Interplay between axonal Wnt5-Vang and dendritic Wnt5-Drl/Ryk signaling controls glomerular patterning in the Drosophila antennal lobe.

Despite the importance of dendritic targeting in neural circuit assembly, the mechanisms by which it is controlled still remain incompletely understood. We previously showed that in the developing Drosophila antennal lobe, the Wnt5 protein forms a gradient that directs the ~45˚ rotation of a cluster of projection neuron (PN) dendrites, including the adjacent DA1 and VA1d dendrites. We report here that the Van Gogh (Vang) transmembrane planar cell polarity (PCP) protein is required for the rotation of the DA1/VA1d dendritic pair. Cell type-specific rescue and mosaic analyses showed that Vang functions in the olfactory receptor neurons (ORNs), suggesting a codependence of ORN axonal and PN dendritic targeting. Loss of Vang suppressed the repulsion of the VA1d dendrites by Wnt5, indicating that Wnt5 signals through Vang to direct the rotation of the DA1 and VA1d glomeruli. We observed that the Derailed (Drl)/Ryk atypical receptor tyrosine kinase is also required for the rotation of the DA1/VA1d dendritic pair. Antibody staining showed that Drl/Ryk is much more highly expressed by the DA1 dendrites than the adjacent VA1d dendrites. Mosaic and epistatic analyses showed that Drl/Ryk specifically functions in the DA1 dendrites in which it antagonizes the Wnt5-Vang repulsion and mediates the migration of the DA1 glomerulus towards Wnt5. Thus, the nascent DA1 and VA1d glomeruli appear to exhibit Drl/Ryk-dependent biphasic responses to Wnt5. Our work shows that the final patterning of the fly olfactory map is the result of an interplay between ORN axons and PN dendrites, wherein converging pre- and postsynaptic processes contribute key Wnt5 signaling components, allowing Wnt5 to orient the rotation of nascent synapses through a PCP mechanism.


Introduction
The prevailing view of neural circuit assembly is that axons and dendrites are separately guided by molecular gradients to their respective positions whereupon they form synapses with each other [1][2][3][4]. However, careful observation of developing neural circuits reveals that the process may be more complex. For example, in the developing retina outer plexiform layer (OPL) the axon terminals of rods and cones, and dendrites of their respective postsynaptic cells, the rod and cone bipolar cells, are initially intermingled in the nascent OPL [5]. Even as the rod and cone axons are connecting with their target dendrites, the terminals are segregating into rodand cone-specific sub-laminae, suggesting that the processes of targeting and synaptic partner matching may be coordinated. Whether the two processes are functionally linked and what mechanisms might be involved are unknown.
The stereotyped neural circuit of the Drosophila olfactory map offers a unique opportunity to unravel the mechanisms of neural circuit development. Dendrites of 50 classes of uniglomerular projection neurons (PNs) form synapses with the axons of 50 classes of olfactory receptor neurons (ORNs) in the antennal lobe (AL) in unique glomeruli [6,7]. This precise glomerular map is thought to be established during the pupal stage by the targeting of PN dendrites [8][9][10]. We previously reported that during the establishment of the fly olfactory map, two adjacent dendritic arbors located at the dorsolateral region of the AL, the DA1 and VA1d dendrites (hereafter referred to as the DA1/VA1d dendritic pair), undergo rotational migration of~45˚around each other to attain their final adult positions [11]. This rearrangement (in the lateral/90˚! dorsal/0˚! medial/270˚! ventral/180˚direction) occurs between 16 and 30 hour After Puparium Formation (hAPF), a period of major ORN axon ingrowth to the AL [8,12,13]. We showed that a Wnt5 signal guides this rotation by repelling the dendrites [11]. Wnt5 is expressed by a set of AL-extrinsic cells and forms a dorsolateral-high to ventromedial-low (DL>VM) gradient in the AL neuropil which provides a directional cue to align the dendritic pattern relative to the axes of the brain. We also showed that the Derailed (Drl)/ Ryk kinase-dead receptor tyrosine kinase, a Wnt5 receptor [14][15][16][17], is differentially expressed by the PN dendrites, thus providing cell-intrinsic information for their targeting in the Wnt5 gradient. Interestingly, drl opposes Wnt5 repulsive signaling so that dendrites expressing high levels of drl terminate in regions of high Wnt5 concentration and vice versa. To further unravel the mechanisms of PN dendritic targeting, we have screened for more mutations that disrupt the rotation of the DA1/VA1d dendritic pair.
Here we report that mutations in the Van Gogh (Vang) gene disrupted the rotation of the DA1/VA1d dendritic pair, thus mimicking the Wnt5 mutant phenotype. Vang encodes a fourpass transmembrane protein [18,19] of the core Planar Cell Polarity (PCP) group, an evolutionarily conserved signaling module that imparts polarity to cells [20,21]. The loss of Vang suppressed the repulsion of the VA1d dendrites by Wnt5, indicating that Vang is a downstream component of Wnt5 signaling. Surprisingly, Vang acts in the ORNs, which suggests an obligatory codependence of ORN axon and PN dendritic migration. We also show that the drl gene is selectively expressed in the DA1 dendrites where it antagonizes Vang and appears to convert Wnt5 repulsion of the DA1 glomerulus into attraction. The opposing responses of the DA1 and VA1d glomeruli likely create the forces by which Wnt5 directs the rotation of the glomerular pair. Our work shows that converging pre-and postsynaptic processes contribute key signaling components of the Wnt5 pathway, allowing the processes to be co-guided by the Wnt5 signal.

Vang promotes the rotation of DA1/VA1d dendrites
We have shown that during wild-type development the adjacent DA1 and VA1d dendrites rotate around each other, such that DA1 moves from its original position lateral to VA1d at 18 hAPF to its final position dorsolateral to VA1d in the adult, an~45˚rotation [11]. We also showed that this rotation requires the Wnt5 gene, for in the null Wnt5 400 mutant the rotation is abolished, resulting in an adult DA1/VA1d angle of 76.03˚± 3.6˚(N = 29, vs 29.32˚± 2.5˚, N = 22 in the Wnt5 400 /+ heterozygous control, Student's t test p<0.0001) (See Materials and Methods and S1 Fig for quantification) (Fig 1A-1C). The DA1 and VA1d pair of dendrites were visualized by expressing UAS-mCD8::GFP under the control of Mz19-Gal4, which specifically labels the DA1, VA1d and DC3 dendrites [22,23]. To elucidate the molecular mechanisms by which Wnt5 controls the rotation of the PN dendrites, we screened a panel of signal transduction mutants for similar defects in DA1/VA1d rotation. We found that animal homozygous for Vang mutations exhibit a DA1/VA1d phenotype that mimicked that of the Wnt5 400 mutant. For example, in the null Vang 6 allele, the DA1/VA1d angle was 54.72˚± 2.8˚(N = 24, vs 27.78˚± 4.6˚, N = 18 in the Vang 6 /+ heterozygous control, t-test p<0.0001) (Fig 1D-1F) suggesting that Vang might function in the Wnt5 pathway. Since the Vang 6 allele, which encodes a truncated 128 amino acid product [19], displayed a highly penetrant phenotype, we examined this allele further. We therefore examined the positioning of glomeruli in different regions of the Vang 6 AL by expressing UAS-mCD8::GFP under the control of various Or-Gal4 drivers [6] (Fig 1G-1O, see Materials and Methods for quantification). We observed that glomeruli in the lateral region of the AL, such as the VA1v glomerulus (Fig 1G-1I), showed the greatest displacement compared with glomeruli in other regions, suggesting that Vang primarily controls neurite targeting in the lateral AL. Since the Wnt5 protein is highly concentrated at the dorsolateral region of the AL [11], the Vang mutant defects are consistent with Vang playing a role in Wnt5 signaling. We hypothesized that Vang mediates Wnt5 signaling in the control of the DA1/VA1d dendritic rotation.
To obtain further evidence for Vang's role in regulating the rotation of the DA1/VA1d dendritic pair, we stained ALs during a time of active glomerular rotation (24 hAPF) [11] with an antibody directed against the N-terminal 143 amino acids of Vang [24]. We observed that the Vang staining has a punctate appearance and is highly concentrated in the dorsolateral region of the AL between 0-9 μm from the AL anterior surface (Fig 2A). Co-labeling of the DA1/ VA1d dendrites with the Mz19-Gal4 driver showed that they reside between~3-6 μm in this high Vang expression domain. Vang staining in the neuropil begins to decline at 10 μm but strongly highlighted the nerve fiber layer (arrow) and the antennal nerve at 8-12 μm (arrow and arrowheads in Fig 2A and 2B), as well as the antennal commissure at 22 μm depth. The antibody stained the Vang 6 mutant ALs (Fig 2C), likely because the Vang 6 allele encodes a truncated protein. Nonetheless, the strong reduction in staining intensity compared with wildtype ALs attested to the antibody's specificity. We concluded that Vang is expressed in the AL during the period of active AL neuropil rotation, where it colocalized with the DA1 and VA1d dendrites. The Vang expression pattern is consistent with the hypothesis that Vang mediates Wnt5 control of the DA1/VA1d dendritic rotation.
To confirm the above findings, we used mosaic techniques to induce ORNs or PNs lacking the Vang gene and examined the effects on the DA1/VA1d dendritic rotation. Induction of either Vang f04290 or Vang 6 mutant ORN axons using the ey-FLP/FRT technique, which induces large clones in the antenna [27], resulted in the DA1/VA1d dendritic pair exhibiting larger angles (54.73˚± 3.26˚, N = 30 and 51.03˚± 2.62˚, N = 28 respectively) compared with animals innervated by wild-type ORN axons (20.24˚± 2.51˚, N = 20, wild type vs. Vang f04290 , p<0.0001; wild type vs. Vang 6 , p<0.0001; Vang 6 vs Vang f04290 , p = 0.7297; one-way ANOVA with post hoc Tukey test) (Fig 4A-4C, 4G and 4H), confirming that Vang is required in the ORNs for DA1/VA1d dendritic rotation. Next, we induced Vang 6 mutant PN clones using the MARCM system [28] with GH146-Gal4 as the PN marker. We observed that Vang 6 mutant PN neuroblast and single-cell clones extended their dendrites into AL and innervated the glomeruli normally (Fig 4D-4F). Importantly, ALs innervated by large vang 6 PN clones exhibited normal dendritic pattern, as judged by the angles of the DA1/VA1d dendrites (27.50˚± 4.89˚, N = 12) compared with those of the control (23.29˚± 5.27˚, N = 7, t-test p = 0.56) (Fig 4E, 4G and 4H). Thus, our transgenic rescue and mosaic experiments showed that Vang functions in the ORNs to non-autonomously promote the rotation of the DA1/VA1d dendritic pair.

Vang is not required for ORN axon growth or correct ORN-PN pairing
A possible explanation for the Vang mutant phenotype is that Vang is required for ORN axon growth to the AL, the failure of which indirectly disrupted glomerular patterning. Mutations in Vang have been shown to result in abnormal projection of mushroom body axons [29]. To determine if ORN axons entered the AL in the Vang 6 mutant, we labeled eight different ORN axon terminals in the AL using Or-Gal4 drivers. We found that Vang 6 mutant axons entered the AL normally, although their terminals were shifted in the AL neuropil (Fig 1). To investigate if the Vang mutation disrupted the proper matching of the ORN axons and PN dendrites, we simultaneously labeled pre-and postsynaptic partners of glomeruli for which specific markers were available. We achieved this by labeling the DA1, VA1d and DM1 dendrites with Mz19-Gal4 driving UAS-mCD8::GFP and simultaneously the ORN axons targeting the VA1d and VA1v glomeruli with the Or88a-CD2 and Or47b-CD2 transgenes respectively. We observed that Or88a axons were strictly paired with VA1d PN dendrites in the Vang 6 mutant as in the wild type (Fig 5A and 5B). Likewise, the Or47b axons strictly innervated the VA1v glomerulus, and never strayed into the VA1d territory in the Vang 6 mutant ( Fig 5C). Thus, Vang is not required for ORN axon projection into the AL or their correct pairing with their postsynaptic partners. We propose that Vang functions in the context of paired axons and dendrites allowing the neurites to coordinately respond to the Wnt5 signal. This idea is consistent with our observation that PN dendritic rotation occurs between 16 and 30 hAPF [11], the period of major ORN axon invasion into the AL [8,12].

Vang acts downstream of Wnt5 to repel the VA1d glomerulus
The close resemblance of the Vang and Wnt5 mutant phenotypes raised the questions of whether and how Vang might function in the Wnt5 signaling pathway to regulate the rotation of the DA1 and VA1d glomeruli. To address these questions, we asked if loss of Vang would block Wnt5 signaling. We previously showed that overexpression of Wnt5 in the DA1 and VA1d dendrites with the Mz19-Gal4 driver split the VA1d dendrites into two smaller arbors probably due to repulsion between the dendrites (Fig 6A and 6B) [11]. Interestingly, Wnt5 overexpression had no effect on the DA1 dendrites, indicating that the DA1 and VA1d dendrites respond differentially to the Wnt5 signal. The VA1d defect provided an opportunity to assess if Vang is needed for the Wnt5 gain-of-function phenotype. Whereas only 9.37% (3/32) of the VA1d dendrites in the Mz19>Wnt5 animals were intact, this fraction rose to 45.65% (21/46) in the Mz19>Wnt5; Vang 6 /Vang 6 animals (Fig 6C and 6F). Moreover, the distances between the split VA1d arbors in the Mz19>Wnt5; Vang 6 /Vang 6 animals were smaller than those in the Mz19>Wnt5 animals (11.83 μm ± 0.61 μm, N = 25, vs 21.06 μm ± 0.96 μm, N = 29, t-test p<0.0001) (Fig 6B, 6C and 6G). Despite severe distortion, the VA1d dendrites were faithfully paired with their Or88a axon partners, reinforcing the idea that Wnt5 signaling does not play a role in ORN-PN matching (Fig 6D and 6E). Overexpression of Wnt5 in the DA1 and VA1d dendrites did not affect the development of the Or88a neurons (S2 Fig). We conclude that Wnt5 signals through Vang to repel the VA1d glomerulus.
To further probe the relationship between Wnt5 and Vang, we examined the DA1/VA1d rotation in animals lacking both genes. We observed that the rotation in the Wnt5 400 ; Vang 6 double mutant (92.20˚± 4.1˚, N = 41) is more severely disrupted than that in either single mutant (76.03˚± 3.62˚in Wnt5 400 , n = 29, t-test p = 0.0031 and 54.12˚± 2.8˚in Vang 6 , n = 34, t-test p<0.0001) (Fig 7). The enhanced phenotype of the Wnt5 400 ; Vang 6 mutant suggested that Wnt5 and Vang could function independently to promote DA1/VA1d rotation (Fig 7F). We currently do not know how Vang acts independently of Wnt5. However, it interesting that Wnt5 directs the rotation of the DA1/VA1d glomeruli through both Vang-dependent and Vang-independent pathways. Since Wnt5 acts through Vang in the VA1d glomerulus, we hypothesized that Wnt5 acts through a Vang-independent mechanism in the DA1 glomerulus.

drl promotes the rotation of DA1-VA1d glomeruli
The Drl atypical receptor tyrosine kinase has been shown to bind Wnt5 and mediates its signaling in the migration of a number of cell types [14][15][16][17]. We previously demonstrated that

PLOS GENETICS
Wnt5, Drl and Vang signaling directs glomerular rotation Drl is differentially expressed by PN dendrites wherein it antagonizes Wnt5's repulsion of the dendrites [11]. To delineate the AL region where drl functions, we examined the positioning of several glomeruli in the null drl 2 mutant by expressing UAS-mCD8::GFP under the control of various Or-Gal4 drivers (Fig 8A-8I). We observed that, as in the Vang mutant, the lateral glomeruli showed the strongest displacements in positions compared with the control indicating that drl primarily regulates neurite targeting in the lateral AL (Fig 8A-8C). To better characterize the neuropil defect in this region we employed the Crispr/Cas9 technique [30] to create the null drl JS allele on the Mz19-Gal4 chromosome (Materials and Methods), which allowed us to assess the DA1/VA1d dendritic arrangement in the drl mutant. To our surprise, we observed that the DA1/VA1d dendritic pair in the drl JS /drl 2 null mutant showed strong deficits in DA1/VA1d rotation, resembling the Wnt5 400 null phenotype (Fig 8J-8L). Indeed, measurement of the DA1/VA1d angle of the drl JS /drl 2 mutant (76.55˚± 3.75˚, N = 31) showed that it was even slightly larger than that of the Wnt5 400 mutant (69.22˚± 5.03˚, N = 32, t-test p = 0.2493). The similarity of the drl and Wnt5 mutant phenotypes indicates that drl cooperates with Wnt5 in promoting the rotation of the DA1/VA1d glomeruli.

PLOS GENETICS
Wnt5, Drl and Vang signaling directs glomerular rotation drl acts in the DA1 dendrites to promote glomerular attraction to Wnt5 How does drl mediate Wnt5 function in glomerular rotation? We hypothesized that drl functions in the DA1 dendrites, to regulate migration of the DA1 glomerulus towards the Wnt5 source. In support of this idea, antibody staining of the Drl protein showed that it is highly expressed by the DA1 dendrites but not the VA1d dendrites (Fig 2D and 2E) [11]. The domain of high Drl expression occupies the anterior dorsolateral domain of the 24 hAPF ALs (0-8 μm from anterior), a region in which Wnt5 is also highly expressed [11]. The hypothesis predicts that ablation of drl in DA1 alone would disrupt the rotation of the DA1/VA1d glomeruli. To test this hypothesis, we used MARCM to induce drl JS homozygosity in either DA1 or VA1d dendrites and assessed the effects on DA1/VA1d glomerular rotation. The DA1 and VA1 dendrites could be independently identified since their cell bodies are located in the lateral and anterodorsal PN clusters respectively. We observed that drl JS mutant VA1d dendrites were associated with glomerular pairs with small angles (17.13˚± 3.68˚, N = 16), that is, with the DA1 glomerulus closely associated with the dorsolateral AL (Fig 9C-9F). In contrast, drl JS mutant DA1 dendrites are associated with glomeruli with wide variations in angles (58.43˚± 8.33˚, N = 14, t-test p = 0.0003), that is, with the DA1 glomerulus not closely associated with the dorsolateral AL (Fig 9A, 9B, 9E and 9F). Thus, drl appears to act in the DA1 dendrites to confer directionality of the DA1 glomerulus towards the Wnt5 source.

PLOS GENETICS
Wnt5, Drl and Vang signaling directs glomerular rotation drl likely converts Wnt5 repulsion of the DA1 glomerulus into attraction How does drl promote the migration of the DA1 glomerulus towards Wnt5? We propose two models by which drl could accomplish this task. In the first model, drl acts as a positive effector of Wnt5 attractive signaling. In the second model, drl neutralizes Wnt5 repulsive signaling and/or converts the repulsive signaling into an attractive one. Careful examination of the DA1/VA1d targeting defects in drl JS /drl 2 null mutant revealed differences with that of the Wnt5 400 mutant, inconsistent with the idea that drl acts as a positive effector of Wnt5 signaling. First, the DA1 glomerulus is often displaced medially from the AL lateral border (6.26 μm ± 1.85 μm from border, N = 28) (Fig 10H), a defect not seen in the Wnt5 400 mutant. This resulted in the frequent reversal in the positions of the DA1 and VA1d glomeruli (Fig 10B), or displacement of both glomeruli medially from the AL lateral border ( Fig  10A). Second, the mean DA1/VA1d angle in the drl JS /drl 2 null mutant is slightly but not significantly larger than that of Wnt5 400 null mutant (Fig 10E and 10G). Instead the defects are more consistent with the second model, which predicts increased Wnt5 repulsion of the DA1 glomerulus in the drl mutant, thus driving the glomerulus ventromedially. To test this hypothesis, we simultaneously removed both drl and Wnt5 functions and examined the displacement of the DA1/VA1d glomeruli. We observed that the DA1 glomerulus is restored to AL lateral border in the Wnt5 400 ; drl JS /drl 2 double mutant, as it is in the Wnt5 homozygote (0.00 μm from border, N = 30 and N = 25 respectively) (Fig 10C and 10H). We also observed that the DA1/VA1d angle in the Wnt5; drl double mutant (70.66˚± 4.39˚, N = 29) is more similar to that of the Wnt5 mutant (69.22˚± 5.03˚, N = 32, t-test p = 0.83) than that of the drl mutant (76.55˚± 3.75˚, N = 31, t-test p = 0.3102) (Fig 10E and 10G). We conclude that Wnt5 repels the DA1 glomerulus ventromedially and that drl antagonizes the Wnt5 repulsive activity. Since we showed above that drl promotes the migration of DA1 towards Wnt5, we conclude that drl acts in the DA1 glomerulus to convert Wnt5 repulsion of the DA1 glomerulus into attraction. Taken together, our results suggest that Wnt5 orients the rotation of the VA1d/DA1 glomeruli by attracting the DA1 glomerulus through Wnt5-drl signaling and repelling the VA1d glomerulus through Wnt5-Vang signaling.
To investigate the mechanism by which drl converts Wnt5 repulsion of the DA1 glomerulus into attraction, we examined the mechanism by which Wnt5 repels the DA1 glomerulus. A likely scenario is that Wnt5 repels the DA1 glomerulus through Vang. To test this idea, we simultaneously removed both drl and Vang functions and measured the displacement of the

PLOS GENETICS
Wnt5, Drl and Vang signaling directs glomerular rotation DA1 glomerulus from the AL lateral border as well as the DA1/VA1d rotational angle. We found that in the Vang 6 drl JS /Vang 6 drl 2 double mutant, the DA1 glomerulus is restored to the AL lateral border (0.156 μm ± 0.132 μm, N = 28, t-test p = 0.0017) (Fig 10D and 10H). We conclude that drl neutralizes the Wnt5-Vang repulsion of the DA1 glomerulus. Interestingly, measurement of the DA1/VA1d angles in the Vang drl double mutant showed that the rotation of the glomeruli is more severely impaired (93.41˚± 3.65˚N = 29) than that of either single mutants (76.55˚± 3.75˚in drl 2 , N = 31, t-test p = 0.0021 and 54.57˚± 2.75˚in Vang 6 , N = 35, t-test p<0.0001) (Fig 10F and 10G). The enhanced phenotype of the Vang drl double mutant indicated that drl and Vang act in parallel pathways to promote DA1/VA1d rotation. The

PLOS GENETICS
Wnt5, Drl and Vang signaling directs glomerular rotation parallel functions are in accord with our model that drl acts in the DA1 glomerulus while Vang acts in the adjacent VA1d glomerulus to promote DA1/VA1d rotation. Simultaneous loss of Vang and drl would be expected to exacerbate the DA1/VA1d rotational defect.

Discussion
Elucidating the mechanisms that shape dendritic arbors is key to understanding the principles of nervous system assembly. Genetic approaches have revealed both intrinsic and extrinsic cues that regulate the patterning of dendritic arbors [31,32]. In contrast, there are only a few reports on the roles of axons in shaping dendritic arborization [33,34]. In this paper we provide evidence that final patterning of the fly olfactory map is the result of an interplay between ORN axons, PN dendrites and the Wnt5 directional signal (Fig 11). We show that the Vang PCP protein [20,21] is an axon-derived factor that mediates the Wnt5 repulsion of the VA1d dendrites. We also show that the Drl protein is specifically expressed by the DA1 dendrites where it antagonizes the Wnt5-Vang repulsion of the DA1 glomerulus and likely converts it into an attractive response. The differential responses of the DA1 and VA1d glomeruli to Wnt5 would produce the forces by which Wnt5 effects the rotation of the glomerular pair. We present the following lines of evidence in support of this model of olfactory neural circuit development.
Immunostaining showed that Vang is expressed at the same time and place as Wnt5 and Drl, and concentrated in the dorsolateral AL where major dendritic reorganization occurs [11]. Mutations in Vang strongly disrupted the pattern of glomeruli in the AL, mimicking the Wnt5 mutant phenotype. Notably, mutation of Vang suppressed the strong repulsion of the VA1d dendritic arbor caused by Wnt5 overexpression, indicating that Vang acts downstream of Wnt5 to repel the VA1d dendrites. Unexpectedly, using cell type-specific transgenic experiments and mosaic analyses we found that Vang functions specifically in the ORNs, indicating an obligatory codependence of ORN axon and PN dendritic targeting. Unlike the VA1d glomerulus, which expresses low levels of drl and is repelled by Wnt5, the adjacent DA1

PLOS GENETICS
Wnt5, Drl and Vang signaling directs glomerular rotation glomerulus expresses high levels of drl. Mosaic analyses showed that drl acts specifically in the DA1 dendrites to confer directionality of the DA1 glomerulus towards Wnt5. Finally, in the absence of drl, the DA1 glomerulus is displaced away from the Wnt5 source, a defect that is suppressed by the removal of either Wnt5 or Vang. Taken together, we propose that drl likely converts Wnt5 repulsion of the DA1 glomerulus into attraction by inhibiting Wnt5-Vang repulsive signaling.
We envision that Vang and Drl act cell autonomously to regulate axonal and dendritic guidance respectively and cell non-autonomously to modulate each other's functions. Both Vang and Drl/Ryk have well-documented cell autonomous functions in neurite guidance. For example, vertebrate Vangl2 was localized to the filopodia of growth cones [35,36] and Drosophila Vang mediates the repulsion of mushroom body axon branches in respond to Wnt5 [29,37]. Similarly, Drl and Ryk mediate the functions of Wnt5 and Wnt5a respectively in the targeting of dendrites [16,38] and axons [5,14,17,39,40]. Both proteins also have well-documented cell non-autonomous functions. Vertebrate Vangl2 acts as a ligand to steer migrating neurons [41,42]. We and others have shown that Drl could sequester Wnt5 using its extracellular Wnt Inhibitory Factor (WIF) motif [38,[43][44][45]. Indeed, in this manner Drl may reduce Wnt5-Vang interaction, thus neutralizing Wnt5 repulsion of the glomeruli. How would Drl convert a glomerulus's response to Wnt5 from repulsion to attraction? Increasing Ryk levels were proposed to titrate out Fz5 in chick retinal ganglion axons, thus converting growth cone response to Wnt3 from attraction to repulsion [46]. Whether Drl function through a similar mechanism in the DA1 dendrites will require further investigation.
The opposing functions of Vang and Drl in a migrating glomerulus satisfies Geirer's postulate for topographic mapping, which states that targeting neurites must detect two opposing forces in the target so that each neurite would come to rest at the point where the opposing forces cancel out [47]. Thus, the relative levels of Drl and Vang activities in a glomerulus may determine its targeting position in the Wnt5 gradient (Fig 11). For the DA1/VA1d glomerular pair, the relative levels of Drl and Vang activities would result in the migration of DA1 up the gradient and VA1d down the gradient, that is, the rotation of the glomerular pair. The opposing effects of Wnt5 on the targeting glomeruli could allow the single Wnt5 gradient to refine the pattern of the olfactory map.
The rotation of the DA1/VA1d glomeruli bears intriguing resemblance to the PCP-directed rotation of multicellular structures such as mouse hair follicles and fly ommatidia, whose mechanisms remain incompletely understood [48][49][50]. Our demonstration of the push-pull effect of Wnt5 on the glomeruli suggests that similar mechanisms may be involved in other PCP-directed rotations. Planar polarity signaling has emerged as an important mechanism in the morphogenesis of many tissues [20,51,52]. However, apart from the molecules of the core PCP group (Vang, Prickle, Frizzled and Dishevelled) the identities of other signaling components are subjects of debate. A key question is the extracellular cue that aligns the core PCP proteins with the global tissue axes. Although Wnt ligands have been implicated, a definitive link between them and PCP signaling has been difficult to establish [20,21,53]. Our work showing that Wnt5 and Vang act together to direct the orientation of nascent glomeruli adds to two other reports [54,55] that Wnt proteins play instructive roles in PCP signaling. Another debate surrounds the role of Drl/Ryk role in PCP signaling. First identified as signal transducing receptors for a subset of Wnt ligands in Drosophila [17,56], vertebrate Ryk was subsequently shown to act in PCP signaling [57,58]. These reports, combined with the lack of classical wing-hair PCP phenotypes in the drl mutant, led to the proposal that Ryk's PCP function is a vertebrate innovation [59]. Our demonstration of drl's role in Wnt5-Vang signaling suggests, however, that the Drl/Ryk's PCP function is likely to be evolutionarily ancient.

Fly strains and Crispr/Cas9 knockout of the drl gene
All mutant and transgenic fly lines were obtained from the Bloomington Drosophila Stock Center except for UAS-Vang, which was a gift from B. A. Hassan. The Wnt5 400 allele is a null allele generated by the imprecise excision of an adjacent P element, resulting in the deletion of most of the Wnt5 open reading frame [14]. The drl 2 allele is a null allele generated by the imprecise excision of a P element in the 5' non-coding region, resulting in the deletion of the 5' regulatory sequences and the first exon of drl [60]. To generate a new drl null allele on the Mz19-Gal4 chromosome, exons 2, 3 and 4 of the drl locus (encompassing~90% of the drl open reading frame) were excised from the chromosome by Crispr/Cas9-mediated deletion using the sgRNAs GACAAGTGAAGGGGTGCTGT and GACACCTGTAGTGAGAGGTA following a published protocol [61]. Ten individual offspring from Crispr/Cas9 fathers were crossed to Adv/CyO virgins to establish lines. The lines were screened by PCR using deletionspanning primers to identify potential drl mutants. The PCR products were sequenced and one mutant, drl JS , with the expected precise deletion of the drl locus in the Mz19-Gal4 background was chosen for study. The drl JS mutation failed to complement the AL phenotype of the drl 2 mutation, consistent with drl JS being a null allele.

Quantification of glomerular rotation
Two different quantification strategies were employed. To quantify the displacements of single glomeruli labeled by the Or-Gal4 drivers [6], the angle subtended at the VA6 glomerulus (close to the center of the AL) by the dorsal pole and the labeled glomerulus (in the dorsal/0˚! lateral/90˚! ventral/180˚! medial direction/270˚) was measured. To quantify the rotation of the DA1 and VA1d glomeruli [22] around each other, the angle subtended at the VA1d glomerulus by the dorsal pole and the DA1 glomerulus (in the dorsal/0˚! lateral/90˚! ventral/ 180˚! medial/270˚direction) was measured. Data were collected, analyzed and plotted using the Prism statistical software. For two-sample comparisons, unpaired Student's t-tests were applied. For comparisons among more than two groups, one-way AVOVA tests were used followed by Tukey's test. Rose diagrams were plotted using the Excel program.