Research Article

Pre-Target Axon Sorting Establishes the Neural Map Topography

See allHide authors and affiliations

Science  31 Jul 2009:
Vol. 325, Issue 5940, pp. 585-590
DOI: 10.1126/science.1173596

Mapping the Neuronal Map

In vertebrates, sensory information is topographically represented as a neural map in the brain. How is the neural map formed in the brain? Nearly a half-century ago, Roger Sperry proposed the “chemoaffinity” model, in which the positional cues on the target determine the axonal projection site, thereby establishing the topographic neural map. However, molecular mechanisms of topographic map formation remain controversial. Imai et al. (p. 585, published online 9 July; see the Perspective by Miyamichi and Luo) now report that the topographic map is formed by axon-axon interactions before the axons reach the target. In the mouse olfactory system, the topography of the map is determined by the relative expression levels of a guidance receptor, Neuropilin-1, and its repulsive ligand, Semaphorin-3A, expressed in axons. Topographic organization occurs even in the absence of the target, the olfactory bulb. These findings require that Sperry's model, which suggests that only the targets determine the topography of neural maps, needs to be reconsidered.


Sensory information detected by the peripheral nervous system is represented as a topographic map in the brain. It has long been thought that the topography of the map is determined by graded positional cues that are expressed by the target. Here, we analyzed the pre-target axon sorting for olfactory map formation in mice. In olfactory sensory neurons, an axon guidance receptor, Neuropilin-1, and its repulsive ligand, Semaphorin-3A, are expressed in a complementary manner. We found that expression levels of Neuropilin-1 determined both pre-target sorting and projection sites of axons. Olfactory sensory neuron–specific knockout of Semaphorin-3A perturbed axon sorting and altered the olfactory map topography. Thus, pre-target axon sorting plays an important role in establishing the topographic order based on the relative levels of guidance molecules expressed by axons.

In the vertebrate nervous system, sensory information is spatially encoded in the brain, forming topographic maps that are fundamental for cognition and higher-order processing of sensory information (1, 2). Molecular mechanisms of topographic map formation have been extensively studied in the visual system. The visual image on the retina is roughly preserved in the tectum, which receives retinal ganglion cell axons. Nearly 50 years ago, Sperry proposed the “chemoaffinity hypothesis,” in which target cells present chemical cues to guide axons to their destinations (3). Axonal projection of retinal ganglion cells is instructed by several pairs of axon guidance molecules that demonstrate graded expression in the retina and tectum (1, 2).

Olfactory information is also encoded in a topographic map formed on the olfactory bulb (OB), a part of the forebrain. In rodents, odors are detected with ~1000 types of odorant receptors (ORs) expressed in olfactory sensory neurons (OSNs) in the olfactory epithelium (4). Each OSN expresses only one functional OR gene (5, 6). Furthermore, OSNs expressing a given type of OR converge their axons to a specific glomerulus on each glomerular map in the OB (79). During olfactory development, OSN axons are guided to approximate locations in the OB by the combination of dorsal-ventral patterning, based on anatomical locations of OSNs in the olfactory epithelium (10), and anterior-posterior patterning, regulated by OR-derived cyclic adenosine monophosphate (cAMP) signals (11, 12). The glomerular arrangement along the dorsal-ventral axis appears to be determined by axon guidance molecules expressed in a graded manner along the dorsomedial-ventrolateral axis in the olfactory epithelium, such as Robo-2 (13) and Neuropilin-2 (14). Unlike dorsal-ventral positioning, anterior-posterior positioning of glomeruli is independent of positional information in the olfactory epithelium. Instead, OR-specific cAMP signals determine the expression levels of Neuropilin-1 (Nrp1) in OSN axon termini, forming a gradient of Nrp1 (11). Thus, the olfactory system also uses gradients of axon guidance molecules to form the topographic map.

How then do guidance molecules regulate topographic map formation? Does map formation solely depend on axon-target interaction? Topographic order emerges in axon bundles, well before they reach the target (15, 16). Here, we studied the pre-target sorting of OSN axons and its role in topographic map formation in the mouse olfactory system.

Nrp1 regulates axonal projection of OSNs along the anterior-posterior axis. OR-derived cAMP signals regulate the axonal projection of OSNs along the anterior-posterior axis in the OB; low cAMP leads to anterior positioning and high cAMP leads to posterior positioning (11). Furthermore, the levels of Nrp1 in OSN axon termini correlated with the level of cAMP signals (11).

We found that the Nrp1 levels determine the glomerular positioning along the anterior-posterior axis. When Nrp1 was overexpressed in OR-I7–expressing OSNs (fig. S1), projection sites shifted posteriorly relative to the control (Fig. 1A and fig. S1). In contrast, when Nrp1 was knocked out specifically in I7 OSNs, the projection sites shifted anteriorly relative to the control (Fig. 1A and fig. S2). In the pan-OSN Nrp1 knockout, however, projection sites for I7 often split into anterior and posterior areas (fig. S3). If absolute Nrp1 levels determine glomerular positioning, all glomeruli should form in the anterior OB in the pan-OSN knockout, and the results for I7 OSNs should be the same between the I7-specific knockout and pan-OSN knockout. These results indicate that the relative Nrp1 levels among axons determine the OSN projection sites.

Fig. 1

Projection and axon sorting of OSNs. (A) Genetic manipulation of Nrp1 expression levels in OSNs. Whole-mount fluorescent views of OBs (medial surface) were analyzed [age, postnatal day 14 (P14)]. For the loss of function of Nrp1, an I7-ires-Cre mouse (labeled with gap-YFP; yellow) was crossed to the floxed Nrp1 line. The projection site for the I7 (labeled with gap-CFP; cyan) was analyzed as an internal control in the same animal. In the Nrp1flox/flox background, Nrp1 was specifically knocked out in OSNs expressing I7-ires-Cre. As a result, projection sites were shifted anteriorly in the OB. In the Nrp1flox/WT background, where the Nrp1 level was decreased by half, the anterior shifts were intermediate to those in the Nrp1flox/flox background. In the Nrp1WT/WT background, CFP- and YFP-labeled OSN axons coconverged. For the gain of function of Nrp1, we generated I7-ires-Nrp1 (labeled with gap-mCherry; red) and compared the projection site to that of the control construct, I7-ires-Cre. In the double transgenic mice, I7-ires-Nrp1 glomeruli were found posterior to the I7-ires-Cre glomeruli. Statistical data are in fig. S2. Scale bars, 500 μm. (B) Pre-target axon sorting of OSNs. The axons of dorsal-zone (D-zone) OSNs course through the bundle on the dorsal roof of the olfactory epithelium before projecting to the D domain of the OB. The D domain comprises DI and DII domains in the OB. The DI domain is located in the most dorsal part of the OB; the DII domain is just ventral to DI and is further divided into DII-P (Nrp1-high; posterior) and DII-A (Nrp1-low; anterior). DI axons were stained with DBA lectin (cyan). DII-P axons were immunostained with antibodies to Nrp1 (red). For labeling of DII-A axons, OSN axons expressing I7(RDY)-ires-gapYFP were immunostained with antibodies to YFP (green). Three types of axons projecting to the DI, DII, and DII-P regions in the OB are intermingled in the bundle near the olfactory epithelium but segregate to form a tripartite organization as they approach the OB. Coronal sections (#1 to #4, each separated by 64 μm) are shown. DI, DII-A, and DII-P axons are sorted to the medial, central, and lateral areas, respectively. Scale bars, 100 μm. OE, olfactory epithelium; A, anterior; P, posterior; D, dorsal; V, ventral; M, medial; L, lateral; ires, internal ribosome entry site.

Pre-target axon sorting in the bundle. How do the relative levels of Nrp1 determine the anterior-posterior positioning of glomeruli in the axonal projection of OSNs? To determine where the organization occurs for the olfactory map topography, we analyzed the axon bundles of dorsal-zone (D-zone) OSNs that project to the dorsal domain (D domain) of the OB. The D domain OB comprises two regions, DI and DII; DI is represented by class I ORs, and DII is represented by class II ORs. Class I and class II ORs are phylogenetically distinct and their glomeruli are segregated in the OB (17). We subdivided DII into two areas on the basis of Nrp1 expression level (18): DII-P is the posterior portion innervated by Nrp1-high axons, and DII-A is the anterior region innervated by Nrp1-low axons. Thus, the D domain can be divided into three areas: DI, DII-A, and DII-P (Fig. 1B).

Axon bundles that project to the D-domain OB were analyzed in neonatal mice by staining serial coronal sections from the anterior olfactory epithelium through the OB. Within the bundle, DI axons stained with DBA-lectin (fig. S4), DII-P axons stained with Nrp1 antibodies, and DII-A axons stained with neither. To visualize the Nrp1-low axon group (DII-A), we stained for yellow fluorescent protein (YFP)–positive OSNs that express the mutant OR gene I7(RDY)-ires-gapYFP; the I7(RDY) receptor is unable to activate downstream cAMP signals, thereby causing a lack of Nrp1 expression (11). Early in the trajectory of OSN projection, DI, DII-A, and DII-P axons were intermingled within the bundle. As the bundle progressed posteriorly, however, the axons segregated into a tripartite organization before they reached the target OB (Fig. 1B). Within the bundle, DI, DII-A, and DII-P axons were sorted to the medial, central, and lateral areas, respectively. Furthermore, there was a topographic order even within the DII-P axon group. Anterior-posterior positioning of glomeruli in the OB is well correlated with the central-lateral localization of axons in the bundle (fig. S5). Thus, the olfactory map topography emerges within the axon bundle prior to the axon-target interactions in the OB.

Heterotypic OSN axons segregate even without the OB (19, 20). We analyzed the Gli3 mutant mouse (Pdn/Pdn), where the OB is completely absent, and found that the pre-target axon sorting indeed occurred (Fig. 2A) (21). In the Pdn/Pdn mutant, OSN axons demonstrated a graded anterior-posterior topography among the olfactory fibers in the cranial cavity (Fig. 2B), which supports the notion that the anterior-posterior topography can be formed, at least partially, without axon-target interactions.

Fig. 2

Anterior-posterior topography is established without the OB. The OB-less mutant Pdn/Pdn was analyzed. The Pdn/Pdn mouse has an insertional mutation within the Gli3 gene coding for a transcription factor (21), resulting in agenesis of the OB. In the cranial cavity of these mutant mice, OSN axons form a fibrocellular mass (FCM). (A) Coronal sections of axon bundles from the wild-type and Pdn/Pdn mutant mice were analyzed (age, P0). OSN axons expressing I7(RDY)-ires-gapYFP were immunostained with antibodies to YFP. Axonal segregation occurred without the OB in the Pdn/Pdn mutant (n = 6/6). Scale bars, 100 μm. (B) Horizontal OB sections of the wild-type and mutant mice (age, P0) were immunostained with antibodies to Nrp1 (red) and gap43 (green); gap43 is an axonal marker for OSNs. In the glomerular layer of the wild-type OB, Nrp1 shows an anterior-low–posterior-high gradient (11). Similarly, in the mutant mouse (Pdn/Pdn), Nrp1 demonstrates an anterior-low–posterior-high gradient within the FCM in the absence of the OB. Nrp1 and gap43 immunoreactivities were quantified along the anterior-posterior axis of the FCM (along the dotted line). Mean signal intensities are normalized to 50. Data are means ± SD (n = 8). OE, olfactory epithelium; Ctx, cerebral cortex. Scale bars, 500 μm.

Nrp1 regulates pre-target axon sorting. Because Nrp1-high (DII-P) and Nrp1-low (DII-A) axons are segregated within the bundles, we investigated whether Nrp1 is involved in the pre-target axon sorting. Wild-type I7 axons were normally found among the Nrp1-positive DII-P axons in the lateral area in the bundle (Fig. 3A and fig. S6A). When Nrp1 was specifically knocked out in I7-expressing OSNs, these I7 axons moved to the central area (DII-A). In contrast to the wild-type I7, axons of the I7(RDY) mutant OSNs, which do not express Nrp1, were found in the central area (DII-A) of the bundle (Fig. 3B and fig. S6B). By simply expressing Nrp1 in these mutant OSNs with I7(RDY)-ires-Nrp1, axons moved to the lateral area (DII-P). Thus, Nrp1 indeed determines the sorting of DII-A (central) and DII-P (lateral) axons within the bundle.

Fig. 3

Nrp1 regulates axon sorting in the bundle. (A) The loss-of-function experiment. Axons for I7-ires-Cre were sorted to the lateral peripheral area (DII-P). When Nrp1 was knocked out specifically in I7-ires-Cre OSNs, axons were sorted to the central region (DII-A). (B) The gain-of-function experiment. Axons for I7(RDY) that do not express Nrp1 were sorted to the central area of the bundle (DII-A). When Nrp1 was expressed specifically in I7(RDY) OSNs in the I7(RDY)-ires-Nrp1 mouse, axons were sorted to the lateral periphery (DII-P). Quantitative analyses are in fig. S6. Coronal sections of axon bundles (age, P0) were analyzed. Changes in axon sorting in the bundle are schematically shown at the right. Scale bars, 100 μm.

Sema3A expressed by OSNs is required for axon sorting. Nrp1 is the receptor for the secreted repulsive ligand Semaphorin-3A (Sema3A) (22). Sema3A knockout disrupts proper targeting of OSNs (23, 24). We studied the possible involvement of Sema3A in pre-target axon sorting. Sema3A is expressed not only in the target, but also in OSNs. Single-cell microarray analysis revealed that Nrp1 and Sema3A genes are regulated in a complementary manner by OR-derived cAMP signals (fig. S7) (25). In situ hybridization of embryonic olfactory epithelium demonstrated complementary expression of Nrp1 and Sema3A (fig. S8A). To visualize the Sema3A-expressing OSN axons, we generated bacterial artificial chromosome (BAC) transgenic mice in which an axonal marker, gap-YFP, is expressed under the control of the Sema3A promoter (fig. S1). In the transgenic mouse, Sema3A-positive axons were found in the Nrp1-low area (DII-A) in the bundle (Fig. 4A and fig. S8B), which suggests that Nrp1-high axons (DII-P) are repelled by Sema3A-expressing axons (DII-A).

Fig. 4

OSN-derived Sema3A regulates not only pre-target axon sorting in the bundle, but also glomerular positioning in the OB. (A) Left panel: Segregation of Nrp1-expressing and Sema3A-expressing axons in the bundle (see also fig. S8B). The axonal marker protein gap-YFP was expressed under the Sema3A promoter in the BAC transgenic mice (age, P0). Nrp1-high DII-P axons were sorted to the lateral periphery (DII-P), whereas Sema3A-high axons were sorted to the central area (DII-A). Right panel: Axon sorting in the bundle in wild-type and OSN-specific Sema3A knockout mice. The conditional knockout mice were generated by crossing the OMACS-Cre line and the floxed Sema3A line (fig. S9). In the OSN-specific Sema3A knockout (age, P0), Nrp1-expressing axons (DII-P) spread out within the bundle, although they are still dense in the lateral region. I7(RDY) axons normally sorted to the central region (DII-A) are found in the medial region in the axon bundle. Scale bars, 100 μm. (B) Sorting and projection of I7-expressing OSN axons. In the wild-type mouse, I7 axons are sorted to the lateral periphery of the bundle. In contrast, in the OSN-specific Sema3A knockout, I7 axons are sorted to the central area, where DII-A axons are normally found (age, P0). Sorting patterns of I7 axons are schematically shown at the right (see also fig. S10B). Projection sites for I7 axons were analyzed in the OB (whole-mount medial views; age, P14). In the OSN-specific Sema3A knockout, I7 glomeruli were found anterior to the wild-type control. Results are summarized at the right; I7 glomeruli for the OSN-specific Sema3A knockout (14 bulbs, magenta) and wild-type littermates (12 bulbs, blue) are plotted on a schematic diagram of the medial OB (see also fig. S11A). Note that I7 OSN-specific Sema3A knockout did not affect the glomerular positioning (fig. S11B). Scale bars, 100 μm (axon bundles), 500 μm (OBs).

Because Sema3A is expressed in different cell types in olfactory tissues, we generated an OSN-specific knockout for Sema3A to investigate whether Sema3A in OSN axons is required for axon sorting within the bundle. We used the OMACS gene promoter, which is activated in the D-zone olfactory epithelium during early embryogenesis, to drive the expression of Cre recombinase (fig. S9) (26, 27). The OMACS-Cre mouse was crossed to the floxed Sema3A mouse (28) to generate the OSN-specific conditional knockout. In wild-type mice, Nrp1-positive DII-P axons are sorted to the lateral periphery of the bundle. In the conditional Sema3A knockout, however, the DII-P axons were no longer confined to the lateral periphery (Fig. 4A). Concomitantly, DII-A axons spread to more medial areas in the bundle. Tripartite compartmentalization was not evident in the OSN-specific Sema3A knockout. Thus, Sema3A derived from OSN axons is required for pre-target axon sorting in the bundle.

Pre-target axon sorting affects the topographic map formation in the OB. We asked whether perturbation of axon sorting mediated by OSN-derived Sema3A affects glomerular map formation in the OB. In wild-type mice, I7 axons that coexpress gap-CFP (cyan fluorescent protein) were sorted to the lateral-peripheral compartment (DII-P) within the bundle. In the conditional Sema3A knockout, however, I7 axons were found in the central area where the DII-A axons are normally found (Fig. 4B and fig. S10B). When I7 glomeruli were analyzed in the OB, anterior shifts were observed in the conditional knockout (Fig. 4B and fig. S11A), consistent with the shift of I7 axons in the bundle. We also analyzed Nrp1-negative DII-A axons in the conditional knockout (fig. S12). These axons were confined to the central area in the wild-type bundle. However, they were scattered to both the central and medial areas in the conditional knockout. As a result, glomeruli were found not only in the anterior but also in the dorsal OB in the conditional knockout. Thus, positional changes within the axon bundle correlate well with positional changes of glomeruli in the OB.

Discussion. We found that pre-target axon sorting plays an important role in the organization of the topographic map. Nrp1 and its repulsive ligand Sema3A are both expressed in OSNs and are involved in axon sorting before targeting on the OB. Within the axon bundles of D-zone OSNs, DII-A axons (Nrp1 low, Sema3A high) are sorted to the central compartment of the bundle, whereas DII-P axons (Nrp1 high, Sema3A low) are sorted to the lateral-peripheral compartment. This sorting appears to occur, at least in part, by the repulsive interaction between Sema3A and Nrp1. In addition to the repulsive interactions, Sema3A and Nrp1 signals may induce homophilic adhesion of axons with Nrp1 itself (29) or with other molecules such as L1 (30). Furthermore, additional guidance receptors such as Plexin-A1 (16) may be involved in the sorting of DII-A and DII-P axons (fig. S7). We assume that similar mechanisms are also at work in the sorting of DI and DII axons (Fig. 1B).

Once OSN axons are sorted in the bundle, they need to be oriented along the correct axis before projecting onto a topographic map on the OB. This probably requires positional cues that are derived from the target or that are found along the pathway between the olfactory epithelium and the OB. In the Sema3A total knockout, Nrp1-positive DII-P axons spread rather uniformly across diameter of the bundle (Fig. 5B) and consequently mistarget to the anterior region in the OB (23, 24). The effect is different in the OSN-specific Sema3A knockout, where DII-P axons at least gravitate toward the lateral region in the bundle (Fig. 5B). Thus, Sema3A expressed by cells outside of the bundle likely functions as an additional guidance cue to orient the sorted axons along the correct axis for projection onto OB. In early embryos, but not in postnatal mice, Sema3A is expressed in the anterior OB (Fig. 5A). Furthermore, Sema3A is found in ensheathing glial cells along the medial side of the axon bundles (Fig. 5A) (23). Involvement of such intermediate cues has been reported for the thalamocortical projection (31, 32).

Fig. 5

Sema3A expressed by the non-OSN cells in the mouse olfactory system. (A) Sema3A mRNA is detected not only in OSNs, but also in the OB and ensheathing glia (EG) that surround axon bundles (age, E15). Sema3A expression is high on the medial side (Nrp1-low) of the axon bundle encircled by dotted line (left). In the OB, Sema3A is expressed in the anterior domain (encircled by dotted line) during early embryonic stages (right). Scale bars, 100 μm. The mouse olfactory system (lateral view) is schematically shown at the top. (B) Coronal sections of axon bundles stained with antibodies to Nrp1 (age, P0). In the total knockout of Sema3A, Nrp1-positive axons are spread uniformly within the bundle (right). However, in the OSN-specific Sema3A knockout, Nrp1-positive axons gravitate to the lateral side of the bundle (middle). Thus, non-OSN Sema3A may help to orient the axon bundle organization. Quantification of Nrp1 immunoreactivities (along the dotted lines) is shown in fig. S10A. Scale bars, 100 μm. (C) Possible mechanisms for topographic map formation in the brain. In the mouse olfactory system, a guidance receptor, Nrp1, and its repulsive ligand, Sema3A, are expressed in a complementary manner in OSNs. We propose that pre-target axon-axon interactions regulate the sorting of heterotypic axons. Target and intermediate cues may direct correct orientations of the map.

In the Drosophila olfactory system, early-arriving OSN axons from the antenna repel late-arriving axons from the maxillary palp with Sema1a and Plexin-A, thereby segregating the two types of axons in the target (33). In the vertebrate retinotectal projection, target-derived guidance cues (e.g., ephrin-As) are thought to provide positional information. However, surgical and genetic studies indicated that projection sites for retinal ganglion cells are determined by relative, but not absolute, levels of guidance receptors, known as “axonal competition” (34, 35). Furthermore, retinal axons are presorted before the target recognition (15). Although retinal ephrin-As are thought to antagonize Eph-A receptors in cis (36), they may also function in trans to segregate heterotypic axons (3739). We propose that the axon-axon interaction is a general strategy to establish the topographic order based on the relative levels of guidance molecules expressed by axons (Fig. 5C).

Supporting Online Material

SOM Text

Figs. S1 to S12


  • * These authors contributed equally to this work.

References and Notes

  1. See supporting material on Science Online.
  2. We thank D. D. Ginty and A. L. Kolodkin for Nrp1 mutant mice; M. Taniguchi and T. Yagi for Sema3A mutant mice; R. Y. Tsien for mCherry plasmid; and our laboratory members for valuable comments. Supported by the PREST program of the Japan Science and Technology Agency (R.K.), the Global COE program (K.K.), the Mitsubishi Foundation, and a Specially Promoted Research Grant from the Ministry of Education, Culture, Sports, Science and Technology of Japan.
View Abstract

Stay Connected to Science

Navigate This Article