Alteration of Nrp1 signaling at different stages of olfactory neuron maturation promotes glomerular shifts along distinct axes in the olfactory bulb

In vertebrates, topographically organized maps in the brain process various sensory modalities (Accolla et al., 2007; Carleton et al., 2010; Luo and Flanagan, 2007; Vincis et al., 2012). These maps usually reflect an organization that is already present at the level of the sensory structures, but not always (O'Leary et al., 1999). The olfactory system is one of these exceptions, since functionally identical sensors – that is neurons expressing the same odorant receptor (OR) gene – are scattered in the sensory epithelium of the nose (Buck and Axel, 1991). It is only at the level of the sensory neuron axonal projections in the olfactory bulb (OB) that an organized topographical map emerges: generally, two glomeruli in each OB correspond to each sensory population expressing a given OR (Mombaerts et al., 1996; Ressler et al., 1994; Vassar et al., 1994). These two glomeruli are in stereotyped locations that reflect a mirror image defined by a symmetry axis cutting the bulb approximately sagittally. Given the size of the OR gene repertoires in mammalian genomes (over 1000 OR genes in the mouse) and the thousands of sensory neurons expressing each of these receptor genes, the system faces a remarkable wiring and pathfinding problem. This complex task is to be completed in utero and perinatally, but also during the life of the animal, since olfactory sensory neurons (OSNs) are constantly renewed through adulthood.

Significant progress has been made in our understanding of how OSNs target to the appropriate glomeruli in the bulb. Different mechanisms are at work to establish the bulbar map and involve both activity-dependent sorting of axonal projections and genetically determined cues (Chen and Flanagan, 2006; Imai and Sakano, 2008; Sakano, 2010). Two levels of guidance have been identified. The first directs axons to their target zones in the bulb and the second guides local axon sorting, which ends in glomerular segregation (Feinstein et al., 2004; Feinstein and Mombaerts, 2004; Imai et al., 2006, 2009; Nakashima et al., 2013; Nishizumi and Sakano, 2015; Rodriguez-Gil et al., 2015; Serizawa et al., 2006; Takeuchi et al., 2010; Wang et al., 1998). The current model explaining global targeting (at least on the medial side of the bulb) proposes the existence of guidance cues that determine the position of the target along two axes in the bulb: one dorsoventral and the other anteroposterior. The dorsoventral axis roughly correlates with the anatomical distribution of the sensory neurons in the neuroepithelium, and with the graded and complementary expression of neuropilin 2 (Nrp2) and semaphorin 3F (Sema3F) along this axis (Takeuchi et al., 2010). Neurons located in the dorsal zone of the nasal cavity target to the dorsal bulb, while those located more ventrally project to more ventral parts. A second axis directs sensory fibers along the anteroposterior length of the bulb. Elegant studies have suggested that this anteroposterior targeting is dependent on the level of cAMP that is produced by the spontaneous activity (that is agonist-independent activity) of the OR expressed by each sensory neuron (Imai et al., 2006, 2009; Nakashima et al., 2013). These cAMP concentrations are then translated into specific levels of Nrp1 (among others guidance molecules) that are responsible for guiding olfactory axons along the anteroposterior axis of the bulb (Imai, 2012). The timing during the maturation of the OSN at which Nrp1 plays this role has however not been defined.

To precisely evaluate the role played by Nrp1 in the establishment of the bulbar topographical map, we followed the path of genetically defined olfactory sensory populations deleted for Nrp1 at different stages during the building of the bulbar map, and competing with functionally identical sensory populations expressing Nrp1. Our data show a crucial time-dependent function played by Nrp1 in the establishment of the olfactory axonal map topography.

Nrp1 is differentially transcribed by OSN populations. As previously reported, robust expression is observed in neurons that target the medial and the lateral parts of the bulb, with those located anterolaterally and posteromedially expressing Nrp1 the most strongly (Fig. 1A,B) (Miller et al., 2010; Pasterkamp et al., 1998; Schwarting et al., 2000). Glomeruli located on the medial side of the bulb exhibit a gradient of Nrp1 expression along the anteroposterior axis (Fig. 1B) (Imai et al., 2006). However, this is true at a global anteroposterior level, not always at the glomerular level, since one finds glomeruli that are intensely marked for Nrp1 surrounded by glomeruli expressing barely detectable levels of Nrp1 (Fig. 1B,D,G,J). To precisely evaluate the role of Nrp1 in the building of the olfactory map, we analyzed the axonal projections of three different Nrp1-expressing sensory populations. We chose M71-, M72- and MOR23-expressing neurons because (1) they express Nrp1 (Fig. 1C-K) (Dal Col et al., 2007), (2) their axonal projections are robust, stereotyped and have already been extensively studied, in particular using the knock-in alleles M71 (Olfr151)^lacZ, M72 (Olfr160)^lacZ and MOR23 (Olfr16)^GFP (Feinstein et al., 2004; Feinstein and Mombaerts, 2004; Vassalli et al., 2002), and (3) their projections are located in areas easily visualized by whole-mount analyses.

To disrupt Nrp1 expression, we opted for an approach to specifically remove Nrp1 from OSNs. We made use of a mouse line bearing Nrp1^flox/flox alleles (Gu et al., 2003) crossed to a mouse expressing Cre recombinase specifically in maturing OSNs (via an Omp^cre allele) (Li et al., 2004). To validate the Omp-driven deletion approach, bulbs from Nrp1^flox/flox;Omp^+/cre mice were stained with an Nrp1 antibody; a lack of Nrp1 staining was observed in the glomerular layer (Fig. S1A-G). M71^lacZ;Nrp1^flox/flox;Omp^+/cre, M72^lacZ;Nrp1^flox/flox;Omp^+/cre and MOR23^GFP;Nrp1^flox/flox;Omp^+/cre mice were generated (n=10, 22 and 20, respectively). The M71, M72 and MOR23 axonal projections were analyzed by whole-mount analyses, and were compared with their respective controls (n=20, 28 and 12, respectively; Fig. 2). Homogenous glomeruli were formed by M71, M72 and MOR23 neurons. Whereas no major targeting alteration was observed in MOR23^GFP;Nrp1^flox/flox;Omp^+/cre mice (Fig. 2J,O), a drastic phenotype was observed in M71^lacZ;Nrp1^flox/flox;Omp^+/cre and M72^lacZ;Nrp1^flox/flox;Omp^+/cre mice, in which glomeruli were shifted dorsally (Fig. 2F-I,L,N). M72 axonal fibers often crossed the dorsal midline to innervate the contralateral M72 glomerulus (Fig. 2H,I,M). M71-expressing neurons exhibited an even more drastic dorsalization of their glomeruli, since the lateral and the medial glomeruli were fused on the dorsal part of the bulb (Fig. 2F,G,K). Considering a position of the mouse head in which the dorsal part of the snout is horizontal (which corresponds to an angle of +113° for the cribriform plate; Fig. 2L), the M71 and M72 mean glomerular shift corresponds to a dorsalization of 490±210 and 468±127 µm (±s.d.) with +7 and +15° relative to the dorsoventral axis, respectively (Fig. S4A,B). This shift was very different from that we were expecting, since previously observed I7 glomeruli lacking Nrp1, on which the anteroposterior model was built, were anteriorized (Imai et al., 2009). This published and our own approach differed in the timing at which Nrp1 was excised, since the Cre recombinase was under the control of a promoter active in mid-mature neurons [the I7 OR] or in late-mature neurons (Omp), respectively. We thus hypothesized that the timing at which Nrp1 is deleted could be critical in the mistargeting. We therefore expanded our Cre-mediated Nrp1 allele inactivation to target the different phases of OSN maturation.

For early deletion of Nrp1 during OSN maturation we chose the Gng8 (Gγ8) gene, which is expressed in precursor and immature olfactory neurons (Hanchate et al., 2015; Ryba and Tirindelli, 1995). We generated a knock-in mouse line bearing a Gng8^cre allele, in which a polycistronic cassette containing an IRES followed by the Cre recombinase coding sequence (CDS) was inserted by homologous recombination immediately after the Gng8 stop codon (Fig. 3A). To evaluate this novel transgenic line, it was crossed with a mouse line bearing an R26^flRFP reporter allele to generate Gng8^cre;R26^flRFP mice. Red fluorescence in the olfactory epithelium was evaluated. In Gng8^cre;R26^flRFP mice over 90% of the olfactory epithelium was labeled, including the whole apical and medial layers and the upper part of the basal layer (Fig. 3B, Fig. S2A,B). This was expected since neurogenesis takes place at the very basal part of the epithelium, with neurons moving apically while differentiating. To further validate the Gng8-mediated early Nrp1 deletion approach, bulbs from Nrp1^flox/flox;Gng8^cre/cre mice were stained with an Nrp1 antibody; a lack of Nrp1 staining was observed in both the glomerular and the nerve layers (Fig. S1H-J).

For expression of Cre recombinase at the middle stage of the axonal projection process, we opted for a knock-in mouse line bearing an M71^cre allele that left the M71 CDS intact (Fig. 3B). OR transcription can be observed prior to OMP expression (Hanchate et al., 2015; Rodriguez-Gil et al., 2015). To control for the stage at which the specific knock-in lines express the Cre recombinase, Gng8^cre;R26^flRFP, M71^cre;R26^flRFP and Omp^cre;R26^flRFP mice were analyzed, and the positions along the apicobasal axis of the septal sensory epithelium of the youngest red-fluorescent neurons were located (that is, those closest to the basal lamina with a one-neuron-wide apicobasal sliding window; Fig. 3C). As expected, a coverage of the different OSN developmental phases was observed, with Gng8^cre being apparently transcribed before M71^cre, and M71^cre before Omp^cre (Fig. 3C).

In order to evaluate the effects of an earlier excision of the floxed Nrp1 exon, M71^lacZ;Nrp1^flox/flox;Gng8^+/cre, M72^lacZ;Nrp1^flox/flox;Gng8^+/cre and MOR23^GFP;Nrp1^flox/flox;Gng8^+/cre mice were generated (n=12, 22 and 18, respectively) and compared with their respective controls (n=8, 5 and 14, respectively; Fig. 4A-L). We initially assessed a potential general disruption of the glomerular arrangement by analyzing sections of P0 OBs from Nrp1^flox/flox;Gng8^cre/cre mice. We did not observe any obvious alterations (data not shown). Given the early inactivation of the Nrp1 gene with Gng8^cre, we expected a similar, although more pronounced, dorsalization phenotype than that observed using the Omp^cre driver. In Nrp1^flox/flox;Gng8^+/cre mice, we indeed observed a fusion of the lateral and medial M71 glomeruli (Fig. 4G,H,M,N) and a higher frequency of contacts between the contralateral M72 glomeruli (Fig. 4I,J,O,P). The M71 and M72 dorsal glomerular shift corresponds to a distance of 698±84 and 537±229 µm with an angle of −5 and +8, respectively (Fig. S4C,D), relative to the dorsoventral axis. Surprisingly, we also observed the presence of an anteriorized glomerulus in M71-, M72- and MOR23-labeled mice with an early deletion of Nrp1 (Fig. 4G-R). Relative to the endogenous glomeruli and following the anteroposterior axis (i.e. relative to the dorsal part of the snout), the mean positions of the M71, M72 and MOR23 supplementary anterior glomeruli were at a distance of 1827±97, 1688±227 and 442±98 µm with an angle of +30, +33 and −41°. Since the number of OSNs expressing a given receptor may affect their targeting or at least their propensity to form a stable glomerulus, we evaluated the possible alteration of OSN numbers in an Nrp1 null background. All X-Gal-stained OSNs located on the rostral part of the dorsal turbinate II up to the groove linking it to the basal turbinate II were counted on whole mounts of Gng8^cre;Nrp1^+/+ and Gng8^cre;Nrp1^flox/flox main olfactory epithelium. We found no significant difference between the two genotypes (189±32 for Nrp1 wild type versus 178±28 for Nrp1 null; n=4 and n=3 P50 mice, respectively).

As a parallel approach to the removal of Nrp1 at different stages of OSN development, we forced a reduction of Nrp1 expression in OSNs. We assumed that in mice bearing a single functional Nrp1 allele, the number of Nrp1 transcripts would be decreased. We analyzed the axonal projections of M72 fibers in Omp^+/cre;Nrp1^+/+;M72^lacZ and Omp^+/cre;Nrp1^+/flox;M72^lacZ mice. We found no significant differences in glomerular position between mice bearing one or two Nrp1 alleles (Fig. S3A-D), which suggests either that the readout of Nrp1 levels is not fine tuned or that there is compensatory expression of Nrp1 from the remaining allele when one is ineffective.

Although precisely timed during OSN development and restricted to OSNs, the late (Omp-driven) and early (Gng8-driven) Nrp1 deletion did affect all OSNs. To investigate the possible contribution of the non-labeled and possibly misrouted Nrp1-deficient OSNs to the observed phenotypes, we took advantage of the monogenic expression of OR genes to exclusively inactivate Nrp1 in OSNs expressing the M71 OR gene, which is highly transcribed and starts to be expressed early during OSN development (Fig. 3C). R26^flRFP;Nrp1^flox/flox;M71^cre/cre mice were generated. All of the M71 lateral and medial glomeruli were fused dorsally, and the presence of a very rostral glomerulus was observed (n=22; Fig. 5P). This phenotype was very similar to that observed in M71^lacZ;Nrp1^flox/flox;Gng8^+/cre mice.

In addition, we also took advantage of the monoallelic expression of OR genes and generated compound heterozygotes by adding an M71 wild-type allele driving the expression of a GFP fluorophore. R26^flRFP;Nrp1^flox/flox;M71^cre/GFP mice were thus generated, allowing the projections of Nrp1-deficient M71 axons (in red) and of M71 wild-type axons (in green) to be visualized in the same animal (Fig. 5). We observed the formation by Nrp1-deficient fibers of dorsally located M71 glomeruli (either linked or fused) in more than 90% of the OBs analyzed (n=62; Fig. 5E-L,P, Fig. S5A,B), similar to what was observed in M71^lacZ;Nrp1^flox/flox;Gng8^+/cre and M71^lacZ;Nrp1^flox/flox;Omp^+/cre mice. Wild-type M71 fibers of R26^flRFP;Nrp1^flox/flox;M71^cre/GFP mice did often project to two glomeruli, the position of which was relatively similar to that observed in wild-type mice. However, a significant proportion of the Nrp1-deficient fibers innervated the wild-type M71 glomeruli (Fig. 5F-H, Fig. S5A-C). Even more surprising, numerous Nrp1-expressing M71 axons were rerouted towards the dorsal Nrp1-deficient glomerulus (Fig. 5F,H,K-O, Fig. S5B). Finally, a small anterior glomerulus formed by Nrp1-deficient fibers was observed (Fig. 5I,J,P), reminiscent of what was seen in Nrp1 early-deleted mice. In this context, the formation of an anterior glomerulus and the dorsalization must come about through cell-autonomous mechanisms and not depend on any other populations of axons with Nrp1 having been deleted.

Sema3A and VEGF165 represent the two main ligands for Nrp1 (Schwarz and Ruhrberg, 2010). Given the widespread role played by Sema3A in the development of the nervous system, and of the olfactory system in particular (Imai et al., 2009; Pasterkamp et al., 1998; Schwarting et al., 2000, 2004; Taniguchi et al., 2003), we evaluated its role in the projection of M71- and M72-expressing fibers. We took two parallel and complementary approaches: an Nrp1 hypomorphic mutant allele (Nrp1^sema) that encodes a product unable to recognize Sema3A but still able to respond to VEGF (Renzi et al., 2000), and a constitutive Sema3a null allele (Sema3a^del) (Fig. 6, Fig. S6). M71^GFP;Nrp1^sema/sema mice (n=18, Fig. S6B) and M72^lacZ;Nrp1^sema/sema mice (n=12, Fig. 6E,F) were analyzed. Similar to what was observed in Nrp1^flox/flox;Gng8^+/cre animals, lateral and medial axonal projections of M71 and M72 did form an anteriorized glomerulus and projected to a dorsalized glomerulus relative to the wild-type glomerular position (Fig. 6A-D,E,F,I,J, Fig. S6B,D). A similar pattern was observed in M71^GFP;Sema3a^del/del and M72^lacZ;Sema3a^del/del mice (n=9, Fig. S6C,D and n=8, Fig. 6G,H,K,L, respectively). However, the M71 and M72 projection patterns in Nrp1^sema/sema and Sema3a^del/del animals were more variable and disorganized (dorsalized glomeruli were often observed as pairs or triplets and not frequently linked or fused) than those observed in late or Nrp1 early-deleted mice (Fig. 6E-L, Fig. S6B-D). Interestingly, the projection pattern of M72-expressing fibers in constitutive mutants (Sema3a^del/del or Nrp1^sema/sema) or in the early Nrp1 deletion context is reminiscent of what is observed in Adcy3^del/del animals (Fig. S7) (Dal Col et al., 2007). This is consistent with what we know about this olfactory cyclase, which is not only crucial for olfactory transduction, but also necessary for Nrp1 expression in OSNs (Dal Col et al., 2007).

Here, we dissected the role played by Nrp1 in the establishment of the topographic map built by axonal projections of OSNs. Using parallel genetic approaches in vivo, we singled out and followed the axonal fibers of three functionally different sensory populations that were each defined by the expression of a specific OR gene. We altered the function of Nrp1 in these populations at different steps during axonal targeting and in different competitive environments. We observed that, as predicted by the current model explaining the anteroposterior positioning of glomeruli in the bulb, M71-, M72- and MOR23-expressing sensory neurons formed anteriorized glomeruli relative to their wild-type position when the interaction between Nrp1 and Sema3A was impaired constitutively or early during OSN maturation. However, in addition to these anterior glomeruli we observed that both M71- and M72-expressing neurons also formed medial glomeruli located at their wild-type anteroposterior position. Surprisingly, these glomeruli exhibited a dorsal shift to such an extent that the medial and the lateral glomeruli were often fused.

In a previous study, Nrp1 was shown to play a role in sorting fibers inside axon bundles projecting toward the OB (Imai et al., 2009). In that report, in a pan-OSN Nrp1 knockout background the axonal projections of I7-expressing OSNs exhibited split projections, with one glomerulus located in its wild-type position and the other anteriorized relative to the latter. When the Nrp1 deletion was restricted to a specific OSN population (I7-expressing OSNs), a single anteriorized glomerulus was observed (Imai et al., 2009). The hypothesis provided to explain the difference between the two phenotypes was that the relative Nrp1 levels among axons determine OSN projection sites. However, our observation of an anterior in addition to a posterior glomerulus in Nrp1^flox/flox;M71^cre/GFP;R26^flRFP bulbs is difficult to reconcile with this view. We propose other explanations, not necessarily exclusive, for the presence of the two well-defined and stereotyped M71, M72 or MOR23 glomeruli on the medial or dorsal sides of the Nrp1-deficient bulb. First, two different populations of neurons expressing the same OR but differing in their expression of other proteins might coexist in the olfactory sensory epithelium and might be revealed in an Nrp1-deficient context. Second, the anterior or dorsal projections could reflect axons that targeted the bulb at different times during development. The two target positions would be maintained during the life of the animal, with the early projections guiding later incoming fibers. Finally, depending on their location in the neuroepithelium, sensory neurons, which are likely to encounter different signals at different times during fiber migration, could simply be prone to innervate different domains in the absence of Nrp1.

The absence of an anterior glomerulus in an Nrp1 late-deletion context suggest that young sensory neurons are differentially sensitive to Nrp1 expression relative to older ones, or at least that the lack of Nrp1-derived signals leads to different outcomes depending on when the signals fade. Sensory neurons would thus target anteriorly when Nrp1 is deleted early during their differentiation and dorsally when deleted late. The presence of two glomeruli in M71^lacZ;Nrp1^flox/flox;Gng8^+/cre mice would thus reflect neurons that lost Nrp1 as soon as Gng8 was expressed, and other neurons with Nrp1^flox alleles that took longer to be recombined.

We previously described an altered M72 and MOR23 projection pattern in Adcy3^del/del mice. Among various transcriptional differences in guidance-related genes between Adcy3^+/+ and Adcy3^del/del OSNs, we found (among other guidance molecules) Nrp1 to be significantly downregulated. Interestingly, the Nrp1-deficient phenotype reported here is very similar to that observed in Adcy3^del/del mice, suggesting that the downregulation of Nrp1 plays a crucial role in the Adcy3^del/del axonal guidance alterations.

We reported here the presence of a dorsal fused glomerulus in Nrp1^flox/flox;M71^cre/GFP;R26^flRFP mice. Given that we did not observe M71 glomeruli in the dorsal OB of wild-type mice, the OB dorsal domain might be repulsive to Nrp1-positive fibers (e.g. M71-expressing neurons). However, the presence of Nrp1-positive fibers in the dorsal fused glomerulus of these Nrp1^flox/flox;M71^cre/GFP;R26^flRFP mice indicates that this potential repulsion may be overcome by other signals, such as potential homotypic interactions between like fibers (i.e. fibers expressing the same OR).

As previously discussed, the current model of ‘global' axonal targeting of OSNs involves two axes on the OB: an Nrp2-dependent dorsoventral axis and an Nrp1-dependent anteroposterior axis. The rostralization of the glomeruli we report here is compatible with this model, but the dorsalization of M71 and M72 glomeruli suggests an additional role played by Nrp1 in the targeting along the dorsoventral axis. However, considering two fixed and orthogonal axes to explain OSN targeting might tackle the problem with an inadequate tool which, although agreeable to our topographical view of the world and although useful as a descriptor or even as a working hypothesis, might not really reflect cues provided by the biological signals that direct the targeting of olfactory sensory axons in the bulb.

The Gng8^cre allele was generated by gene targeting in E14 embryonic stem cells (ESCs). The targeting vector (TV) was built by modifying via recombineering techniques a bacterial artificial chromosome (BAC, bMQ314G20) containing the Gng8 CDS. The final TV contained a 5′ homology arm (HA) corresponding to 8.9 kb of genomic sequence ending with the Gng8 stop codon. This 5′ HA was followed by a PacI-flanked reporter cassette comprising an internal ribosome entry site (IRES) sequence followed by the Cre recombinase CDS and an Frt-NeoR-Frt (FNF) selection cassette. The 3′ HA corresponding to 5.3 kb of genomic sequence downstream of the Gng8 stop codon followed the reporter cassette. The linearized TV was electroporated into ESCs and G418-resistant colonies were screened by Southern blot using a 469 nt RFLP probe starting with the start codon of the Ptgir gene located 11 kb downstream of Gng8 (Fig. 3). Positive clones showed a 14 kb EcoRV fragment, whereas the wild-type allele produced a 46 kb band on the gel. The FNF selection cassette was further excised by crossing Gng8^cre mice with a transgenic mouse line expressing FLP recombinase in the germ line. All animals presented in this study lacked the FNF cassette. All mice were bred on a mixed C57BL/6-CBA background, except for Sema3a null mice (generated by crossing mice bearing Sema3a^flox and CMV^cre alleles) that were bred on a CD-1 background to increase the survival of mutant pups (Sibbe et al., 2007; Vieira et al., 2007). The MOR23 (Olfr16)^GFP, M72 (Olfr160)^lacZ, M71 (Olfr151)^lacZ, M71 (Olfr151)^GFP, M71 (Olfr151)^cre, Omp^cre, CMV^cre, Nrp1^flox, Nrp1^sema, Sema3a^flox, Adcy3^del and R26^flRFP alleles were previously described (Dupe et al., 1997; Feinstein et al., 2004; Feinstein and Mombaerts, 2004; Gu et al., 2003; Li et al., 2004; Luche et al., 2007; Taniguchi et al., 1997; Vassalli et al., 2002; Wong et al., 2000; Zheng et al., 2000). Age- and sex-matched littermates were used when comparing mutant and wild-type animals.

All animal procedures were performed in accordance with the guidelines and regulations of the institution and of the state of Geneva.

Animals were dissected and either directly imaged for fluorescence or fixed with 4% paraformaldehyde and stained with X-Gal as previously described (Rodriguez et al., 1999). No association was observed between projection patterns of the right and the left bulbs of a given animal. Therefore, position data were pooled between right and left bulbs. Whole-mount images from X-Gal staining were taken on a Leica MZFLIII stereo microscope. For fluorescent reporters, images were taken with a Zeiss SteREO Lumar V12 stereo microscope or a Zeiss LSM700 confocal microscope. All medial view images were cropped and tilted to align the dorsal part of the nasal cavity with the image top frame.

After dissection, head samples were degased and fixed overnight in 4% paraformaldehyde, then transferred to 15% sucrose for 12 h, followed by 30% sucrose for 12 h. Samples were subsequently embedded in O.C.T. (VWR), frozen and 14-16 µm cryosections prepared using a cryostat. Slides were conserved at −80°C until use. For immunostaining, slides were pre-incubated for 30 min in blocking solution at room temperature (2% normal goat serum, 3% BSA, 0.1% Triton X-100 in PBS) followed by incubation overnight at 4°C with primary antibodies diluted in the same blocking solution. Primary antibodies were: goat anti-Nrp1 (1:100, R&D Systems AF566), chicken anti-GFP (1:400, Abcam ab13970), rabbit anti-RFP (1:400, Abcam ab62341) and chicken anti-β-galactosidase (1:500, Abcam ab9361). Slides were washed three times for 15 min each with 0.1% Triton X-100 in PBS. A secondary incubation of 90 min at room temperature with one of the following antibodies was performed: donkey anti-goat A555 (1:400, Invitrogen A21432), goat anti-chicken A488 (1:400, Invitrogen A11039) or goat anti-rabbit A555 (1:400, Invitrogen A21428). Finally, slides were washed three times for 15 min each with 0.1% Triton X-100 in PBS, treated for 5 min with DAPI (1:5000 in PBS) and mounted with DABCO (Sigma) in glycerol. Images were acquired using a Zeiss Axioplan 2 or a Leica DM5500 microscope.

To generate standardized OB images of mapped glomerular positions, we first generated a two-dimensional prototypic OB medial outline out of 80 representative images of medial OBs taken from 7-week-old C57BL/6 male mice. The OB outline of these 80 images was traced in Adobe Illustrator, compiled in an ImageJ stack (Schneider et al., 2012), and processed using the ‘stackreg' alignment plugin (Thevenaz et al., 1998) to generate a reference outline. The OB outline and the glomerular positions of all experimental medial OB images were then aligned to this reference using the same procedure. M71 axonal projections in OBs from mice bearing M71^lacZ/lacZ or M71^+/lacZ alleles were considered equally, as we did not observe any differences between them (data not shown). The same applies to mice bearing the M72^lacZ/lacZ or M72^+/lacZ alleles.

The x,y coordinates of each glomerulus were extracted by exporting Adobe Illustrator image files (.ai) into scalable vector graphics files (.svg). Datasets containing the coordinates of each glomerular population were then built for each genotype and for each glomerular population. These datasets were then imported into RStudio, and the ‘aspace' module (Randy Bui et al., 2012) was used to compute the mean coordinates (x,y) of a selected glomerular population as well as a standard deviational ellipse (x, y, width, height and rotation angle) that represents the glomerular dispersion in two dimensions. One standard deviational ellipse covers ∼68% of the glomeruli when a spatial normal distribution is observed (i.e. most glomeruli condensed in the center, fewer in the periphery).

We thank Francisco Resende and Véronique Jungo for expert technical help.