Linking Cell Fate, Trajectory Choice, and Target Selection: Genetic Analysis of Sema-2b in Olfactory Axon Targeting

Summary

Neural circuit assembly requires selection of specific cell fates, axonal trajectories, and synaptic targets. By analyzing the function of a secreted semaphorin, Sema-2b, in Drosophila olfactory receptor neuron (ORN) development, we identified multiple molecular and cellular mechanisms that link these events. Notch signaling limits Sema-2b expression to ventromedial ORN classes, within which Sema-2b cell-autonomously sensitizes ORN axons to external semaphorins. Central brain-derived Sema-2a and Sema-2b attract Sema-2b-expressing axons to the ventromedial trajectory. In addition, Sema-2b/PlexB-mediated axon-axon interactions consolidate this trajectory choice and promote ventromedial axon bundle formation. Selecting the correct developmental trajectory is ultimately essential for proper target choice. These findings demonstrate that Sema-2b couples ORN axon guidance to postsynaptic target neuron dendrite patterning well before the final target selection phase, and exemplify how a single guidance molecule can drive consecutive stages of neural circuit assembly with the help of sophisticated spatial and temporal regulation.

Introduction

Neural circuit assembly relies on the coordinated efforts of diverse developmental processes. Neurons must first acquire distinct fates, which eventually determine their wiring specificity and physiological properties. Axons then navigate along specific pathways towards their target cells, often across long distances. Finally, axons choose specific synaptic partners within the target zone. While significant progress has been made during the past few decades in our understanding of each of these steps (reviewed in Sanes and Yamagata, 2009; Jukam and Desplan, 2010; Kolodkin and Tessier-Lavigne, 2011), less is known about the cellular and molecular mechanisms that seamlessly coordinate such distinct developmental processes.

The olfactory system relies on highly organized inputs from diverse classes of olfactory receptor neurons (ORNs), and thus offers an excellent context to explore the relationships between cell fate, axon pathway choice, and target selection. In Drosophila, 50 classes of ORNs, most of which express a single odorant receptor, target their axons precisely to 50 corresponding glomeruli in the antennal lobe (Couto et al., 2005; Fishilevich and Vosshall, 2005; Silbering et al., 2011). ORN axons synapse on projection neuron (PN) dendrites, most of which arborize within a single glomerulus (Stocker et al., 1990; Jefferis et al., 2001). Thus, a key feature in the olfactory circuit is the precise establishment of one-to-one pairs between 50 ORN classes and 50 PN classes. PN dendrites pattern the developing antennal lobe first. By 18 hours after puparium formation (18h APF), when pioneering ORN axons reach the developing antennal lobe, dendrites of individual PNs already occupy specific areas within the antennal lobe corresponding to their future glomerular positions (Jefferis et al., 2004). The secreted semaphorins Sema-2a and Sema-2b are expressed in a gradient within the antennal lobe and signal through transmembrane Sema-1a to instruct PN dendrite targeting along the dorsolateral-ventromedial axis (Komiyama et al., 2007; Sweeney et al., 2011). Local binary determinants such as Capricious further segregate PN dendrites into discrete glomeruli (Hong et al., 2009). Several distinct mechanisms of ORN axon targeting have also been identified. For example, Sema-1a also mediates repulsive axon-axon interactions to segregate ORN axons from different sensory organs (Lattemann et al., 2007; Sweeney et al., 2007). Hedgehog signaling coordinates peripheral ORN cell body position with antennal lobe glomerular targeting (Chou et al., 2010). Teneurin-mediated homophilic attraction matches PN dendrites with corresponding ORN axons during final target selection (Hong et al., 2012). With the exception of Teneurin-mediated synaptic partner matching, it remains unclear how and when axon-derived and target-derived cues cooperate, or how the development of PN dendrites and ORN axons is coordinated.

ORNs expressing a specific odorant receptor exhibit characteristic olfactory responses (Hallem and Carlson, 2006), which are sent to particular glomeruli and relayed by postsynaptic target PNs to stereotyped areas in higher olfactory centers (Jefferis et al., 2007). ORNs are housed in specific sensilla in the 3^rd segment of the antennae and in the maxillary palps. Most sensilla contain 2–3 individual ORNs belonging to distinct classes. The Notch pathway diversifies ORN fates within individual sensilla (Endo et al., 2007). Loss of mastermind (mam), a transcriptional co-activator that mediates Notch signaling, causes ORNs to adopt a “Notch-OFF” fate, whereas loss of numb, an antagonist of Notch signaling, causes ORNs to adopt a “Notch-ON” fate. Notch-OFF and Notch-ON ORN classes target their axons to distinct glomeruli. However, the specific molecules and mechanisms that mediate such targeting are currently unknown. By analyzing Sema-2b function in ORN axon targeting, we now report new molecular and cellular mechanisms that link ORN cell fate, trajectory choice, and target selection. Our findings also demonstrate that Sema-2b couples PN dendrite patterning with ORN axon guidance well before the final target selection phase, thus newly connecting previously disparate steps in olfactory circuit wiring.

Results

Sema-2b Is Selectively Expressed by the Ventromedial ORN Axon Bundle

While examining the role of secreted semaphorins in PN dendrite targeting (Sweeney et al., 2011), we observed a striking distribution pattern for Sema-2b in adult ORNs. Pioneering antennal ORN axons first contact the ventrolateral edge of the developing antennal lobe at 18h APF (Jefferis et al., 2004). They then circumscribe the ipsilateral antennal lobe and navigate across the midline before invading both the ipsi- and contralateral antennal lobes. At 24h APF, ORN axons labeled with the pan-ORN pebbled-GAL4 driver (Sweeney et al., 2007) took either a ventromedial or dorsolateral trajectory around the antennal lobe (Figure 1A₁). ORN axons thus segregated into two distinct bundles of roughly equal size, with a mean dorsolateral/ventromedial ratio of 0.90 (Figure 1H). Notably, Sema-2b protein was highly enriched in the ventromedial axon bundle but was undetectable from the dorsolateral bundle (Figure 1A₂). It was also present within the ventromedial antennal lobe (Figure 1A₂, arrow) where it is contributed by a central source (Sweeney et al., 2011). Based on this highly specific distribution pattern, we hypothesized that Sema-2b plays a crucial role in ORN axon trajectory choice.

Figure 1. Sema-2b Is Selectively Expressed in Ventromedial ORNs and Specifies Their Axon Trajectories.

(A) A₁, developing ORN axons choose either a ventromedial or dorsolateral trajectory (yellow arrowheads) as they circumnavigate the antennal lobe (AL) at 24h APF; all ORN axons are labeled by pebbled-GAL4 driving UAS-mCD8:GFP. A₂, only ventromedial axons express high levels of Sema-2b protein. Sema-2b is also enriched within the ventromedial antennal lobe (arrow). A₃, merge of A₁ and A₂. In all figures, D = dorsal; L = lateral. Midline is to the left. Scale bar = 10 μm for all images of pupal brains.

(B) In sema-2b^−/− mutants, ORN axons primarily take the dorsolateral trajectory; the ventromedial axon bundle is markedly reduced. The absence of Sema-2b staining in sema-2b mutant (B₂) confirms antibody specificity.

(C and D) ORN-specific expression of UAS-sema-2b (C) or UAS-sema-2b^TM (D) rescues ventromedial axon trajectory in sema-2b^−/− mutants.

(E–G) At 29°C, ORN-specific overexpression of secreted (F) or membrane-tethered (G) Sema-2b in a wild-type background causes more axons to choose the ventromedial trajectory compared to control (E).

(H) Quantification of dorsolateral / ventromedial axon bundle ratio. See Supplemental Methods for details. Boxes indicate geometric mean (middle line) and 25–75% range, while whiskers indicate maximum and minimum values. Geometric means/sample sizes are as follows: 25°C control: 0.90/16; sema-2b^−/−: 4.77/33; sema-2b^−/−, ORN > sema-2b: 0.97/14; sema-2b^−/−, ORN > sema-2b^TM: 0.81/27; 29°C control: 0.74/16; ORN > sema-2b: 0.45/24; ORN > sema-2b^TM: 0.46/25. Inset: Quantification of pan ORN > Sema-2b overexpression levels (in a sema2b^−/− background) in ventromedial (VM) vs. dorsolateral (DL) axon bundles for Figure 1C. Sema-2b fluorescence intensity was normalized to mCD8:GFP fluorescence. *** p < 0.001; * p < 0.05, one-way ANOVA with Bonferroni’s Multiple Comparison Test.

Sema-2b Acts in ORNs to Specify Ventromedial Trajectory Choice

Indeed, in 24h APF sema-2b homozygous mutant (sema-2b^−/−) brains, the dorsolateral axon bundle was enlarged at the expense of the ventromedial bundle, shifting the average dorsolateral/ventromedial ratio to 4.77 (Figure 1B and 1H). Sema-2b is thus required for ventromedial trajectory choice. Given that both the central brain and peripheral ventromedial ORNs contribute Sema-2b (Figure 1A₂), we next tested its cell-type specific function by using pebbled-GAL4 to drive UAS-sema-2b expression in sema-2b^−/− mutants. This restored the ventromedial axon bundle size to wild-type (WT) levels (Figure 1C and 1H; mean ratio=0.97), demonstrating that ORN-derived Sema-2b alone is sufficient to rescue the ventromedial axon trajectory defects of sema-2b^−/− mutants in the absence of central Sema-2b. A membrane-tethered version of Sema-2b (Sema-2b^TM; Wu et al., 2011) also rescued trajectory with similar efficiency (Figure 1D and 1H; mean ratio=0.81), suggesting that secretion is not obligatory for Sema-2b function. Sema-2b thus acts locally at the membrane rather than as a long-distance secreted cue.

While pan-ORN Sema-2b expression restored the dorsolateral/ventromedial axon bundle ratio, we observed that Sema-2b protein was not expressed equally between the two bundles, with three-fold lower expression in the dorsolateral bundle relative to the ventromedial bundle (Figure 1C₂; quantified in Figure 1H inset). This may have facilitated rescue, and suggests that endogenous post-transcriptional mechanisms down-regulate Sema-2b in dorsolateral ORNs (see Discussion).

We further tested whether ORN-specific Sema-2b overexpression could re-direct dorsolateral axons to the ventromedial trajectory. Because overexpression at 25°C did not cause significant phenotypes (data not shown), likely due to the post-transcriptional down-regulation in dorsolateral ORNs, we raised experimental flies at 29°C to increase UAS-Sema-2b transcription levels. Under these conditions, secreted and membrane-tethered Sema-2b both biased axons towards the ventromedial trajectory, in some cases almost completely removing the dorsolateral bundle (mean dorsolateral/ventromedial ratio = 0.45 and 0.46 for secreted and membrane-tethered Sema-2b, respectively). Sufficiently high levels of Sema-2b can thus force dorsolateral axons to choose the ventromedial trajectory. Taken together, our expression, loss- and gain-of-function studies indicate that Sema-2b is required in ventromedial ORNs for their proper trajectory choice, and plays an instructive role in specifying ventromedial trajectory choice.

Glomerular Targeting Is Differentially Affected in sema-2b Mutants Based On ORN Class

What are the consequences of incorrect trajectory choice for ORN class-specific glomerular targeting? To systematically examine this question, we labeled individual ORN classes in the adult brain with odorant receptor (Or) promoter-driven GAL4 lines. In WT, Or92a-GAL4 and Or67b-GAL4 each label a “ventromedial” ORN class that sends its axons along a ventromedial trajectory and targets to the VA2 and VA3 glomeruli in the ventromedial antennal lobe, respectively (Figure 2A, top row). In sema-2b^−/− brains, the majority of axons for both ORN classes took dorsolateral trajectories around the antennal lobe, and terminated in specific ectopic loci in the dorsolateral antennal lobe (Figure 2A, bottom). Residual axons that took the ventromedial trajectory terminated in approximately correct target regions.

Figure 2. Sema-2b Is Required for Trajectory Choice and Glomerular Target Selection of Ventromedial ORNs.

(A–C) Glomerular targeting of axons (labeled in green) from specific ORN classes in control (top panels) or sema-2b^−/− mutant (bottom panels) adult brains. nc82 staining labels the antennal lobe neuropil (magenta). ORN classes and their glomerular targets (in parentheses) are indicated above the images. D, dorsal; L, lateral. Scale bar = 20 μm (for all adult brain images). (A) Axons of the “ventromedial” ORN classes Or92a and Or67b normally take ventromedial trajectories to target the VA2 and VA3 glomeruli, respectively. In sema-2b^−/− mutant brains, the majority of axons aberrantly project dorsolaterally to incorrect targets. (B) Axons of the “dorsolateral” classes Or67d and Or88a normally target to DA1 and VA1d, respectively; their trajectory and targeting remain largely unperturbed in sema-2b^−/− mutants. (C) Axons of the “dorsomedial” classes Or22a and Or47a normally choose ventromedial trajectories (arrowheads) to DM2 and DM3, respectively. In sema-2b^−/− mutants, they aberrantly select dorsolateral trajectories (arrowheads), but target to the correct glomeruli. See Figure S1 for phenotypes of the 19 additional ORN classes analyzed.

(D) Schematic of the adult antennal lobe representing all 25 ORN classes examined in sema-2b^−/− mutants, listed by target glomeruli and separated into three section planes along the anterior-posterior axis. Red, classes with both targeting and trajectory error; blue, unaffected classes; orange, classes with mostly trajectory errors; purple, the CO₂-sensitive ORNs which target to only ipsilateral V glomerulus.

(E) Quantification of phenotypes in sema-2b^−/− mutant brains. Black, percentage of brains exhibiting mistargeted axons; gray, percentage of antennal lobe hemispheres exhibiting ectopic axon trajectories. Neither of these phenotypes was observed in wild-type controls. Number of mutant brains analyzed for each Or class is listed above each column group. Compare with eyFlp MARCM phenotypes in Figure S2. Also see Figure S3 for analysis of mistargeting specificity.

In contrast to the severe phenotypes of ventromedial classes, “dorsolateral” ORN classes exhibited little or no trajectory and targeting defects: both Or67d-GAL4 and Or88a-GAL4 label dorsolaterally projecting ORN classes that target to the DA1 and VA1d glomeruli, respectively. Neither class was affected in sema-2b mutants (Figure 2B), consistent with the absence of Sema-2b expression in the dorsolateral-targeting ORNs during development (Figure 1).

In a third group of ORN classes exemplified by Or22a and Or47a, axons normally choose ventromedial trajectories and target to dorsomedial glomeruli (Figure 2C). Interestingly, their trajectories were severely affected in sema-2b^−/− brains, just as in ventromedial classes, but their glomerular targeting remained largely normal. Glomerular targeting of dorsomedial ORN classes can thus be independent of trajectory choice.

To extend these results more globally, we analyzed 19 additional ORN classes, all of which reinforced the class-specific sema-2b^−/− phenotypes described above (Figure S1). Altogether, 10 different ventromedial classes exhibited severe and highly penetrant trajectory and targeting phenotypes (red, Figure 2D and 2E, Figure S1A), while 8 dorsolateral classes were mostly unaffected and 6 dorsomedial classes exhibited severe trajectory choice phenotypes with only mild targeting defects (blue and orange, Figure 2D and 2E, Figure S1B–C). Furthermore, ORN-specific removal of sema-2b using eyFlp-driven MARCM (Mosaic Analysis with a Repressible Cell Marker; Lee and Luo, 1999; Figure S2C) caused targeting and trajectory phenotypes nearly identical to those of whole animal homozygous mutants, both qualitatively and quantitatively (Figure S2). This demonstrates further that Sema-2b is required in ORNs for axon trajectory and glomerular targeting.

In summary, axon trajectory phenotypes of sema-2b^−/− mutants have class-specific consequences for final target selection. For dorsolateral ORN classes, glomerular targeting is normal since Sema-2b was not required for their trajectory choice (Figure 2E, middle). For dorsomedial classes, initial axon trajectory choice is incorrect but glomerular targeting is minimally affected, because target glomeruli are close to the midline commissure region where axons from ventromedial and dorsolateral trajectories reconvene (Figure 2E, right). For ventromedial ORNs, incorrect trajectory choice has devastating consequences for targeting; axons that aberrantly choose dorsolateral trajectories cannot take alternate routes to their targets, but instead terminate in the dorsolateral antennal lobe far away from their normal targets (Figure 2E, left).

Mistargeted ORN Axons Retain Target Specificity When Sema-2b Is Perturbed

Despite the widespread glomerulus distortion or displacement of sema-2b^−/− brains, we observed that mistargeted axons from ventromedial ORN classes projected to specific regions of the antennal lobe rather than random locations. Indeed, in eyFlp MARCM, wayward sema-2b^−/− axons of each ORN class targeted to specific sets of glomeruli that remained identifiable (Figure S3A). These data suggest that axons that choose incorrect trajectories can still select preferred targets, perhaps by responding to local cues in other regions of the antennal lobe. We further tested mistargeting specificity by co-labeling Or67b ORNs with one of three other ORN classes in sema-2b mutants (Figure S3B). In most cases, co-labeled Or92a axons in the mutant brain mistargeted medially but non-adjacently to the ectopic Or67b target site (16/20=80% hemispheres), while Or98a axons mistargeted medially but adjacently (34/40=85% hemispheres, respectively) and Or35a axons mistargeted ventrally to ectopic Or67b axons (30/34=88% hemispheres). In all cases, ectopic axon termini from different ORN classes did not intermingle.

Since Sema-2b overexpression caused a significant decrease in the dorsolateral/ventromedial ratio (Figure 1F, 1H), we predicted that such overexpression would cause adult targeting defects in dorsolateral ORN classes. Indeed, a large fraction of Or67d and Or88a ORN axons mistargeted to the ventral antennal lobe when Sema-2b was overexpressed in all ORNs during development (Figure S3C). Furthermore, these mistargeted axons always projected to immediately adjacent regions much like in wild-type controls (18/18 hemispheres), with Or67d axons medial to Or88a axons in 89% of cases (16/18 hemispheres).

These observations suggest that mistargeted axons in Sema-2b loss- and gain-of-function contexts still exhibit targeting specificity, and can retain relative inter-class spatial relationships. They thus support a sequential model of ORN axon guidance in which Sema-2b primarily specifies initial axon trajectory and thereby constrains target choice, while additional molecular cues subsequently govern targeting to specific glomeruli.

Loss of Plexin-B, the Sema-2b Receptor, Causes Nearly Identical Trajectory and Glomerular Choice Defects

We next explored the cellular mechanisms by which Sema-2b specifies ORN trajectory choice. Plexin B (PlexB) has been identified as a receptor for the secreted semaphorins Sema-2a and Sema-2b (Ayoob et al., 2006; Wu et al., 2011). Evidence from the embryonic ventral nerve cord supports a model in which Sema-2b acts as an attractive cue through PlexB to promote fasciculation of Sema-2b-expressing axons in specific longitudinal bundles (Wu et al., 2011). To test whether PlexB plays a role in ORN axon trajectory choice, we first examined ORN axon trajectories in plexB^−/− mutants at 24h APF. Just as in sema-2b^−/− mutants (Figure 1B), the dorsolateral axon trajectory was enlarged at the expense of the ventromedial trajectory (Figure 3A, compare middle and top rows; quantified in Figure 3B). Adult plexB^−/− mutants also exhibited class-specific glomerular targeting defects nearly identical to those of in sema-2b mutants (Figure 2): ventromedial ORN classes showed severe defects in both trajectory and target choice, dorsolateral classes were unaffected, and dorsomedial classes exhibited severe trajectory choice defects but largely normal target selection (Figure 3C–E; Figure S4). The striking similarities between sema-2b^−/− and plexB^−/− mutant phenotypes suggest that PlexB acts as a Sema-2b receptor in ORN axon trajectory choice, corroborating previous studies in the embryonic ventral nerve cord (Wu et al., 2011).

Figure 3. PlexB Acts in ORNs and Shares Nearly Identical Developmental and Adult Phenotypes with Sema-2b.

(A) As in Figure 1A, developing ORN axons choose a ventromedial or dorsolateral trajectory at 24h APF in controls (top). In plexB^−/− mutants, ORN axons primarily choose the dorsolateral trajectory, with a drastic reduction in the ventromedial axon bundle (middle). ORN-specific expression of UAS-plexB at 25°C rescues ventromedial axon trajectory in plexB^−/− mutants (bottom). All ORN axons are labeled by pebbled-GAL4 driving UAS-membrane-tagged tdTomato (mtdT), while N-Cadherin staining labels the antennal lobe neuropil.

(B) Quantification of dorsolateral / ventromedial axon bundle ratio, as in Figure 1H. Geometric means/sample sizes are as follows: control: 0.90/17; plexB^−/−: 3.45/19; plexB^−/−, ORN > plexB: 1.49/18. * p < 0.05; ** p < 0.01; N.S., not significant; one-way ANOVA with Bonferroni’s Multiple Comparison Test.

(C) Representative ventromedial, dorsolateral, and dorsomedial ORN classes labeled in control (top) or plexB^−/− brains (bottom). As in sema-2b^−/− brains, Or92a and Or67b axons exhibit trajectory and targeting phenotypes, while Or67d axons are largely unaffected, and Or22a axons primarily exhibit trajectory defects only. See Figure S4 for phenotypes of additional classes.

(D–E) Schematic and quantification of trajectory and targeting defects for the 9 ORN classes examined in plexB^−/− mutants, labeled as in Figure 2.

Plexin-B Also Acts in ORNs to Regulate ORN Trajectory Choice

We next asked which cells require PlexB for ORN trajectory choice. None of the anti-PlexB antibodies we produced allowed us to detect endogenous PlexB in the pupal brain. plexB’s location on the 4^th chromosome also precludes MARCM-based mosaic analyses. We therefore tested cell type-specific PlexB requirements using a transgenic rescue approach. Just as we determined for Sema-2b, PlexB expression in all ORNs using pebbled-GAL4 was sufficient to rescue the ORN axon trajectory defects of plexB^−/− mutants at 24h APF (Figure 3A, bottom; quantified in Figure 3D), indicating that PlexB acts ORN-autonomously to regulate axon trajectory choice. These data suggest that Sema-2b and PlexB mediate ORN axon-axon interactions to regulate trajectory choice.

Sema-2b Can Act Cell-Autonomously in ORNs to Specify Ventromedial Trajectory

How can a secreted protein such as Sema-2b act within ORNs to instruct ventromedial trajectory choice of individual axons? The fact that membrane-tethered Sema-2b can fully rescue the sema-2b^−/− trajectory defects (Figure 1D) indicates that Sema-2b acts at short range. Below, we use two experiments to test whether Sema-2b can act cell-autonomously in ORNs to specify ventromedial trajectory.

First, we used hsFLP-based MARCM with late heat-shock timepoints to induce very small sema-2b^−/− clones (Figure 4A). In all six ventromedial classes examined, labeled sema-2b^−/− clones exhibited ectopic dorsolateral trajectories and severe mistargeting phenotypes just as with eyFlp MARCM clones or in whole animal mutants (Figure 4B and Figure S5A; quantified in Figure 4C). This indicates that Sema-2b deletion in a small number of isolated ORNs is sufficient to cause both trajectory and glomerular targeting defects. Our use of specific Or-GAL4 lines identified the class of labeled ORNs and aided phenotypic analysis. However, because each Or-GAL4 only labeled a subset of hsFlp-induced mutant cells (Figure 3A), it remains possible that loss of Sema-2b in unlabeled “background” ORN clones contributed to the observed defects.

Figure 4. Cell-Autonomous and Non-Autonomous Sema-2b functions.

(A) Schematic of loss-of-function hsFlp MARCM. Left: hsFlp MARCM generates small and sparse ORN clones (dotted lines; compare with eyFlp MARCM in Figure S2), a subset of which is labeled with class-specific Or-GAL4 lines. In control MARCM, all cells and thus all clones are WT; Right: in sema-2b^−/− hsFlp MARCM, all labeled cells are sema-2b^−/−, but only a subset of sema-2b^−/− ORNs expressing a given Or-GAL4 is labeled. Green, Or-GAL4-labeled ORNs; grey, sema-2b^−/− clones; red, control (sema-2b^+/- or sema-2b^+/+).

(B) Control (top) and sema-2b^−/− (bottom) hsFlp MARCM small clones of mCD8:GFP-labeled axons (green). Or92a and Or67b Sema-2b^−/− axons aberrantly choose dorsolateral trajectories and mistarget dorsolaterally within the antennal lobe. As with whole animal mutants or eyFlp MARCM, some axons can remain unaffected and target properly. Animals between 0–24h APF were heat-shocked for 10 minutes at 37°C to induce clones.

(C) Quantification of hsFlp MARCM trajectory and targeting phenotypes, labeled analogously to previous figures. Different classes exhibit varying penetrance; Or49b (VA5) and Or98a (VM5v) axons are the most severely affected.

(D) Schematic of the pebbled-GAL4-based MARCM overexpression approach. Left: as before, hsFlp MARCM generates small and sparse ORN clones in a sema-2b^−/− background. Thus, all cells are mutant and a small number of ORNs are labeled with mCD8:GFP. Right: in experimental animals, labeled ORNs are the only cells in the antennae that express UAS-sema-2b^TM. Grey, sema-2b^−/− tissue; green, pebbled-GAL4-labeled ORNs; red, Sema-2b^TM-expressing ORNs.

(E) Representative 24h APF antennal lobes from sema-2b^−/− animals with control (top) or Sema-2b^TM-overexpressing (bottom) ORNs. 0h APF animals were heat-shocked for 5–10 minutes at 37°C to induce clones. Left, ORN axons labeled with mCD8:GFP; Right, overlay of left panel with Sema-2b (red) and NCad (blue) staining; Arrowheads, labeled axons.

(F) Quantification of axon trajectory phenotypes in pebbled-GAL4 hsFlp MARCM. See Experimental Procedures for details. n (antennal lobe hemispheres) = 32 (sema-2b^−/− control), and 41 (sema-2b^−/−, UAS-sema-2b^TM).

(G) Schematic of the reverse eyFlp MARCM approach. Left, WT control eyFlp MARCM as in Figure S2. Right: in sema-2b^−/− eyFlp reverse MARCM, up to 50% of ORNs are homozygous mutant for sema-2b, but only WT ORNs are labeled by class-specific Or-GAL4 lines. Green, Or-GAL4-labeled ORNs; grey, sema-2b^−/− clones; red, control (sema-2b^+/- or sema-2b^+/+).

(H) Control eyFlp MARCM (top) and sema-2b^−/− eyFlp reverse MARCM (bottom) clones of mCD8:GFP-labeled axons (green). In the presence of background sema-2b^−/− clones, WT Or92a and Or67b axons exhibit trajectory and targeting phenotypes similar to those of mutant axons.

(I) Quantification of eyFlp reverse MARCM phenotypes. Most classes exhibited trajectory and targeting phenotypes comparable to those of hsFlp MARCM, with the exception of Or49b and Or98a ORNs, which remained largely normal. Notably, these two classes were the most severely affected in hsFlp MARCM (C). See Figure S5 for MARCM analyses of additional ORN classes and cell body counts for MARCM rescue.

In a complementary experiment, we expressed Sema-2b^TM in small pebbled-GAL4–labeled hsFlp MARCM clones in sema-2b^−/− mutants. In this scenario, only labeled ORNs express Sema-2b (Figure 4D), allowing us to test whether isolated Sema-2b+ ORNs can rescue trajectory phenotypes in sema-2b^−/− mutants. Fasciculation and filopodia extension of developing ORN axons hinders unequivocal axon counting within the antennal lobe. We therefore quantified the number of labeled ORNs by counting their somata within the antenna; by adjusting heat shock timing and duration, we reduced the number of labeled ORNs to an average of ~10 isolated cells per antenna (Figure S5B–C), each of which contains a total of ~1300 ORNs. In a control experiment where all cells were sema-2b^−/−, MARCM-labeled axons predominantly selected the dorsolateral trajectory (Figure 4E, top panels; Figure 4F), as expected from the developmental sema-2b^−/− phenotype (Figure 1B). However, in experimental animals in which all cells were sema-2b^−/− except a few MARCM-labeled ORNs expressing Sema-2b^TM, all labeled axons took the ventromedial trajectory in the majority of cases (Figure 4E, bottom panel; Figure 4F). Remarkably, all of the 9 examples of Sema-2b+ single cell clones chose the ventromedial trajectory, while all 3 examples of single cell clones in controls chose the dorsolateral trajectory.

Given that pebbled-GAL4 labels all ORNs, a full rescue should restore a roughly 1:1 dorsolateral vs. ventromedial axon ratio (Figure 1C). The fact that all Sema-2b+ axons selected the ventromedial trajectory in nearly all experimental animals is more reminiscent of our overexpression experiments (Figure 1F–G). One possible explanation is a difference in overexpression levels: because the UAS-sema-2b^TM transgene is distal to the FRT site used for MARCM, two copies of UAS-sema-2b^TM transgenes were expressed in all MARCM clones. Since most unlabeled (and thus sema-2b^−/−) axons selected the dorsolateral trajectory, another possibility is that reduced competition for limiting central cues or physical space facilitates ventromedial trajectory choice in Sema-2b+ axons. Regardless of the specific mechanisms, sparsely restoring Sema-2b expression in less than 1% of ORNs or even in single ORNs was sufficient for ventromedial targeting, demonstrating that Sema-2b can act cell-autonomously to instruct ventromedial axon trajectory choice (see Figure 7 and Discussion).

Figure 7. Summary of Sema-2b Function in ORN Axon Trajectory.

(A) Sema-2b is expressed only in ventromedial ORN classes (right, red cell) due to its repression by Notch signaling in dorsolateral classes (right, blue cell). Sema-2a and Sema-2b from projection neurons and degenerating larval ORN axons (left, Central Sema-2a/2b, red shading) regulate PN dendrite patterning earlier in development (Sweeney et al., 2011), but remain ventromedially enriched within the antennal lobe and function as orientation cues for incoming Sema-2b+ axons (red), biasing them ventromedially. See expansion of dashed box [1] in Figure 7E, left. D, dorsal; L, lateral.

(B) Sema-2b and its receptor PlexB mediate axon-axon interactions to reinforce the ventromedial bias generated in (A) and promote ventromedial axon bundle formation. See expansion of dashed box [2] in Figure 7E, right.

(C) Proper developmental trajectory choice is essential for final target selection in ventromedial (red) ORN classes. In WT (left), the ventromedial trajectory of Sema-2b+ axons constrains them against selecting possible targets in the dorsolateral region of the antennal lobe (dotted outline). In sema-2b mutants (right), ventromedial ORNs choose dorsolateral trajectories that prevent them from reaching proper targets. Instead, they mistarget to specific dorsolateral target areas, likely in response to cell-surface cues that remain intact. The dorsolateral class (black) is unaffected in sema-2b mutant.

(D) Summary of key experiments and phenotypes for axon trajectory choice at 24h APF. (Top left) In WT, Sema-2b+ axons (red) choose the ventromedial trajectory, while Sema-2b– axons (black) choose the dorsolateral trajectory. (Top right) In sema-2b, sema-2a/2b, or plexB mutants, axons predominantly select the dorsolateral trajectory, indicating a ‘dorsolateral default’ in the absence of Sema-2b/PlexB signaling. Orange shading in sema-2b antennal lobe depicts remaining Sema-2a. (Bottom left) Pan-ORN Sema-2b overexpression in WT brains causes most ORNs to select the ventromedial trajectory. Similarly, sparse or single MARCM-based Sema-2b overexpression in a sema-2b^−/− background causes labeled axons to select the ventromedial trajectory. Orange shading depicts remaining central Sema-2a. Together with the mutant analyses above and ORN-specific PlexB rescue (Figure 3A), these data suggest that Sema-2b/PlexB signaling suppresses the dorsolateral default. (Bottom right) The stochastic trajectory choice of entire bundles in the absence of central Sema-2a/2b demonstrates the importance of these target zone-derived cues in orienting incoming Sema-2b+ axons. Intact axonal Sema-2b/PlexB signaling consolidates all axons into single bundles.

(E) A speculative model of cellular mechanisms of Sema-2b function. Left: For pioneering axons, Sema-2b expressed in cis may form a complex with PlexB, sensitizing it to the attraction of centrally-derived Sema-2a and Sema-2b in trans (arrow), creating a ventromedial trajectory bias (Figure 7A, [1]). Right, for late-arriving axons, cis Sema-2b may likewise sensitize PlexB to trans Sema-2b from preceding ventromedial (VM) axons. Sema-2b/PlexB thus mediate attractive axon-axon interactions that consolidate initial trajectory choice (Figure 7A, [2]).

Sema-2b Also Acts Non-Cell-Autonomously in ORNs

While Sema-2b can act cell-autonomously to direct ventromedial ORN targeting, Sema-2b produced from ventromedial-targeted ORN axons may also serve as a ligand to attract other PlexB/Sema-2b+ ORN axons (Figure 3 and see Figure 7). To test potential non-cell-autonomous functions of Sema-2b, we performed eyFlp reverse MARCM (Komiyama et al., 2004), in which up to 50% of all ORNs are sema-2b^−/− and only WT ORN axons are labeled (Figure 4G). If Sema-2b acts strictly cell-autonomously, WT ORN axons should target normally in these mosaic animals. However, we found significant targeting defects in multiple ventromedial ORN classes. This is consistent with the idea that Sema-2b from ORNs also acts as a ligand to mediate axon-axon interactions, thereby promoting ventromedial trajectory and targeting. Interestingly, we observed variable penetrance between different classes (Figure 4H–I and Figure S5D). Or49b and Or98a ORNs, targeting to VA5 and VM5v respectively, were largely normal in eyFlp reverse MARCM (Figure 4I) but exhibited the most penetrant phenotypes in hsFlp MARCM (Figure 4C). The simplest interpretation is that axons from Or49b and Or98a ORNs arrive at the antennal lobe earlier than axons of other classes, and therefore rely primarily on cell-autonomous Sema-2b function to detect central cues (see next section).

Sema-2a and Sema-2b from the Central Brain Orient ORN Axon Trajectory Choice

While Sema-2b/PlexB-mediated attractive ORN axon-axon interactions can account for why Sema-2b+ axons select a common trajectory, they cannot explain why Sema-2b+ axons always choose the ventromedial trajectory. A possible scenario would be for an external cue to bias the initial choice and orient it with respect to the antennal lobe, after which Sema-2b/PlexB mediated axon-axon interactions can further consolidate Sema-2b+ axons into a common trajectory.

We have previously shown that Sema-2a and Sema-2b act redundantly to regulate PN dendrite targeting before the arrival of pioneering adult ORN axons. Specifically, Sema-2a and Sema-2b derived from degenerating larval ORNs and developing adult PNs form a ventromedial high–dorsolateral low concentration gradient, thereby directing Sema-1a dependent PN dendrite targeting along this axis (Sweeney et al., 2011). Indeed, both Sema-2a and Sema-2b were still enriched ventromedially within the antennal lobe at 24h APF (Figure 1A₂; Figure S6A) and could thus act as orienting cues for incoming adult axons. To test this idea, we first examined ORN axon trajectory choice in sema-2a^−/−, sema-2b^−/− double mutants. We found similar axon trajectory choice defects (Figure 5A; quantified in Figure 5B) as in sema-2b mutants (Figure 1B, 1H) or plexB mutants (Figure 3A–B). These data indicate that in the absence of Sema-2a/2b/PlexB signaling, most ORN axons take the dorsolateral trajectory by default.

Figure 5. Central Sema-2a and Sema-2b Bias Initial Axon Trajectory Choice.

(A) 24h APF sema-2a^−/−, sema-2b^−/− double mutant brains in which pebbled-GAL4 drives mCD8:GFP in all adult ORNs. Axons predominantly choose dorsolateral trajectories just as in sema-2b or plexB mutants.

(B) Quantification of dorsolateral / ventromedial axon bundle ratio, as in Figure 1H. control = 1.02, n = 6; sema-2a^-/, sema-2b^−/− = 3.91, n = 11; ** p < 0.01; unpaired t-test.

(C) 24h APF sema-2a^−/−, sema-2b^−/− double mutant brains in which pebbled-GAL4 drives mCD8:GFP as well as Sema-2b expression in all adult ORNs, but not in larval ORNs due to the presence of Orco-GAL80 (see Figure S6). Left, all ORN axons labeled by UAS-mCD8:GFP; Middle, Sema-2b staining; Right, overlay of left and middle panels. Top row, representative image in which ORNs choose both ventromedial and dorsolateral trajectories (4/27=14.8% hemispheres); middle row, representative image in which all ORN axons choose the ventromedial trajectory (12/27=44.4% hemispheres); bottom row, representative image of an antennal lobe in which all ORN axons choose the dorsolateral trajectory (11/27=40.8% hemispheres). See Figure S6 for Sema-2a expression patterns and loss-of-function phenotypes.

Next, we used pebbled-GAL4 to express Sema-2b in ORNs of sema-2a^−/−, sema-2b^−/− double mutants, to specifically test the role of central Sema-2a/2b while keeping Sema-2b/PlexB mediated ORN axon-axon interactions intact. To restrict Sema-2b expression to adult ORNs only, we also included Orco-GAL80, which can suppress pebbled-GAL4 in larval ORNs throughout their development (Figure S6F–G). We found that the trajectory of Sema-2b-expressing ORNs fell into three groups under these conditions. In 15% of cases (n=4/27), axons were split between the dorsolateral and ventromedial trajectories (Figure 5C, top). In the remaining 85% cases, however, all axons exclusively chose either the ventromedial (12/27) or dorsolateral (11/27) trajectory, with nearly equal probability (Figure 5C middle and bottom). These results indicate that centrally derived Sema-2a and Sema-2b are essential to bias ORN axon trajectory choice. sema-2a^−/− single mutants did not exhibit significant ORN mistargeting phenotypes, and Sema-2b overexpression was sufficient to bias axons to the ventromedial trajectory even in sema-2a mutants (Figure S6B–E). Thus, central Sema-2a and Sema-2b act similarly and redundantly in orienting incoming ORN axons. In the absence of these central cues, ORN axons were equally likely to choose ventromedial or dorsolateral trajectories. The fact that all Sema-2b+ axons bundle together in 85% of cases supports the idea Sema-2b/PlexB-mediated axon-axon signaling — which remained intact in this experiment—was sufficient to consolidate axons into single bundles, even though trajectory choice became stochastic.

Sema-2b Expression Is Negatively Regulated by Notch Signaling in ORNs

What mechanisms regulate Sema-2b expression and restrict it to ventromedial ORNs? A previous study demonstrated that the Notch pathway diversifies ORN identities and organizes their axonal projections (Endo et al., 2007). Specifically, asymmetric Notch pathway activation divides ORNs within each sensillum into Notch-ON and Notch-OFF classes with distinct axonal trajectories and targeting decisions. In the majority of cases, Notch-ON and Notch-OFF ORNs from the antenna respectively select dorsolateral and ventromedial axon trajectories. These observations led us to hypothesize that 1) the Notch pathway regulates Sema-2b expression, and 2) that Sema-2b acts downstream of Notch to specify axon trajectory choice.

To test these hypotheses, we utilized AM29-GAL4, which labels the two ORN classes that reside within the ab10 sensillum: the Notch-ON ORN takes a dorsolateral trajectory and projects to the DL4 glomerulus, while the Notch-OFF ORN takes a ventromedial trajectory and projects to DM6 (Endo et al., 2007). To test whether the Notch pathway regulates Sema-2b expression, we first used AM29-GAL4-based MARCM labeling to specifically visualize pairs of ORNs within the ab10 sensillum. We found that Sema-2b protein was enriched as perinuclear particles in only one cell of each AM29⁺ ORN pair at 36h APF (Figure 6A₁). These perinuclear particles likely correspond to Sema-2b protein in the secretory pathway, as they were absent from both cells in sema-2b^−/− paired MARCM clones (Figure 6B₁). To examine whether the Notch pathway regulates Sema-2b distribution, we used MARCM to generate AM29⁺ paired clones mutant for mastermind (mam), a core component of the Notch-activated transcriptional complex (Kovall, 2008). Strikingly, Sema-2b was present in both cells of mam^−/− paired clones (Figure 7C₁), indicating that Notch signaling normally represses Sema-2b in one cell of each AM29⁺ pair (Figure 6A₃–6C₃).

Figure 6. Sema-2b is Repressed by the Notch Pathway and Mediates Trajectory Choice Downstream of the Notch Pathway.

(A) A₁, Left, wild-type AM29+ paired clones generated by hsFlp MARCM and labeled with mCD8:GFP. Middle, Sema-2b protein is selectively enriched in one cell of each paired clone (12/12 clones). Arrowheads point to Sema-2b puncta. Right, overlay of left and middle panels. Elav immunostaining labels neuronal nuclei; ORN cell bodies are outlined in yellow. Scale bar = 5 μm. A₂, axonal projections from wild-type AM29+ paired clones. One cell takes a ventromedial trajectory and selects the DM6 glomerulus, while the other takes a dorsolateral trajectory and selects the DL4 glomerulus (16/16 clones). Yellow arrowheads mark axon trajectories. A₃, summary schematic of Sema-2b expression and axonal projection patterns for wild-type paired clones. Green, MARCM-labeled AM29+ paired clones; yellow, Sema-2b enrichment; blue, activated Notch signaling according to Endo et al. (2007).

(B) B₁, immunostaining confirms absence of Sema-2b signal in sema-2b^−/− paired clones; B₂, both axons from sema-2b^−/− mutant clones take dorsolateral trajectories (17/22=77.3% clones), but target correctly to DM6 and DL4 (22/22 clones); B₃, summary schematic.

(C) C₁, Sema-2b is expressed in both cells of mam^−/− paired clones (arrowheads, 15/15 clones); top, anterior section; bottom, posterior section of the same paired clone. C₂, both axons of mam^−/− paired clones take ventromedial trajectories and select the DM6 glomerulus (22/22 clones). C₃, summary schematic; removing mam de-represses Sema-2b expression, and thus causes axons to choose ventromedial trajectories.

(D) D₁, as expected, Sema-2b is absent in mam^−/−, sema-2b^−/− double mutant clones; D₂, both axons from double mutant clones take dorsolateral trajectories (32/37=86.5% clones); top, example in which axons do not innervate the DL4 glomerulus (17/37=46% clones); bottom, example with normal targeting to DM6 and DL4 (20/37=54% clones). D₃, summary schematic; Sema-2b removal reverts the mam^−/− phenotype with respect to trajectory choice, but only partially “rescues” target selection as represented by the dashed DL4 glomerulus.

(E) Quantification of trajectory choice phenotypes. Orange/blue hatched, one dorsolateral trajectory and one ventromedial trajectory; orange, ventromedial trajectory only; blue, dorsolateral trajectory only.

(F) Quantification of target selection phenotypes. Orange/blue hatched, one DL4 terminus and one DM6 terminus; orange, DM6 only; blue, DL4 only. See Figure S7 for epistatic analyses of additional ORN classes.

Sema-2b Mediates Notch Pathway Output for ORN Trajectory Choice

To examine the functional consequences of Notch-mediated Sema-2b repression in ORN axon targeting, we analyzed axon trajectory and glomerular targeting for AM29⁺ paired MARCM clones. In WT paired clones, one ORN took the ventromedial trajectory to DM6, whereas its partner took a dorsolateral trajectory to DL4 (Figure 6A₂, A₃, 16/16 clones). In mam^−/− paired clones, both axons adopted a ventromedial trajectory and targeted to DM6 (Figure 6C₂, C₃, 22/22 clones) in accordance with previous results (Endo et al. 2007). In sema-2b^−/− paired clones, however, both axons projected dorsolaterally to their targets in most cases (Figure 6B₂, B₃; 17/22=77.3% clones), confirming that Sema-2b is essential for ventromedial axon pathway choice. Because sema-2b and mam reside on the same chromosome arm, we used MARCM to test double mutants. Remarkably, in sema-2b^−/−, mam^−/− double mutant paired clones, both neurons chose a dorsolateral axon trajectory, a phenotype opposite to that of mam^−/− paired clones (Figure 6D₂, D₃; 32/37 clones; Figure 6E, compare rightmost two columns). Together with the fact that mam negatively regulates Sema-2b expression, this epistasis experiment demonstrates that sema-2b acts downstream of mam. Furthermore, since additional loss of sema-2b reverted mam^−/− axon trajectory choice phenotypes, these results also indicate that Sema-2b alone accounts for Notch pathway-mediated axon trajectory choice.

Interestingly, Sema-2b removal only partially reverted the mam^−/− phenotype in glomerular targeting, as double mutant axons skipped the DL4 glomerulus in 46% of clones despite taking the correct dorsolateral trajectory that passed by the target DL4 glomerulus (17/37 clones; Figure 6D₂, compare top and bottom; quantified in Figure 6F). Notch signaling thus regulates other factors in addition to Sema-2b to direct glomerular target choice.

To test how generally Notch regulates Sema-2b, we extended our phenotypic and epistasis analyses to the ab1 and at2 sensilla. In each case, we used two Or-GAL4 lines to label a dorsolateral (Notch ON) and a ventromedial (Notch OFF) ORN class each, and used hsFLP MARCM to manipulate the genotype of labeled ORNs. Just as in AM29⁺ paired clones, we found that Sema-2b was required for ventromedial trajectory and was epistatic to mam (Figure S7). Thus, Sema-2b mediates axon trajectory choice downstream of Notch signaling in multiple ORN classes.

Discussion

Our analyses of Sema-2b function have revealed multiple cellular and molecular mechanisms that contribute to ORN targeting specificity. Chronologically, Notch signaling first limits Sema-2b expression to ventromedial ORNs (Figure 7A, right). Secreted Sema-2a and Sema-2b from the central brain then bias the trajectory choice of these Sema-2b+ ORN axons once they reach the antennal lobe (Figure 7A, left), after which Sema-2b/PlexB-mediated ORN axon-axon interactions consolidate this trajectory choice and promote ventromedial bundle formation (Figure 7B). Ultimately, choosing an appropriate developmental trajectory is crucial for proper target selection in ventromedial ORN classes (Figure 7C).

From ORN Fate to Trajectory Choice: the Notch Pathway Specifies Differential Sema-2b Expression

The Notch pathway regulates numerous developmental events in both invertebrates and vertebrates, and its classic roles in specifying cell fate in the peripheral and central nervous systems is well established (Louvi and Artavanis-Tsakonas, 2006). The diversification of cell fates within individual sensilla (Endo et al., 2007) is an excellent example of how Notch determines neuronal fate, with specific consequences for axon targeting. While Notch has been proposed to act through a transcription-independent cytosolic pathway to regulate axon patterning (Kuzina et al., 2011), ORN axon targeting appears to utilize canonical transcriptional regulation, given its dependence on the transcriptional co-activator Mastermind (Endo et al., 2007). Canonical Notch activity has been shown to regulate axon and dendrite development in multiple contexts in vertebrates and invertebrates (e.g.: Sestan et al., 1999; Redmond et al., 2000; Langen et al., 2013; Li et al., 2013).

Here, we identified Sema-2b as a crucial downstream target in Notch-mediated ORN axon targeting (Figure 7A). Sema-2b expression is negatively regulated by Notch signaling, thus establishing a molecular difference between Notch-ON and Notch-OFF ORNs. Moreover, sema-2b and mam mutant ORNs exhibit opposite trajectory choice defects, while sema-2b, mam double mutant ORNs phenocopy sema-2b mutant ORNs. These data cumulatively indicate that Sema-2b acts downstream of the Notch pathway, and that Sema-2b is the primary mediator of Notch pathway effects on ORN trajectory choice.

However, Sema-2b is unlikely to account for all aspects of Notch-mediated ORN fate diversification, as our double mutant analysis indicates that synaptic target choice relies on additional Notch targets independent of Sema-2b (Figure 6D₂-F). The specific mechanism by which Notch signaling represses Sema-2b expression in select ORNs is currently unknown. Notch may directly repress Sema-2b transcription, or indirectly reduce Sema-2b mRNA or protein stability post-transcriptionally. Indeed, our observation that Sema-2b protein levels were significantly lower in dorsolateral ORN (Notch ON) axons compared to ventromedial (Notch OFF) axons despite pan-ORN GAL4-driven expression (Figure 1H inset) favors an indirect post-transcriptional regulation of Sema-2b by the Notch pathway. While these possibilities remain to be investigated in future studies, our identification of an axon guidance molecule downstream of Notch provides an instructive example of how a fate decision is translated into an axon guidance decision.

Sema-2b Acts Across Multiple Steps to Instruct Axon Trajectory Choice

Our genetic analyses have uncovered multiple mechanisms that ensure proper axon trajectory choice when ORN axons first arrive at the antennal lobe (Figure 7D). Below, we first summarize these findings based on our genetic data, and then place these findings in developmental and cell biological context.

In WT (Figure 7D, top left), Sema-2b+ axons (red) choose the ventromedial trajectory and Sema-2b– axons (black) choose the dorsolateral trajectory. However, most axons in sema-2b mutants, plexB mutants, or sema-2a sema-2b double mutants choose the dorsolateral trajectory (Figure 7D, top right). This indicates that the dorsolateral trajectory becomes the default in the absence of Sema-2b/PlexB signaling. Since overexpressing Sema-2b in all ORNs causes most axons to take ventromedial trajectories, while sparse overexpression in isolated ORNs in sema-2b mutants (Figure 7D, bottom left) or PlexB expression in all ORNs in plexB mutants (not shown in Figure 7) can restore the ventromedial trajectory, we deduce that axonal Sema-2b and PlexB signaling vetoes the dorsolateral default pathway. Finally, since removing central Sema-2a/2b while maintaining Sema-2b/PlexB signaling in ORNs results in stochastic trajectory choice of the entire axon bundle (Figure 7D, bottom right), we infer that axonal Sema-2b/PlexB is sufficient to cause all Sema-2b+ axons to take the same trajectory, while central Sema-2a/2b is normally required to bias the trajectory choice of Sema-2b+ axons ventromedially.

Axon-axon interactions are widely used to establish wiring specificity in complex circuits of vertebrates and invertebrates (Sanes and Yamagata, 2009). In the fly olfactory system, for example, Sema-1a produced by early-arriving antennal ORN axons acts as a repulsive cue to constrain glomerular target selection of late-arriving maxillary palp ORN axons (Sweeney et al., 2007). Here, we show that Sema-2b/PlexB mediated ORN axon-axon interactions regulate the trajectory choice of individual ORN axons when they first arrive at the antennal lobe, well before their final glomerular target selection. This mechanism is thus reminiscent of pre-target axon sorting as described for mammalian ORN axon targeting along the anterior-posterior axis of the olfactory bulb (Imai et al., 2009).

However, axon-axon interactions alone are insufficient to produce highly stereotyped neural maps and likely require external cues for proper orientation. We have identified central Sema-2a and Sema-2b as important orienting cues. In their absence, most Sema-2b+ ORNs can still form bundles, presumably through Sema-2b/PlexB mediated axon-axon interactions, but their trajectory choice becomes stochastic (Figure 5). This experiment thus illustrates how target-derived and axon-derived cues cooperate to specify ORN trajectory choice: central Sema-2a/2b bias Sema-2b+ axons towards ventromedial trajectories (Figure 7A, left); ORN axon-derived Sema-2b subsequently acts as a cue to attract more Sema-2b+ axons to the same trajectory through the PlexB receptor (Figure 7B). Indeed, our mosaic analyses support both cell-autonomous and non-autonomous roles for Sema-2b in ORN axon targeting (Figure 4).

In summary, Sema-2b mediates multiple complementary processes during ORN and PN development. 1) Between 0–18h APF, larval ORN- and PN-derived Sema-2b, along with Sema-2a, form a ventromedial > dorsolateral concentration gradient in the developing antennal lobe to instruct PN dendrite targeting along this axis (Sweeney et al., 2011). 2) The Sema-2a/2b gradient persists through 24h APF to bias the trajectory choice of pioneering Sema-2b+ ventromedial ORN axons. 3) Sema-2b on ORN axons acts as a ligand to mediate bundle formation and consolidate trajectory choice. 4) Sema-2b in ORN axons also acts cell-autonomously in steps 2–3 to guide Sema-2b+ axons to take the ventromedial trajectory (Figure 7A–B). Thus, Sema-2b acts four times in consecutive stages of olfactory circuit wiring, and serves as a molecular link between PN dendrite targeting and ORN axon targeting well before final synaptic matching.

A Possible Mechanism of Sema-2b Function: Cell-Autonomously Sensitizing PlexB to External Semaphorins

While secreted semaphorins are classic extracellular ligands for axon guidance (Kolodkin and Tessier-Lavigne, 2011), our mosaic genetic analyses demonstrate surprisingly that Sema-2b can act cell-autonomously to instruct ORN axon trajectory choice. In this context, Notch regulation of Sema-2b in individual ORNs ensures differential expression of this instructive molecule in different ORN classes.

How can a secreted molecule act cell-autonomously to instruct a guidance choice? Given the nearly identical phenotypes of plexB and sema-2b mutants (this study) and high-affinity binding of Sema-2b to PlexB-expressing neurons (Wu et al., 2011), we propose that Sema-2b cell-autonomously primes or sensitizes PlexB, such that it acts as an attractive receptor for external Sema-2a and Sema-2b. In one potential scenario, Sema-2b and PlexB expressed in cis (within the same cell) may form a complex that signals attraction in response to external Sema-2a/2b in trans (Figure 7E). This model explains both how centrally-derived Sema-2a/2b orients pioneering Sema-2b+ ORN axons (Figure 7A, [1]; Figure 7E, left) and how “follower” axons utilize ORN-derived Sema-2b for axon-axon interactions and bundle formation (Figure 7A, [2]; Figure 7E, right). This proposal most parsimoniously explains our genetic and mosaic analyses and is fully consistent with Sema-2b+ axon tract formation in the embryonic ventral nerve cord (Wu et al., 2011). However, it awaits future biochemical studies for further validation.

Cis-interactions between an axon guidance molecule and a receptor have previously been documented for vertebrate Ephrin/Eph receptor signaling (Hornberger et al., 1999; Marquardt et al., 2005; Carvalho et al., 2006; Kao and Kania, 2011) and for a vertebrate class-6 transmembrane semaphorin (Haklai-Topper et al., 2010). While these studies suggest cis-attenuation of receptor activity by a membrane-tethered ligand (Yaron and Sprinzak, 2012), our data support the notion that secreted ligands can participate in cis-interactions to promote or sensitize guidance receptor signaling, thus expanding the function of ligand-receptor cis-interactions in axon guidance.

Trajectory Choice Constrains Glomerular Target Choice

Neural circuit wiring is a complex task: in the Drosophila olfactory circuit, each ORN must choose 1 out of 50 alternative postsynaptic targets with high precision after it arrives at the antennal lobe. One strategy to reduce this complexity is to create hierarchical steps, such that axons face fewer simultaneous choices per step. Our genetic analysis of Sema-2b function in ORN axon targeting demonstrates this principle. ORN axons must make a binary decision soon after arriving at the antennal lobe: individual axons take either the ventromedial or dorsolateral trajectory, based on whether they express Sema-2b or not. Proper trajectory is crucial to place ventromedial ORNs in the correct target region for subsequent partner matching and synapse formation. Ventromedial ORNs that select incorrect trajectories early on cannot recover and ultimately make ectopic target choices (Figure 2; Figure 7C).

Interestingly, mistargeted axons in either Sema-2b loss- or gain-of-function conditions retain specificity in target choice (Figure S3), indicating that at least a subset of targeting cues remains functional in both cases. This observation also suggests that cell surface molecules that mediate synaptic partner matching may be distributed throughout the antennal lobe for reiterative use, instead of being restricted to select regions. Indeed, Teneurin-m instructs synaptic partner matching in dorsolateral glomeruli, but is also expressed throughout the antennal lobe in a “salt and pepper” fashion (Hong et al., 2012); ORNs with aberrant trajectories may utilize cues such as Teneurin-m to select specific partners, albeit in incorrect target regions. Early developmental trajectory thus plays an important role in disambiguating cell-surface cues presented by different parts of the antennal lobe. Ultimately, our analysis of Sema-2b illustrates how a relatively small number of molecules can establish complex neural architecture with the help of sophisticated spatial and temporal regulation.

Experimental Procedures

Transgenic Lines

Or-GAL4 or Or-mCD8:GFP lines have been previously described (Couto et al., 2005; Fishilevich and Vosshall, 2005; Komiyama et al., 2004), as have pebbled-GAL4 (Sweeney et al., 2007), UAS-mCD8:GFP (Lee and Luo, 1999), and 10XUAS-IVS-mtdT (Pfeiffer et al., 2010). See Supplemental Information for details on mutant alleles and additional transgenic lines.

Mosaic Analyses

hsFlp MARCM analyses were performed as previously described (Lee and Luo, 1999; Komiyama et al., 2004) with slight modifications. To analyze axon projections in adult brains using hsFlp MARCM, animals between 0–24h APF were heat shocked for 10 minutes at 37°C. For pebbled-GAL4 hsFlp MARCM, newly formed pupae (0h APF animals) were heat-shocked for 5–10 minutes at 37°C and dissected at 24h APF.

Immunofluorescence

Tissue dissection and immunostaining were performed according to previously described methods (Sweeney et al., 2007). See Supplemental Information for antibody information. Confocal images were collected with a Zeiss LSM 510 and processed with Zeiss LSM software, ImageJ, and Adobe Photoshop.

See Supplemental Experimental Procedures for details on: mutant alleles and transgenic lines, immunofluorescence and antibodies, quantification procedures, and genotypes for experiments described in main figures.