlie development of the olfactory map are essentially unknown. We have now characterized the pathways that olfactory axons take to their glomeruli and demonstrate that two growth-cone signaling molecules, Dock and Pak, function to regulate the precise convergence of the axons on their targets.

ONs of the antennae and maxillary palps undergo terminal differentiation during early metamorphosis and become predestined to express particular Or genes and synapse in specific glomeruli(Jhaveri et al., 2000;Vosshall et al., 1999). Between 20 and ∼50 hAPF, their axons leave the nascent antenna in fascicles and enter the AL in search of their targets. PNs, however, acquire their cell fates, which predetermine their glomerular choice, during larval development (Jefferis et al.,2001). During early pupal development their dendrites enter the AL and become precisely paired with ON axons in specific glomeruli. Thus, ONs expressing a given Or gene rendezvous with PNs of a particular identity within a topographically defined glomerulus in the AL.

We find that in the wild type, olfactory axons take stereotyped paths on the surface of the AL to converge on their cognate glomeruli. Detailed characterization of the axon trajectories, using Gal4 drivers expressed in different subclasses of ONs shows that, upon arrival at the anterolateral point of the AL, afferents project directly, with little sidetracking to their postsynaptic targets. As in the mouse and moth, these axon pathways are bilaterally symmetric and invariant from AL to AL(Mombaerts et al., 1996;Oland et al., 1998;Wang et al., 1998). How is this precise wiring pattern formed during development? In one model, each ON initially sends collaterals to multiple glomeruli and then withdraws the inappropriate branches in a process requiring odorant-evoked activity. Alternatively, the invariant pattern of connections is the result of directed axon migrations in response to spatially restricted pathfinding cues in the developing AL. A definitive answer to this question will require developmental study or direct observation of the extending axons. However, several observations are consistent with the notion that olfactory axons navigate directly to their cognate glomeruli. First, a temporal lag between early axon pathfinding and subsequent Or gene expression(Clyne et al., 1999) indicates that an odorant-evoked activity is unlikely to play an important role. Indeed,activity is neither required for formation nor maintenance of the olfactory map in mouse and moth (Belluscio et al.,1998; Lin et al.,2000; Oland et al.,1996). Second, and importantly, our finding that the growth cone guidance genes, dock and Pak, are needed for development of the olfactory map, provides strong evidence that directed axon migration plays a key role in the matching of ON axons with their correct glomeruli. Directed navigation of olfactory axons to their targets is also observed in zebrafish and moth (Dynes and Ngai, 1998;Oland et al., 1998).

In dock and Pak mutants, the stereotyped connectivity of AL neuropil is severely disrupted, leading to an aglomerular phenotype. We present three pieces of evidence indicating that dock andPak function in ONs. First, antibody staining shows that Dock and Pak proteins are expressed in antennal axons during the period in which they are projecting to the brain. Consistent with their requirements in ONs, removal ofdock and Pak activities from only the antennae results in ectopic targeting of olfactory axons. Finally, expression of dock andPak cDNAs specifically in ONs in otherwise mutant animals leads to strong rescue of the mutant AL phenotype. We noticed that although numerous glomeruli were restored upon the expression of the wild-type cDNAs, some glomeruli were not. We believe that the incomplete rescue is due to the expression of SG18.1-Gal4 in only a subset of all the ONs. However,it is also possible that the partial rescue reflects an additional requirement of dock and Pak functions in the brain. A recent study indicates that ONs may be divided into different classes based on the timing of their projections (Jhaveri and Rodrigues, 2002). We did not determine further if dockand Pak are required in all ONs or in only a specific subset. Although dock and Pak are specifically required in ONs, our finding of nonautonomous effects on the morphogenetic changes of the PNs and AL glia is in accord with earlier studies in which ONs were physically or genetically ablated (Graziadei and Monti-Graziadei, 1992;Hildebrand et al., 1979;Stocker and Gendre, 1988). Our data therefore show that proper termination of ON axons is also an important step in the sculpting of the AL neuropil into distinct glomeruli.

We provide evidence that the disruption in AL development in dockand Pak mutants is not an indirect effect of aberrant cell-fate determination or axonogenesis. By contrast, we observed that the precise targeting of ON axons is severely disrupted in dock and Pakmutants. To identify the cause of the mistargeting, we examined the axon pathways of individual ON classes (Or47a, and Or47b) at the single-cell level. Although an additional short branch was observed in 9% ofdock mutant neurons (n=45), the most striking defect observed in single-cell clones (64%, n=45) is the chaotic trajectories exhibited by both the ipsilateral and contralateral axons of the ONs. We conclude that the primary function of dock and Pakin ONs is axon pathfinding, to steer ON axons precisely to their target glomeruli. In mouse, mutations in the odorant receptor genes abolish the ability of olfactory axons to pathfind in the anteroposterior axis without affecting their migration in the dorsoventral axis, leading to the proposal that odorant receptors participate in the recognition of only anteroposterior guidance cues (O'Leary et al.,1999; Wang et al.,1998). However, after examining several hundred ALs for eachdock and Pak mutant, we did not observe any consistent patterns in the mistargeting of ON axons. We did observe that the ON classes are affected to different degrees by the loss of dock andPak activities. Although Or22a and Or47a axons terminate in numerous ectopic glomeruli, Or47b axons terminate in a single glomerulus, albeit mis-shapen, in the approximate position of the wild-type VA1lm. We currently do not know the reason for the differential sensitivity of the ON subtypes to the loss of dock and Pakfunctions. One possibility is that Or47b axons, which are among the first axons to enter the AL, are confronted with fewer developing glomeruli(Jhaveri et al., 2000) and hence fewer guidance choices than Or22a and Or47a axons that enter the AL later. Alternatively, Or47b axons may have less need fordock- and Pak-mediated navigational functions because VA1lm is located near the nerve entry point. Indeed, while the Or47bipsilateral axons frequently terminate accurately on VA1lm, the contralateral axons, which have to project across the entire AL surface, are often misrouted. In contrast to the severe projection defects in the AL, the migration of dock and Pak mutant axons through the antennal nerve takes place normally. It is possible that the lack of requirement ofdock and Pak functions during this phase of axon growth reflects a different guidance mechanism in the antennal nerve.

The observation that the ON axon trajectories are severely disrupted indock and Pak mutants suggests that the genes may mediate the detection or response of the growth cones to guidance cues in the environment. Our results indicate that in these events, dock and Pak are very likely to act in a signaling pathway. First, loss of either dockor Pak functions results in olfactory connectivity phenotypes that are indistinguishable. Second, both dock or Pak function autonomously in ONs. Third, mutations that disrupt the domains of Dock (second SH3 domain) and Pak (N-terminal PXXP domain; Pak⁴), which mediate interaction between the two proteins(Hing et al., 1999), disrupt ON axon targeting. We therefore propose that Dock and Pak are part of a signal transduction cascade that allows ONs to find and precisely pair with the correct postsynaptic partners. Although severely disrupted, the guidance of ON axons in dock and Pak mutants is not completely abolished,indicating that other genes function to steer ON axons to their targets as well.

As the Dock-Pak signaling pathway appears to govern the pathfinding of a number of ON subtypes, it is unlikely to explain the specificity of ON targeting. What transmembrane receptors might control the Dock-Pak signaling pathway and direct the precise pairing between ONs and their postsynaptic targets? Because ONs that target different glomeruli are interspersed in the olfactory epithelium, thus precluding their tagging by simple gradients of molecular cues, it is anticipated that the receptors are molecularly diverse. A highly diverse receptor family is, of course, the odorant receptor family. In mouse, odorant receptors themselves provide the specificity for target selection (Mombaerts et al.,1996; Wang et al.,1998). Odorant receptors are unlikely to play a guidance role inDrosophila, however, as expression of these genes begin long after axon migration has taken place (Clyne et al., 1999). Another family of diverse receptors that functions in cell-cell adhesion and is expressed in synapses is the cadherin superfamily(Yagi and Takeichi, 2000). In mouse, some members of the family, the CNRs, are expressed in the olfactory bulb, indicating that cadherin family proteins may also play a role in olfactory map development (Kohmura et al.,1998). Recently, in a biochemical screen for proteins that bind to Dock, Zipursky and colleagues identified the immunoglobulin superfamily transmembrane receptor Dscam (Schmucker et al., 2000). Interestingly, the expression of the Dscamgene is regulated by a novel, combinatorial splicing mechanism, which allowsDscam to encode up to 38,000 isoforms(Schmucker et al., 2000). Because of the tremendous diversity in its gene products, Dscam has been proposed to play a role in encoding synaptic specificity. Our discovery that the Dock-Pak signaling pathway regulate the projection of ON axons to their cognate glomeruli now presents us with a unique opportunity to begin to assess the possible roles of these receptors in the development of theDrosophila olfactory map.

We would like to thank L. Vosshall, R. Axel, V. Rodrigues, R. Stocker, Y. Rao, M. Ramaswami, J. Dixon and B. Dickson for generous gifts of fly lines and reagents. We also thank M. Kim for technical help. We are also grateful to P. Newmark, G. Robinson and members of the Hing laboratory for helpful discussions. Finally, we would like to express thanks to the anonymous reviewers for their constructive comments that greatly improved this manuscript. This work is supported by a grant from NIH/NIDCD (DC5408-01) (H. H.).