Drosophila Eph receptor guides specific axon branches of mushroom body neurons

Developing neurons reach their targets through the directed growth of their axons, the growth cones of which recognize and respond to guidance cues present in their environment. Neurons often make synaptic contacts with multiple remote targets by extending distinct axon branches that grow in different directions. A notable example of this is the development of neurons within the mushroom body (MB), the olfactory learning and memory center of insects (Heisenberg, 2003). MB neurons bifurcate and extend axons into two distinct areas of the brain,ultimately synapsing with different targets and giving the MB its characteristic bi-lobed appearance. The apparent synchrony of the growth of the two branches suggests that the growth cones from individual branching MB neurons may differentially respond to guidance cues(Wang et al., 2002).

A class of molecules known to play key roles in the organization of many tissues, including axon guidance in the nervous system, is the Eph receptor tyrosine kinase (RTK) family members and their ligands, the Ephrins. Members of these large, phylogenetically conserved families of receptors and ligands are found throughout the animal kingdom, but the families have undergone a considerable expansion within vertebrates, resulting in a total of 14-16 Eph receptors and eight to nine Ephrin ligands depending upon the species. Eph receptors are divided into two subclasses based on sequence and their preferential binding affinity for Ephrin ligands: EphA receptors bind to GPI linked Ephrin A ligands, while EphB receptors bind to transmembrane Ephrin B ligands. Ephs and Ephrins are expressed in many developing tissues and mediate a variety of cell-cell contact-dependent signaling events, including axon attraction and repulsion within the nervous system.

The diverse roles played by Ephs and Ephrins are reflected in the complexity of their signaling. Both the receptor and ligand are membrane bound molecules capable of transducing intracellular signals within a cell. Thus,signaling through both the receptor (forward) and the ligand (reverse) are possible. In some cases, such as rhombomere boundary formation within the hindbrain, activation of Eph/Ephrin signaling occurs when Eph-expressing cells encounter domains of Ephrin expression(Poliakov et al., 2004). However, in many cases Eph/Ephrin signaling occurs within cells expressing both receptor and ligand. The consequence of such ligand/receptor co-expression appears to be context dependent. For example, within the retinotectal system Eph/Ephrin co-expression by retinal ganglion cells may result in the Ephrin ligand masking the Eph receptor, blocking receptor activation by ligand expressed in trans(Hornberger et al., 1999). By contrast, within developing motoneurons, co-expressed ligand and receptor do not appear to interact, thereby allowing both receptor and ligand in principal to signal independently within the same cell(Marquardt et al., 2005). Consistent with this idea, within such co-expressing motoneurons the Eph receptor mediates growth cone collapse and repulsion, while the Ephrin mediates axon growth and attraction(Marquardt et al., 2005).

Eph/Ephrin signaling in axon guidance is perhaps best illustrated by its role in the topographic mapping of retinal ganglion cells in the superior colliculus (SC) (Poliakov et al.,2004). Retinal ganglion cell axons initially overshoot their targets within the superior colliculus and subsequently contact their correct termination zones through the extension of collateral branches. Along the anteroposterior (AP) axis of the SC, retinal ganglion cell targeting relies in part on EphA/Ephrin A signaling mediating repulsion(Brown et al., 2000; Nakamoto et al., 1996; Yates et al., 2001). By contrast, along the mediolateral axis of the SC, opposing gradients of EphB receptors on retinal ganglion cells and Ephrin B ligands on SC cells act to control the direction of branch extension and arborization by mediating either attraction or repulsion, depending upon the position of the branch relative to its termination zone (Hindges et al.,2002; Mann et al.,2002; McLaughlin et al.,2003).

In contrast to animals that have multiple Eph receptors and/or ligands, Drosophila has a single Eph and a single Ephrin. The Drosophila Eph receptor shows equal similarity to both the A and B subclasses (Dearborn, Jr et al.,2002; Scully et al.,1999), while the Drosophila Ephrin ligand is most similar to vertebrate Ephrin B ligands. Like other Ephrin B ligands, Drosophila Ephrin contains a transmembrane domain and a conserved tyrosine phosphorylation site (Bossing and Brand, 2002).

Both Drosophila Eph and Ephrin are expressed within the embryonic CNS at a time when neurons are extending axons towards their targets(Bossing and Brand, 2002; Scully et al., 1999). Two previous studies have suggested a role for Drosophila Eph/Ephrin signaling in neuronal development using RNA interference (RNAi) technology(Bossing and Brand, 2002; Dearborn et al., 2002). Here,we describe the generation of a null mutation in Eph, plus our analysis of Eph/Ephrin function within the Drosophila CNS in individuals lacking all Eph function. We show that Drosophila Eph and Ephrin can act as a functional receptor ligand pair in vivo to mediate axon repulsion. Despite this, we fail to detect axon guidance defects in the embryonic CNS of Eph mutant embryos. However, later in development Eph/Ephrin signaling plays a crucial role in the developing MB by guiding the projection of specific axon branches of individual MB neurons.

39c-18 (Wallrath and Elgin,1995) is a lethal white+ (w+) P element insertion on chromosome IV that was mapped to the bent locus by sequencing fragments generated by inverse PCR(Dalby et al., 1995). 39c-18 was mobilized using Δ2,3(Robertson et al., 1988). Six-hundred w;CyO,Δ2,3/+;39c-18/CiDmales were tested for chromosomes with reversion of 39c-18 lethality by singly crossing them to w-females carrying a lethal allele of bent(39c-18^LR#56/CiD) generated by excision of 39c-18. Five-hundred and eighty w+ non-CyO,Δ2,3 flies were isolated, 57 of which mapped to chromosome IV. Insertion sites for 54 lines were determined by inverse PCR; flanking sequences from three lines did not map to a single site and were discarded.

P114 excisions were generated by crossing w;CyO,Δ 2,3/+;P114/CiD males to w;ey^D/CiD females. One-thousand three-hundred w-excision lines were isolated, 2.5% of which were homozygous lethal. Southern blot analysis demonstrated that none of the lethal lines deleted sequences within the Eph genomic region. Seven out of 16 lethal lines did show rearrangements within the onecut genomic region, including the onecut^x122 and onecut^x49 alleles. Eight additional onecut alleles were identified by non-complementation with onecut^x122 and onecut^x49.

Viable lines were screened for rearrangements within Eph by PCR using primers that amplify the first three exons of the Eph gene(Scully et al., 1999). After additional southern blot analysis of Eph^x652 DNA indicated a deletion of the first three exons, breakpoints were determined by sequencing a PCR fragment generated from Eph^x652 DNA using primers predicted to bracket the excised region. The 5′ breakpoint of Eph^x652 maps to genomic position 627320 (BDGP release 4),368 bp upstream of the onecut translation start site. The 3′breakpoint lies within the third intron of Eph, at genomic position 632468.

Immunohistochemistry on dissected embryos was preformed as previously described (Callahan and Thomas,1994). Larval and adult brains were dissected and fixed as described by Hummel et al. (Hummel et al.,2003) with a total fixation time was 60 minutes.

Primary antibodies used: monoclonal antibody (mAb) anti-BP102 (1:20 dilution) (Seeger et al.,1993); mAb anti-Fas2 (1:30)(Lin et al., 1994); mAb 9E10 anti-c-myc (Dm - FlyBase) (1:50) (Developmental Studies Hybridoma Bank);Cy3-conjugated anti-HRP (1:500) (Jackson Immunoresearch); rabbit anti-β-Gal antibody (1:1000) (Cappel); and rabbit anti-GFP (1:5000)(Molecular Probes).

Ephrin-Fc was produced as described(Kaneko and Nighorn, 2003). Samples were incubated in either purified Ephrin-Fc (1:500 dilution in PBS) or straight supernatant, followed by incubation with either rabbit or mAb anti-human IgG, Fcγ (1:500) (Jackson ImmunoResearch). For fluorescence immunostaining, Alexa Fluor 488-conjugated anti-rabbit (1:500) (Molecular Probes) and Cy3-conjugated anti-mouse (1:500) (Jackson ImmunoResearch) were used. In situ hybridization was carried out as described(Dougan and DiNardo, 1992)using probes generated with fragments corresponding to ∼1 kb of 5′sequences of the Eph cDNA (Scully et al., 1999) or the Ephrin cDNA.

pUAS-Ephrin:myc was constructed from the full-length cDNA clone RE46807 tagged in-frame to six copies of the c-myc epitope in the pUAS vector (Brand and Perrimon,1993). For construction of pUAS-dephrinmyc^ΔIC, a PCR fragment was generated that deleted intracellular sequences from amino acids 611 to 652. This was subcloned as an EcoR1/AgeI fragment into pUAS-Ephrin:myc, replacing the wild-type sequence. For each UAS transgene, multiple lines were generated by P element transformation (Spradling and Rubin,1982). Lines with the strongest anti-myc staining were used.

The following flies were generated for MARCM analysis of MB neurons: hsp70-FLP,elav-Gal4,UAS-mCD8GFP/+ or Y;FRT^G13/FRT^G13,TubP-Gal80;Eph^x652/CiDand hsp70-FLP,elav-Gal4,UAS-mCD8GFP/+ or Y;FRT^G13/FRT^G13,TubP-Gal80;Eph^x652/Eph^x652. Clones within the α/β lobes were generated by heat shocking pupae at 38°C, as described (Lee and Luo,2001). Eighty adult Eph^x652 individuals were examined for marked MB clones. Branching patterns of 41 unambiguously labeled Eph^x652 mutant MBs clones were examined.

Both Eph and Ephrin genes map within 33 kb of one another on the 4th chromosome. To generate mutations in the two genes, we mobilized a P element, 39c-18 (Wallrath and Elgin,1995), located ∼145 kb away from Eph and inserted in the bent locus (see Materials and methods). Fifty-five independent 4th chromosome insertion lines were generated. Insertion sites for all 55 were determined by inverse PCR and represent a unique collection of 4th chromosome P-element insertion lines which will be made available from the Bloomington Drosophila Stock Center. One line, P114, maps within 3 kb of the Eph transcription start site and ∼1 kb upstream of onecut, a gene transcribed on the opposite strand encoding a homeodomain class transcription factor(Nguyen et al., 2000). As both Eph and onecut expression are unaffected in homozygous P114 flies (data not shown), we used this line to generate Eph deletions by imprecise excision. A total of 1300 excision lines were generated, 33 of which, or 2.5%, were found to be homozygous lethal. Both viable and lethal excision lines were assayed by PCR and southern blot analysis for rearrangements within the Eph genomic region. A single viable excision line, Eph^x652, was isolated that removes Eph genomic sequences (Fig. 1A). This excision deletes the first three exons of Eph,thus removing the Eph transcriptional start site and the first 79 amino acids of the coding region. The absence of detectable Eph mRNA and protein expression in Eph^x652 homozygous individuals argues that it acts as a null mutation(Fig. 1C; see below).

Sequencing across the breakpoint of Eph^x652 revealed that this excision also removes 5′ sequences of the onecutlocus, including its transcription start site. It does not, however, remove any onecut coding sequence. Among the excision lines, we identified 15 that extend in the direction of onecut, removing onecut-coding sequence but not removing sequence in the direction of or within the Eph gene (Fig. 1A). All of these excision lines are lethal as homozygotes. In addition, intra-allelic combinations of the excisions, such as onecut^lx122/onecut^lx49, are lethal,demonstrating an essential function for the onecut locus. In contrast to the lethality of the onecut^lx122 and onecut^lx49 alleles, homozygous Eph^x652flies are viable and fertile, and have no obvious morphological defects,suggesting that any effect of Eph^x652 on onecutfunction must be partial. Consistent with this, Eph^x652complements the lethality of the onecut alleles, as Eph^x652/onecut^lx122 and Eph^x652/onecut^lx49 flies are viable and fertile.

The single Drosophila Ephrin gene contains a P element insertion,KG09118, generated by the Berkeley Gene Disruption Project, that lies ∼20 bp downstream of the Ephrin transcription start site(Bellen et al., 2004; Roseman et al., 1995). In situ hybridization of Ephrin antisense RNA probes to homozygous Ephrin^KG09118 embryos reveals that this insertion severely reduces Ephrin mRNA expression(Fig. 1F), indicating that this insertion is an Ephrin allele. Like Eph^x652,homozygous Ephrin^KG09118 individuals are viable and fertile.

Previous in situ hybridization studies showed that Eph is expressed within the CNS at a time when neurons extend axons towards their targets and is expressed by most, if not all, CNS neurons at this stage(Scully et al., 1999). To visualize the distribution of Drosophila Eph receptor, we used Ephrin-Fc, a soluble probe consisting of the extracellular domain of Manduca Ephrin fused to the human immunoglobulin Fc fragment(Kaneko and Nighorn, 2003). The major axon tracts within the CNS, the bilaterally symmetric longitudinal connectives and the two commissures connecting each hemisegment, can be visualized with the panaxonal BP102 antibody(Seeger et al., 1993). Incubation of unfixed, dissected embryos with BP102 and Ephrin-Fc reveals that the Eph receptor is expressed on both longitudinal and commissural axons within the CNS in a pattern similar to that of BP102 staining(Fig. 2A-C). Thus, the Eph receptor is expressed by most, if not all embryonic CNS neurons and is targeted to axons.

Binding of Ephrin-Fc is abolished in homozygous Eph^x652mutants (Fig. 2D,E),demonstrating the specificity of Ephrin-Fc binding to the Eph receptor and confirming that Eph^x652 is null for Ephexpression. By contrast, Ephrin-Fc binding is unaffected in embryos homozygous for the onecut alleles onecut^lx122 and onecut^lx49 (Fig. 2F), further demonstrating that the lethality associated with the onecut alleles is not due to compromised Eph function.

Ephrin mRNA, like Eph, is widely expressed throughout the embryonic CNS, as assayed by in situ hybridization(Fig. 1F) and Ephrin protein is expressed by most if not all CNS neurons, but in contrast to Eph has been reported to be localized to cell bodies rather than axons(Bossing and Brand, 2002). Consistent with this, we have found that when expressed by neurons using the GAL4/UAS transactivation system (Brand and Perrimon, 1993), a myc-tagged version of Ephrin is localized to cell bodies (Fig. 2G-I). Therefore, both Drosophila Eph and Ephrin are present at a time and place that would allow them to control axon pathfinding, but each appears to reside within a distinct compartment of the neuron.

In vertebrates, Eph activation by Ephrin often leads to axon repulsion. To test for repellent activity of Eph/Ephrin signaling, we misexpressed Ephrin in midline glia and assayed the effects on commissural axons that normally cross the midline. For these experiments, we used sim-Gal4 plus UAS-Ephrin:myc to drive expression of myc-tagged Ephrin exclusively within midline glia. In these embryos, many commissural axons fail to cross the midline and thus fail to form normal commissures, similar to what has been reported previously (Fig. 3B)(Bossing and Brand, 2002). These defects are consistent with Ephrin acting as an axonal repellent,preventing axons from crossing the ectopic source of Ephrin at the midline. Similar axon guidance defects are observed when an Ephrin transgene lacking all intracellular sequences, UAS-Ephrin^ΔIC:myc, is expressed by sim-Gal4 (Fig. 3C). This result, combined with the finding that misexpression of Eph by midline glia has no effect on commissural axons (data not shown), confirms that it is forward signaling that causes axons to be repelled and that reverse signaling plays no role in this repulsion. To test whether the repellent activity of Ephrin is mediated by Eph, we misexpressed Ephrin in midline glia using the same combination of transgenes as above, but in an Eph^x652mutant background. In these embryos, the ability of Ephrin to repel commissural axons is abolished, restoring the commissural tracts to wild type in their appearance (Fig. 3D). Thus, Eph and Ephrin are able act as a receptor/ligand pair in vivo and, at least in this assay, Ephrin acts solely through the Eph receptor.

The expression of Eph and Ephrin within the embryonic CNS and the severe guidance defects observed when Eph/Ephrin signaling is ectopically activated suggest a role for these signaling molecules in axon guidance during normal development. To assess this, we first examined the overall architecture of the embryonic CNS of homozygous Eph^x652 mutant embryos using the BP102 antibody to label all axons(Seeger et al., 1993). As homozygous Eph^x652 flies are viable and fertile, we took embryo collections from a homozygous mutant stock, thus eliminating both zygotic and maternal Eph. Homozygous Eph^x652 mutant embryos show no obvious defects in overall axon tract organization compared with wild-type embryos (Fig. 4A,B).

We next assayed axon pathfinding in greater detail using markers that label the axonal projections of subsets of neurons. First, we stained embryos for Fasciclin 2 (Fas2), a cell-adhesion molecule expressed on five distinct bilaterally symmetric bundles of axons that run along the anteroposterior axis of the CNS (Lin et al., 1994). We detected no obvious difference in the Fas2 axon bundles between wild-type and Eph^x652 mutant embryos, although the bundles in Eph^x652 mutants appear slightly less fasciculated(Fig. 4C,D). We next examined the axon trajectories of the Apterous (Ap) neurons, three neurons per abdominal hemisegment that extend axons anteriorly as a tightly fasciculated bundle along the most medial Fas2 bundle(Lundgren et al., 1995; Simpson et al., 2000). The axonal projections of the Ap neurons in Eph^x652 mutant embryos are indistinguishable from wild type(Fig. 4E,F). In addition to the Ap neurons, we examined the trajectories of other subsets of neurons such as the Eagle neurons (Bonkowsky et al.,1999), the VUM neurons(Callahan and Thomas, 1994) and the MP1 neurons (Hidalgo and Brand,1997). In all cases, we could detect no axon guidance defects in Eph^x652 mutant embryos (data not shown).

We conclude from these results that although activation of Eph signaling by ectopic Ephrin expression results in axon repulsion, Eph signaling plays only a limited role, if any, in embryonic axon guidance. Consistent with this, we found no obvious axon pathfinding defects in Ephrin^KG09118mutant embryos (data not shown).

Studies in vertebrates have implicated a role for Eph/Ephrin signaling in the development of the olfactory system. In Drosophila, olfactory neurons within the antennae and maxillary palp project to glomeruli within the antennal lobe. Projection neurons, which form dendritic connections with antennal glomeruli, in turn project their axons to the dendritic region of the bilaterally symmetric mushroom bodies (MB) where olfactory information is processed (Crittenden et al.,1998; Heisenberg,2003; Jefferis et al.,2002).

The neurons giving rise to the distinct lobes of the MB are clonally related (Ito et al., 1997; Lee et al., 1999). Four neuroblasts set aside early in development generate a pool of MB neurons in three distinct mitotic waves, giving rise in sequence to the larval bornγ neurons, the early pupal α′/β′ neurons and finally the later born α/β neurons(Fig. 5A). The initial axonal projections of γ neurons prefigure those of the later bornα′/β′ and α/β neurons. They send axons anteriorly as a tightly packed bundle within a structure known as the peduncle. Axons then bifurcate forming a dorsal and medial branch seen in late embryo/early 1st instar through 3rd instar larval stages. These axonal processes undergo degeneration during metamorphosis, only to re-extend a single axonal projection medially during puparation to give rise to the adult structure (Jefferis et al.,2002).

The second group of neurons, the α′/β′ neurons are generated during late 3rd instar. Like γ neurons,α′/β′ neurons extend an axon through the peduncle, and bifurcate, sending a distinct projection medially, along with a second collateral extension dorsally, generating the α′ and β′lobes. Finally, similar projections by the third group of MB neurons, theα/β neurons, result in the formation of an additional, distinct dorsal and medial projection, forming the α and β lobes. Theα and β lobes are further characterized by strong expression of Fas2 (Crittenden et al., 1998; Noveen et al., 2000).

As revealed by Ephrin-Fc staining, we found that Eph is expressed within the MB, specifically within MB neurons and is targeted to their axons throughout MB development. At late embryo/early 1st instar larval stages, when MB γ neurons already display their characteristic bifurcated appearance,we detect Eph expression by most neurons of the brain, including the developing MB γ neurons (Fig. 5B-D). As in the embryonic CNS, Eph is targeted to axons. By 3rd instar larval stages, Eph staining within the brain becomes restricted to the bifurcated γ neurons (Fig. 5E-G). At early pupal stages, levels of Eph expression within the MB increase and the receptor is present on α′/β′ axons(Fig. 5H-J). Eph receptor is present throughout the MB peduncle, as well as on both dorsal and medial lobe projections at pupal stages. Expression persists as development proceeds, but becomes limited primarily to the α/β neurons at late pupal stages(Fig. 5K-M) and the adult(Fig. 5N-P), as revealed by the colocalization with Fas2. Interestingly, in the adult MB, Eph levels are higher on the dorsal projecting α lobes, particularly within their terminal regions (Fig. 5O,arrow). Thus, Eph receptor is restricted to specific lobes within the MB in the adult and its localization within a single α/β neuron is tightly regulated.

We have examined Ephrin expression by in situ hybridization on larval brains at a time when MB neurons bifurcate, forming dorsal and medial lobes. Ephrin expression within the brain at this time is detectable in many subsets of neurons including MB neurons(Fig. 5S, arrows). These results demonstrate that MB neurons, like embryonic neurons, co-express both ligand and receptor during MB development. Furthermore, as in embryonic neurons, Ephrin:myc is localized to MB cell bodies when expressed by OK107-Gal4 (Fig. 5Q,R), a Gal4 driver expressed specifically by the MB throughout its development (Connolly et al.,1996; Kurusu et al.,2002). Therefore, both Drosophila Eph and Ephrin are present within developing MB neurons, but each appears to reside within distinct neuronal compartments.

To examine the overall structure of the adult MB in both wild type and Eph mutants, we used elav-Gal4 plus UAS-mCD8:GFP. This combination of transgenes labels all five lobes of the adult MB(Fig. 6A,B); the α andβ lobes can be specifically visualized by double labeling with Fas2. We found that Eph^x652 adult flies have severe MB defects(n=50). The most dramatic phenotype is reduced or absentα′ and α lobes, often with an associated increase in the thickness of β′ and β lobes(Fig. 6B,K). This phenotype is present in 60% of adult flies, and primarily exhibits itself unilaterally within the MB. Rarely do we detect severe bilateral defects, although in these cases similar reduced or absent dorsal lobe projections are observed. The severity of MB α lobe projection defects covers a continuous range, from lobes completely missing to lobes that appear normal. Defects in α lobe projections are always associated with defects in the earlier formingα′ lobe, suggesting that α′/β′ neurons may act as pioneers for the later born α/β projections as has been proposed (Wang et al., 2002). MB development in Eph^x652/onecut^x122and Eph^x652/onecut^x49 heterozygotes(n=26) is indistinguishable from wild type(Fig. 6I), indicating that the phenotypes we observe within the MB of Eph^x652 mutants are specifically due to loss of Eph.

Defects in dorsal lobe development of Eph mutants are evident at all stages of MB development. Defects are detected in developing αneuron dorsal projections in early pupae (6C,D), as well as in dorsal projecting branches of the earlier born γ neurons(Fig. 6E-H). By the 3rd instar larval stage, wild-type γ neurons have acquired a bifurcated appearance,generating a dorsal and medial projection(Fig. 6E). In ∼60% of Eph^x652 mutant 3rd instar larvae, we see either reduced or absent dorsal lobe projections (Fig. 6F). Many of these same MBs show obviously thicker medial lobes in association with dorsal lobe defects (Fig. 6F, arrowhead). Similar defects in dorsal lobe projections are apparent within developing late embryonic/early 1st instar larval MBs, at a time when γ neurons are first pathfinding (6G-H). Thus, Eph function is required at all stages of MB development.

The role for Eph signaling in MB development is further supported by the phenotypes we observe in Ephrin mutants. In 40% of homozygous Ephrin^KG9118 adult flies (n=30), we see defects in α lobe projections identical in nature to those of Eph^x652 individuals(Fig. 6J). Taken together,these results indicate a requirement for Eph/Ephrin signaling in the development of dorsal lobe projections within the MB.

To confirm that lack of Eph expression and function in MBs results in the dorsal lobe defects, we attempted to rescue the MB defects of Eph^x652 homozygous adult flies by replacing Eph expression using the Gal4/UAS system. For these experiments, we used elav-Gal4,which is expressed in all postmitotic neurons throughout development, and the MB-specific OK107-Gal4 driver. Expression of Eph within the MB using either of these drivers fails to rescue the α/α′ lobe defects. In fact, expression of Eph within the MB of either wild type or Eph^x652 mutants results in a phenotype similar to the Eph loss of function phenotype. In these individuals, there is a high frequency and often bilateral loss of α lobes and a concomitant increase in β lobe thickness (Fig. 6K). In ∼50% of these MBs, β lobes appear fused at the midline (arrowhead), signifying that medially projecting branches now fail to respect the midline boundary. Such failure to respect the midline boundary is also observed in 10% of Eph^x652 mutant MBs (arrowhead in Fig. 6L). These results argue that MB neurons are sensitive to levels or timing of Eph expression.

MB neurons may require Eph/Ephrin signaling to bifurcate and generate both medial and dorsal lobe projections. Alternatively, MB neurons may bifurcate normally in the absence of Eph function, but fail to extend a branch into the dorsal lobe and instead extend both branches in the medial lobe. To distinguish between these possibilities, we visualized individually marked mutant neurons within the MB.

We visualized individual neurons and their projections using the MARCM system (Lee and Luo, 1999), in which clones of cells are generated by mitotic recombination and identified by expression of UAS-mCD8:GFP. We randomly generated marked clones in wild type and in homozygous Eph^x652 mutants at late pupal stages, thus labeling the trajectories of α/β neurons. Ideally, we would have generated Eph^x652 mutant neurons in an otherwise phenotypically wild-type animal, thus determining the autonomous function of Eph within MB neurons. However, owing to the difficulties in generating clones by recombination of the 4th chromosome, we were limited to examining the fate of individual MB neurons within homozygous Eph^x652 mutant individuals.

In wild type, individual α/β neurons extend a single axonal projection into the peduncle that eventually bifurcates, one branch extending medially in the β lobe, and the other dorsally in the α lobe(Fig. 7A,B). Little or no additional branching is observed along the length of each projection, although limited arborization at the end of each lobe can occur. Individualα/β neurons in Eph^x652 adults project a single axon along the length of the peduncle as in wild type(Fig. 7C, inset) and at the base of the peduncle mutant neurons bifurcate normally. However, instead of projecting one branch dorsally along the α lobe, both branches extend medially along distinct paths within the β lobe(Fig. 7C,D). In addition,supernumerary branches along the lengths of the β lobe projections are present in Eph^x652 mutants (arrowheads in Fig. 7D). Thus, Eph/Ephrin signaling is not required for branch formation, but for correct pathfinding of dorsal lobe branches within the developing MB.

In Eph^x652 MBs that completely lack the dorsal lobe,all of the neurons in a clone show projection defects (n=15). In Eph^x652 mutant MBs where the α lobe is present but reduced, marked MB neurons either project normally or show α lobe projection defects (n=8). In larger clones within such MBs, we observe more labeled neurons projecting axons within the β lobe than theα lobe (Fig. 7E,F),indicating that some neurons within a single Eph^x652mutant MB can extend branches normally within dorsal and medial lobes, while other neurons do not, instead extending both branches within the medial βlobe. These results argue that individual branches of MB neurons make independent pathfinding decisions. Furthermore, these results are consistent with a continual requirement of Eph signaling within MB neurons for the correct guidance of individual branches.

In Eph^x652 mutants in which the overall MB structure appears normal, individually marked neurons within all clones examined extend branches normally in α and β lobes(Fig. 7G,H; n=18). In these MBs, individually marked neurons also have a normal unbranched appearance, suggesting that the ectopic branching observed in phenotypically mutant MBs may be a consequence of incorrect pathfinding and target selection of the α dorsal branch. It is possible that within these overall normal-looking MBs there are neurons with abnormal projections, but their numbers are sufficiently small that the probability of one being included in a marked clone is very low.

Vertebrate Eph RTKs and their Ephrin ligands have been shown to play roles in many aspects of axon guidance, including midline crossing at the optic chiasm, retinotopic map formation and motoneuron innervation of target muscles within the limb (Klein, 2004; Palmer and Klein, 2003; Poliakov et al., 2004). Given the simplicity of Drosophila Eph/Ephrin, with its single Eph receptor and single Ephrin ligand, each sharing similarities to both A and B vertebrate subclasses, one might have expected the role of this prototypical Eph/Ephrin complex to reflect the most basic axon guidance functions of this family of proteins. Surprisingly, our analysis of Eph mutants shows that Eph/Ephrin signaling has a very specific requirement in MB development rather than a general role in axon guidance. Although the receptor/ligand pair is expressed in the embryonic CNS and Eph is capable of signaling when ectopically activated, this signaling appears to have a limited role in the guidance of axons. Similarly, although Drosophila Eph is expressed within the developing visual system, and RNAi knockdown experiments previously suggested a role in retinotopic map formation(Dearborn et al., 2002), we detect no defects in the overall targeting of photoreceptor cells in the lamina and medulla in the absence of Eph function (not shown). Thus,in these developing systems, Eph/Ephrin signaling either acts with other guidance molecules or functions in processes other than axon guidance.

Our analysis of the embryonic CNS in Eph-null mutants shows that Eph signaling plays little or no role in axon guidance. Although we have not examined all guidance events, and in fact subtle guidance and targeting defects may yet be uncovered, the lack of any obvious defects is in marked contrast to a previous study using RNAi to knock down Eph expression(Bossing and Brand, 2002). RNAi is a widely used method to suppress gene expression in many organisms,including Drosophila. Although it is generally accepted that RNAi acts specifically to downregulate target gene expression because of its requirement for near perfect sequence identity with the target transcript,there is evidence that interference can occur with sequences carrying only limited sequence similarity (Jackson et al., 2003). The Eph receptor contains regions of sequence similarity, such as the fibronectin type III repeats, to other proteins expressed within the nervous system(Scully et al., 1999). Perhaps the disparity between our results and the results using RNAi is due to expression changes in genes other than Eph that contain sequences capable of binding the Eph RNAi.

Our analysis of Eph mutants reveals a specific role for Eph signaling in regulating the guidance of individual axon branches within the MB. As in wild type, growing axons of MB neurons in Eph mutants bifurcate at the base of the peduncle, forming two branches. However, these two branches often fail to choose divergent paths and instead both extend medially. This is the first evidence for receptor signaling in the guidance of individual branches of MB neurons. Other receptors such as the Down syndrome cell-adhesion molecule (Dscam) and signaling molecules such as the Rho family of GTPases appear to have broader roles in regulating MB axon fasciculation,branch formation, extension and segregation(Schmucker et al., 2000; Wang et al., 2002; Zhan et al., 2004). For example, Dscam is required for MB neuron branch segregation(Wang et al., 2002; Zhan et al., 2004) through a mechanism thought to involve the mutual repulsion of sister branches(Wang et al., 2002; Wojtowicz et al., 2004). In the absence of Dscam, when branches form, they extend randomly in either dorsal or medial directions and can be fasciculated. By contrast, Eph has a more specific role in the guidance of dorsal projections and sister branches do not tightly associate with each other even when both project medially. Therefore, the mechanisms that ensure the formation and maintenance of distinct axon branches are unaffected in Eph mutants and we would predict that Dscam functions normally in Eph mutant MB neurons.

How branch formation occurs in the Drosophila MB is unclear, as branching MB neurons have yet to be followed in situ and in real time. Individual MB growth cones might respond to specific cues by splitting,resulting in formation of two growth cones that project independently, a possibility supported by the relative synchrony of branch formation(Wang et al., 2002). Alternatively, branches might arise as collateral extensions off existing axons (Noveen et al., 2000). Regardless of the precise mechanism of branching within the MB, sister branches must extend away from each other, one in the dorsal lobe and one in the medial lobe. Although Eph is clearly required for the proper guidance of dorsal branches, the mechanism by which it acts is unknown. In one model, Eph receptor signaling, in response to a localized Ephrin source, acts as an attractive signal to guide the extension of one sister branch in a dorsal direction. In the absence of this attractive cue, both branches extend in the`default', or medial, direction. Alternatively, Eph/Ephrin signaling may act as a repellent guidance signal, in which case Eph-mediated repulsion could steer the future dorsal branch away from the midline, allowing its extension dorsally. If branching does occur by growth cone splitting, both of these models call for the differential localization or activation of the Eph receptor within a single branch, and that this spatially localized signaling leads to the guidance of this branch dorsally. In this regard, we have not detected any clear differences in Eph receptor distribution between dorsal and medial lobes at a time when MB neurons are in the process of projecting axons. However, like branch initiation itself, differential Eph expression could be a transient event and difficult to visualize by examining the entire collection of bifurcating MB neurons.

At present, the precise localization of the Ephrin ligand during branch formation is not known. Our studies of Ephrin RNA expression,however, indicate that, as in the embryonic CNS, MB neurons co-express both receptor and ligand. Therefore, the means by which Eph/Ephrin signaling regulates branch extension dorsally may be more complicated than Eph-expressing MB neurons simply responding to Ephrin in their environment. There are now a growing number of instances where Eph receptors and their corresponding ligands are co-expressed within a given tissue(Knoll and Drescher, 2002; Marquardt et al., 2005). This co-expression allows, at least in theory, the possibility for both trans and cis regulation of the Eph/Ephrin signaling complex, as well as for both forward and reverse signaling in the same cell(Santiago and Erickson, 2002). The ultimate path that a growing axon takes therefore, may depend on the balance between these various signals.

The requirement for Eph and Ephrin in guiding branches within the developing Drosophila MB is reminiscent of the role of Eph/Ephrin signaling in the retinotopic mapping of retinal ganglion cells within the midbrain of vertebrates (Hindges et al.,2002; McLaughlin et al.,2003; Poliakov et al.,2004) and layer-specific branching of thalamic axons within the cortex (Mann et al., 2002). In the retinotectal system, initial imprecise axon extensions of retinal ganglion cells into the midbrain are refined by extension of specific collateral branches that resolve into synaptic arbors along appropriate retinotopic locations along both the anteroposterior and dorsoventral axes. EphA/Ephrin A signaling appears to control, in part, the point where branches form along the AP axis of the midbrain, while graded EphB/Ephrin B cues guide branch extension to appropriate locations along the DV axis. Here, as in the MB of Drosophila, Eph/Ephrin signaling controls not the process of branching itself, but the guidance of branches to appropriate target areas,suggesting a conserved function for Eph/Ephrin in guiding branch extensions. Drosophila MB neurons, with their stereotyped branch formation and extension may provide a system for understanding the mechanisms whereby Eph/Ephrin signaling activates intracellular signaling cascades that ultimately lead to the guidance of individual branches to appropriate targets.

We thank past and present members of the Thomas laboratory for advice;Cynthia Hughes, Renee Read and Shingo Yoshikawa for critical reading of the manuscript; the Goodman laboratory and Brian McCabe for generously providing fly stocks and antibodies; Shingo Yoshikawa for scrt-Gal4; and the Bloomington Drosophila Stock Center for providing fly stocks. This work was supported by an American Cancer Society grant to M.B., and by grants from the NIH to J.B.T.