Cxcl12/Cxcr4 chemokine signaling is required for placode assembly and sensory axon pathfinding in the zebrafish olfactory system

Cell positioning and axon pathfinding are developmental processes crucial for the formation of neural circuits and the proper functioning of the nervous system. During early development, neuronal precursors migrate from their birthplace to their final destination along characteristic routes. Upon reaching their destination, the neuronal precursors differentiate and extend axons to the specific targets. Both processes require navigation through a complex environment. Many molecules that have been discovered initially for their functions in axon guidance turn out to also play important roles in cell migration, and vice versa (de Castro,2003; Guan and Rao,2003), suggesting that these two developmental processes use common guidance mechanisms, at least in part.

The olfactory system is an excellent model for studying both cell migration and axon guidance. Olfactory sensory neurons (OSNs) develop within the olfactory placode, a thickened ectoderm that later gives rise to the olfactory epithelium (OE). Subsequently, OSNs extend axons toward the developing olfactory bulb (OB) with exquisite precision. In zebrafish, a previous fate map study has revealed that a large cellular field located along the lateral edge of the anterior neural plate converges through cell movements to form the olfactory placode (Whitlock and Westerfield, 2000). However, the factors responsible for directing the assembly of the olfactory placode have not yet been identified.

Chemokines are small, secreted proteins that were originally identified as molecules regulating leukocyte trafficking in the immune system. Recently, it has become clear that chemokines and their receptors play prominent roles, not only in the immune response, but also in various cellular events during development of the nervous system (Tran and Miller, 2003). In particular, Cxcl12 (also known as SDF-1,stromal cell-derived factor-1) influences the guidance of both migrating neurons (Bagri et al., 2002; Zhu et al., 2002; Stumm et al., 2003; Belmadani et al., 2005; Knaut et al., 2005; Borrell and Marin, 2006) and growing axons (Xiang et al.,2002; Chalasani et al.,2003; Li et al.,2005; Lieberam et al.,2005) through its receptor Cxcr4. As the olfactory placode is one of the regions where Cxcr4 is expressed during embryogenesis(Chong et al., 2001; Tissir et al., 2004; Schwarting et al., 2006), it is conceivable that Cxcl12/Cxcr4 signaling may play a role in olfactory development. Here, we provide genetic evidence demonstrating that Cxcl12/Cxcr4 signaling mediates the assembly of olfactory placodal precursors into a compact cluster to form the olfactory placode in zebrafish. Furthermore, the loss of Cxcl12/Cxcr4 signaling results in impaired pathfinding of pioneer olfactory axons and failure of OSN axons to project to the OB. Our results indicate that chemokine signaling plays a dual role in controlling cell positioning and subsequent axon pathfinding during early development of the primary olfactory system in zebrafish.

Zebrafish, Danio rerio, were maintained and embryos were collected essentially as described (Westerfield,1995). Embryos were staged according to hours postfertilization(hpf) at 28.5°C and morphological criteria(Kimmel et al., 1995). Collected embryos were maintained in 1/3 Ringer's solution (39 mM NaCl, 0.97 mM KCl, 1.8 mM CaCl₂, 1.7 mM HEPES, pH 7.2) supplemented with 100 units/ml penicillin and 100 μg/ml streptomycin. In some cases, 0.002%phenylthiourea (PTU) was added after 12 hpf to prevent pigmentation. We confirmed that PTU treatment at a dose of 0.002% does not affect olfactory placode assembly and OSN axon projection, although the treated embryos show a slight delay of hatching, as observed in previous work(Elsalini and Rohr, 2003).

A transgenic line, Tg(OMP^2k:gap-YFP)^rw032a(abbreviated as omp:yfp), in which membrane-tethered YFP is expressed under the control of the olfactory marker protein (OMP) promoter, were used to visualize olfactory neurons, including unipolar pioneer neurons and ciliated OSNs (Miyasaka et al., 2005; Sato et al., 2005). odysseus (ody) homozygous adults were crossed with heterozygous adults to obtain ∼50% ody/ody embryos and∼50% ody/+ embryos for synchronized development. ody/ody embryos were identified by the impaired migration of germ cells (Doitsidou et al.,2002; Knaut et al.,2003) and neuromasts (David et al., 2002), as assessed by immunofluorescence staining for Vasa (a marker for germ cells) and PCAM (a marker for neuromasts and OSNs; also known as Ncam3 - ZFIN), respectively. In some cases, ody mutant embryos were genotyped by PCR amplification of DNA fragments containing the ody mutant locus, followed by direct sequencing of the PCR products. Primers used are as follows: ody-f, 5′-TGGAGTTTGGCTTCCAGCGAC-3′;ody-r, 5′-CAGCATAGTCAAAGCGTCCAC-3′. A transgenic line, hsp:cxcl12a(previously called heatshock-SDF-1a)(Knaut et al., 2005) was used for ubiquitous heat-induced misexpression of Cxcl12a.

cDNA fragments for probes against cxcl12a, cxcl12b, cxcr4a, cxcr4b,cxcr7a and cxcr7b were amplified by PCR from cDNA libraries of zebrafish embryos or adult heads. Zebrafish cxcr7 orthologs to the human and mouse Cxcr7 gene were identified by TBLASTN searches using the sixth draft zebrafish genome assembly (Zv6: Sanger Center). The sequences used as probes against cxcr7a and cxcr7b were deposited in GenBank with the accession numbers EF467374 and EF467375, respectively. Digoxigenin (DIG)-labeled cRNA probes were generated by in vitro transcription from their cDNA clones. Whole-mount in situ hybridization was performed as previously described (Miyasaka et al.,2005).

Whole-mount immunohistochemistry was carried out as previously described(Miyasaka et al., 2005), with the following modifications. Embryos were fixed overnight at 4°C in 4%paraformaldehyde in PBS and permeabilized in acetone for 7 minutes at-20°C.

Antibodies used are as follows: rabbit polyclonal anti-GFP antibody(1:1000, a kind gift from Dr N. Tamamaki, Kumamoto University, Kumamoto,Japan); rat monoclonal anti-GFP antibody (1:1000, Nacalai Tesque); mouse monoclonal zns-2 antibody (1:20, supernatant, Developmental Studies Hybridoma Bank); rabbit polyclonal anti-PCAM antibody (rabbit IgG; 0.4 μg/ml)(Mizuno et al., 2001; Miyasaka et al., 2005); rabbit polyclonal anti-Vasa antibody (1:2000)(Knaut et al., 2000); rabbit polyclonal anti-DsRed antibody (1:300, Clontech); Cy3-conjugated donkey anti-mouse IgG antibody (1:300, Jackson ImmunoResearch); Cy3-conjugated goat anti-rabbit IgG antibody (1:300, Jackson ImmunoResearch); Alexa 488-conjugated goat anti-rabbit IgG antibody (1:300, Molecular Probes); Alexa 488-conjugated goat anti-rat IgG antibody (1:300, Molecular Probes); peroxidase-conjugated goat anti-rabbit IgG Fab antibody fragment [Histofine Simple Stain Max PO(R),Nichirei Bioscience].

Antisense morpholino oligonucleotides (MOs: Gene Tools, LLC) were dissolved at a concentration of 4 ng/nl in 1× Danieau buffer [58 mM NaCl, 0.7 mM KCl, 0.4 mM MgSO₄, 0.6 mM Ca(NO₃)₂, 5 mM HEPES, pH 7.6] containing 0.1% Phenol Red, and injected into one-cell stage embryos carrying the omp:yfp transgene at a volume of 1 nl with an air pressure microinjector (FemtoJet, Eppendorf). The volume of MO solution to be injected was calibrated by measuring the diameter of a droplet at the tip of glass needle in the air. MO sequences are as follows: cxcr4b MO,5′-AAA TGA TGC TAT CGT AAA ATT CCA T-3′(Doitsidou et al., 2002); cxcl12a MO, 5′-ACT TTG AGA TCC ATG TTT GCA GTG-3′(Li et al., 2005); cxcl12b MO, 5′-CGC TAC TAC TTT GCT ATC CAT GCC A-3′(Knaut et al., 2003).

For ubiquitous misexpression, hsp:cxcl12a/+ fish were crossed with omp:yfp/+ fish. Embryos obtained from such a cross were heat shocked at either 12 hpf or 19 hpf for 1 hour in a 37°C water bath, raised at 28.5°C, and imaged at 26-29 hpf and 3.5 days postfertilization (dpf) using a confocal laser-scanning microscope (Olympus FV500). To identify hsp:cxcl12a transgenic embryos, genomic DNAs were extracted from the embryos after image acquisition, and genotyped by PCR using the following primers: 5′-CAT GTG GAC TGC CTA TGT TCA TC-3′; 5′-ATC AGA GCG ACT ACT ACG ATC AC-3′.

For mosaic misexpression, omp:yfp/+ embryos were injected at the one- to two-cell stage with a 50 μg/ml solution of phsp:mDsRed or phsp:cxcl12a-mDsRed DNA, and heat shocked at 12 hpf and again at 19 hpf. The second heat-shock treatment was given to facilitate identification of cells expressing mDsRed or Cxcl12a-mDsRed. The embryos were fixed at 24 hpf and analyzed by immunohistochemistry using anti-GFP and anti-DsRed antibodies. The DNA constructs were prepared as follows. To generate phsp:mDsRed, EGFP cDNA of the pHSP70/4-EGFP(Halloran et al., 2000) was removed by AgeI and NotI digestion and replaced with DsRed-Monomer cDNA from the pDsRed-Monomer (Clontech). To generate phsp:cxcl12a-mDsRed, the stop codon of cxcl12a was removed by PCR and the cxcl12a gene was inserted between the SalI and AgeI sites of the phsp:mDsRed so that the DsRed-Monomer was in frame with the last codon of the cxcl12a gene.

Embryos younger than 2 dpf were anesthetized in 0.016% tricaine and mounted in 1.5% low-melting point agarose. Embryos older than 2.5 dpf were anesthetized, inserted into a silicone tube (1 mm inner diameter, 2.5 mm outer diameter, 5 mm length) to allow the embryos to stand, and mounted anterior side down on a glass-bottomed petri dish with a small amount of 1/3 Ringer's solution. The embryos were imaged at each time point using a confocal microscope. The embryos were removed from the agarose or silicon tube and kept at 28.5°C between the time points for observations.

A previous study has shown that mRNA for Cxcr4b is prominently expressed in the olfactory placode at 24 hpf (Chong et al., 2001), suggesting that this receptor might play a role in the development of the olfactory system. To corroborate this possibility, we analyzed the spatiotemporal expression of cxcr4b in the head region in more detail. At 12 hpf, cxcr4b was expressed in a row of cells along the anterolateral edge of the neural plate(Fig. 1A) that corresponds to the presumptive olfactory placodal field(Whitlock and Westerfield,2000). During the next few hours, cxcr4b-expressing cells increased in number and arranged at the edge of the developing anterior neural tube, forming bilateral bands along the anteroposterior axis(Fig. 1C,E). By 20 hpf, most of the cxcr4b-expressing cells had assembled into the olfactory placode(Fig. 1G,I). From 24 to 30 hpf,when the initial olfactory axons were extending toward the developing OB, cxcr4b expression in the olfactory placode was still evident(Fig. 1K,M,O), but cells adjacent to the apical surface, where mature OSNs reside, were devoid of cxcr4b expression (Fig. 1O). From 36 hpf onward, cxcr4b expression in the olfactory placode was greatly diminished, and only a few cells expressing cxcr4b remained in the basal part of the OE at 48 hpf(Fig. 1Q,S).

In the zebrafish genome, there are two genes (cxcl12a and cxcl12b) that encode potential ligands for Cxcr4b. Analyzing the expression patterns of the two Cxcr4b ligands, we found that only cxcl12a (right panels in Fig. 1), but not cxcl12b (see Fig. S1D-F in the supplementary material), was present in close proximity to the cxcr4b-expressing olfactory placode. The developing anterior neural tube was located between the two cxcr4b-expressing stripes and started to express cxcl12aby 14 hpf (Fig. 1D). Over the next ten hours, the cxcl12a expression domain gradually became restricted to the anterior tip of the telencephalon(Fig. 1F,H,J,L) and the two cxcr4b-expressing stripes closely followed this refining cxcl12a expression and, by convergence movements, formed the two olfactory placodes (Fig. 1E,G,I,K). At 24 to 36 hpf, in addition to the telencephalic expression, cxcl12a was expressed by cells residing at the basal margin of the olfactory placode, on the border between the placode and the telencephalon where olfactory axons cross to enter the telencephalon(Fig. 1L,P,R). At 48 hpf, the expression of cxcl12a in the telencephalon was less intense but still visible, while the expression in the margin of the OE was barely detectable(Fig. 1T). Thus, the expression of cxcl12a in close proximity to cxcr4b-expressing cells that will form the olfactory placodes is consistent with the idea that chemokine signaling might play a role in olfactory placode assembly and OSN axon pathfinding.

To assess the role of Cxcr4b-mediated chemokine signaling in the formation of the olfactory placode, we analyzed odysseus (ody)mutants, in which a nonsense mutation in the third intracellular loop leads to a truncation of Cxcr4b and probably creates a loss of Cxcr4b function(Knaut et al., 2003). ody/ody or ody/+ embryos were examined for cxcr4b expression during the period of olfactory placode formation. At 12 and 14 hpf, the arrangement of cells expressing cxcr4b in ody/ody embryos was indistinguishable from that in ody/+ embryos (Fig. 2A,B,F,G). At 16 hpf, some cxcr4b-expressing cells in ody/ody embryos were not arranged along the lateral edge of the developing anterior neural tube and dispersed laterally or anteromedially(Fig. 2H). At 18 hpf, a substantial number of cxcr4b-expressing cells were mispositioned ventromedially (Fig. 2I). At 20 hpf, when the olfactory placode had formed, the ventromedially displaced cells were still detectable at ectopic positions, although reduced in number(Fig. 2J). Total number of cxcr4b-expressing cells in the anterior head region of ody/ody embryos did not differ from that of ody/+embryos during the period of olfactory placode assembly[ody/ody versus ody/+, 51±4 versus 51±2 cells/side (16 hpf), 85±2 versus 83±2 (18 hpf),98±5 versus 103±3 (20 hpf), n=6 for each,mean±s.e.m.], suggesting that the loss of Cxcr4b-mediated signaling causes mispositioning of olfactory placodal precursors without affecting their proliferation.

To examine whether the mispositioned cells in ody/odyembryos develop into olfactory neurons, we used the omp:yfp transgenic line in which membrane-tethered EYFP starts to be expressed at ∼20 hpf in a subpopulation of olfactory neurons, including unipolar pioneer neurons and ciliated OSNs (Miyasaka et al.,2005; Sato et al.,2005). In ody/ody embryos carrying the omp:yfp transgene, YFP-positive cells were less tightly organized than those in wild-type embryos and often dispersed ventromedially from the normal position of the olfactory placode at 24 hpf(Fig. 3A-C). Similarly,reducing the expression level of Cxcr4b by MO-mediated knockdown frequently led to a mispositioning of YFP-positive cells at 24 hpf(Fig. 3D), mimicking the phenotype observed in ody mutants. The fact that mispositioned cells express YFP under the control of the omp promoter suggests that the specification of olfactory neurons is independent of Cxcr4b signaling. To trace the subsequent fate of these ectopically located olfactory neurons, we analyzed ody mutants until 3.5 dpf. The ectopically positioned YFP-positive olfactory neurons observed at 1 dpf were rapidly reduced in number during the second day of development, and they had completely disappeared by 3.5 dpf (Fig. 9E-H).

As Cxcl12a, a ligand for Cxcr4b, is expressed in the adjacent neural tube located medially to the cxcr4b-expressing olfactory placodal precursors, we asked whether Cxcl12a is also required for correct olfactory neuron positioning. To address this issue, we reduced the activity of Cxcl12a by MO injection and analyzed the position of YFP-positive olfactory neurons at 24 hpf. cxcl12a MO-injected embryos (morphants) frequently displayed ventromedially mispositioned olfactory neurons(Fig. 3E), as observed in ody mutants (Fig. 3B,C) and cxcr4b morphants(Fig. 3D). Consistent with the absence of expression in the vicinity of the developing olfactory system,reducing the activity of Cxcl12b, another ligand for Cxcr4b, did not perturb olfactory neuron positioning (Fig. 3F).

Due to phenotypic variability in embryos with compromised cxcr4b-mediated signaling (Fig. 3), we quantitated to what degree olfactory neurons are mispositioned at 24 hpf by measuring the distances from the midline to the medialmost and lateralmost cells (medial and lateral limits), and the distances from the ventral edge of the head to the ventralmost and dorsalmost cells (ventral and dorsal limits) (Fig. 4). In wild-type embryos, virtually no olfactory placode failed to assemble, although two out of 30 placodes showed a ventromedial displacement of a single olfactory neuron (Fig. 4A,G). In ody/+ embryos(Fig. 4B,H) and cxcl12b morphants (Fig. 4F,L), essentially no olfactory placode displayed mispositioning of olfactory neurons. By contrast, a large proportion of placodes in ody/ody embryos (Fig. 4C,I), cxcr4b morphants(Fig. 4D,J) and cxcl12a morphants (Fig. 4E,K) exhibited ventromedially displaced olfactory neurons; the averages of ventral limit (ody/ody, 47±4 μm, n=34; cxcr4b MO, 55±4 μm, n=30; cxcl12a MO, 40±5 μm, n=20; mean±s.e.m.) and medial limit (ody/ody, 40±3 μm, n=34; cxcr4b MO, 40±3 μm, n=30; cxcl12a MO,29±4 μm, n=20) were significantly shifted ventrally and medially, respectively, when compared with wild-type placodes (ventral limit,68±3 μm; medial limit, 51±2 μm; n=30). In ody/ody embryos, the average of lateral limit was slightly shifted laterally when compared with wild type (ody/ody,97±2 μm, n=34; wild type, 90±1 μm, n=30), but neither cxcr4b nor cxcl12a morphants showed such a shift. Thus, ventromedial displacement of olfactory neurons is a common phenotype when reducing or removing Cxcl12a/Cxcr4b signaling. These results indicate that Cxcl12a signaling through Cxcr4b is required for assembly of the olfactory neuron precursors into a compact cluster.

The ventromedial mispositioning in cxcl12a morphants appeared more severe than that in either ody/ody embryos or cxcr4b morphants (Fig. 4C-E,I-K). This observation raises a possibility that olfactory neuron precursors may express a second receptor for Cxcl12a. To address this issue, we analyzed the expression of other Cxcl12a candidate receptors. We found that cxcr4a, the other cxcr4 homolog, was not expressed in the developing olfactory placode (see Fig. S1G-I in the supplementary material). Recent studies in mammals have shown that Cxcl12 also binds to and signals through an alternate chemokine receptor, Cxcr7 (also known as Rdc1 and Cmkor1) (Balabanian et al., 2005; Burns et al.,2006). We searched the zebrafish genome database and identified two genes (cxcr7a and cxcr7b) equally homologous to the mammalian cxcr7. In situ hybridization analyses revealed that neither cxcr7a nor cxcr7b was expressed in the developing olfactory placode (see Fig. S1M-R in the supplementary material). These results indicate that Cxcr4b is the only known Cxcl12a receptor expressed in the olfactory neuron precursors, suggesting that olfactory neuron precursors can sense Cxcl12a only through Cxcr4b during placode assembly.

The Cxcl12a expression domain at the anterior neural tube becomes restricted to the anterior tip of the telencephalon simultaneously with the assembly of the olfactory placode (Fig. 1), suggesting that Cxcl12a might act as a directional cue to guide migration of olfactory placodal precursors. To test this possibility, we first sought to perturb the local source of Cxcl12a by raising the Cxcl12a levels throughout the embryo. We heat shocked hsp:cxcl12a;omp:yfp double-transgenic embryos before (12 hpf) and after (19 hpf) olfactory placode assembly, and analyzed the positioning of YFP-positive olfactory neurons at 26-29 hpf. Embryos in which Cxcl12a was heat induced at 12 hpf (six of 11 embryos) showed impaired assembly of olfactory neurons(Fig. 5B), which resembled the Cxcl12a/Cxcr4b morphant and ody mutant phenotypes(Fig. 3). Heat induction of Cxcl12a at 19 hpf (n=12) did not affect the placode assembly(Fig. 5C). Similarly, heat shocking omp:yfp single transgenic embryos did not cause any defects in olfactory neuron assembly (Fig. 5A). These results suggest that a spatially and temporally localized Cxcl12a source is crucial for correct positioning of olfactory neurons.

Next, to determine whether Cxcl12a attracts migrating olfactory neuron precursors, we analyzed the effects of localized Cxcl12a misexpression. We injected omp:yfp transgenic embryos with a DNA construct containing a monomeric DsRed (mDsRed)-tagged Cxcl12a under the control of hsp70promoter, creating ectopic sources of Cxcl12a after heat shock. In embryos that showed mosaic Cxcl12a-mDsRed expression in the anterior head, we frequently observed abnormal appearance of YFP-labeled olfactory neurons ventral to the correct position (n=14 of 16 embryos). In such cases,some olfactory neurons were found in close proximity to the ventral ectopic Cxcl12a sources (Fig. 6B-D). By contrast, we never found dorsally displaced olfactory neurons, despite the presence of ectopic sources in the dorsal region (open arrowheads in Fig. 6B). In control embryos with mosaic mDsRed expression in the anterior head (n=15), olfactory placode assembly was never perturbed (Fig. 6A). These results indicate that ectopically expressed Cxcl12a in the ventral region can compete with the endogenous Cxcl12a and recruit olfactory neurons ventrally.

As cxcr4b expression persists in the olfactory placode at the initial phase of olfactory axon pathfinding, and cxcl12a is expressed along the route of olfactory axon outgrowth(Fig. 1), we investigated whether the loss of Cxcl12a/Cxcr4b signaling affects OSN axon projection to the OB. OSNs are composed of two major types of neurons, ciliated and microvillous OSNs, which express different types of odorant receptors and display distinct glomerular innervation in a mutually exclusive manner(Sato et al., 2005). Using the PCAM antibody, a marker that labels axons from both ciliated and microvillous OSNs (Miyasaka et al., 2005),we examined axon trajectories of OSNs in ody mutants at 3 dpf, when prototypic glomerular targeting by OSN axons has been established. In ody mutants, we frequently observed an impaired OSN axon projection(Fig. 7). Due to phenotypic variability, we divided ody mutants into three classes: 28% of the ody mutants (n=18 embryos) showed bilateral OSN axon projection defects (class 3); 28% showed projection defects only on one side(class 2); and 44% showed normal axon projection to the OB (class 1). In ody mutants with projection defects, OSN axons failed to exit the OE or occasionally stalled immediately after exiting the OE, accumulating near the OE-telencephalon border (Fig. 7C,D). Intriguingly, embryos with projection defects on one side showed no detectable proto-glomerular innervation defects on the contralateral side (Fig. 7C). OSN axon projection defects were also seen when a subpopulation of OSN axons were labeled by membrane-tethered YFP in ody mutants carrying the omp:yfp transgene (Fig. 9H). Reducing the activity of Cxcr4b by MO injections yielded similar defects in OSN axon projection, although the incidence of the defects was lower when compared with that of ody mutants (4 and 41% of OE,respectively; Table 1). Such a discrepancy in phenotypic incidence between mutants and morphants was not observed in the defects of the olfactory placode assembly (see Fig. 4). cxcl12amorphants also exhibited similar defects in OSN axon projection at a comparable incidence with cxcr4b morphants (3 and 4% of OE,respectively; Table 1), whereas cxcl12b morphants never displayed OSN axon projection defects(n=134 OE). The lower incidence of defects in cxcl12a/cxcr4b morphants than ody mutants that lack Cxcr4b function is probably explained by an incomplete removal of Cxcl12a/Cxcr4b signaling after MO injections. These results indicate that Cxcl12a/Cxcr4b signaling is important to allow OSN axons to exit the OE and to extend toward the OB, and suggest that the OSN axon projection would be ensured by less activity of Cxcl12a/Cxcr4b than that required for the olfactory placode assembly.

A previous study has demonstrated that the earliest axons growing to the presumptive OB originate from a transient population of neurons in the olfactory placode: the unipolar pioneer neurons. Ablation of these unipolar neurons perturbs pathfinding of axons from later developing OSNs(Whitlock and Westerfield,1998), suggesting that the axons of unipolar neurons may provide a track for OSN axons to follow. Therefore, it is possible that Cxcr4b-mediated signaling helps axons of unipolar neurons to pioneer to the presumptive OB. To explore this possibility, we investigated the trajectory of pioneer axons of 1.5 dpf ody mutant embryos expressing YFP in both unipolar neurons and ciliated OSNs. Axons of unipolar neurons were identified by co-labeling with zns-2 antibody (Whitlock and Westerfield, 1998). We found that most of the YFP-positive axons emerging from the olfactory placode at 1.5 dpf were also immunoreactive for zns-2, and that projection of double-labeled pioneer axons were frequently impaired in ody mutant embryos(Fig. 8). The defects in pioneer axon projection from ody mutant olfactory placodes did not occur in an all-or-none manner, differently from PCAM-positive OSN axon projection at 3 dpf (see Fig. 7). In 19% of the ody mutant placodes, all pioneer axons failed to exit the placode and accumulated near the placode-telencephalon border (n=16 placodes; Fig. 8A-C); in 19% some pioneer axons reached the presumptive OB and the remaining axons failed to exit the placode(Fig. 8D-F), and in 62% most pioneer axons showed a normal projection to the presumptive OB(Fig. 8G-I). ody/+(see Fig. 9B) and wild-type(Fig. 8J-L) embryos never exhibited such pioneer axon projection defects.

To assess whether the variable projection patterns of pioneer axons from unipolar neurons are associated with projection patterns of following OSN axons, we carried out time-lapse observation of the YFP-expressing axons from both unipolar neurons and ciliated OSNs until 3.5 dpf. We found that about 44%of the ody mutant olfactory placodes (n=16) extended a smaller number of pioneer axons toward the presumptive OB (arrow in Fig. 9F) than did ody/+ and wild-type olfactory placodes at 1.5 dpf. Whenever we observed the pioneer axons clearly not reaching the presumptive OB at 1.5 dpf,we later found that the following axons of ciliated OSNs from the same placode also failed to grow out (Fig. 9G,H; left side) and eventually accumulated at the border between the OE and the telencephalon by 3.5 dpf (arrowheads in Fig. 9H). By contrast, if a substantial number of the pioneer axons reached the presumptive OB by 1.5 dpf,we never detected impaired axon projection of ciliated OSNs at 3.5 dpf in ody mutant embryos (Fig. 9F-H; right side). In ody/+(Fig. 9A-D) and wild-type (data not shown) embryos, the pioneer axons always successfully navigated to the presumptive OB, and ciliated OSNs also established correct projection to the OB. These results suggest that Cxcr4b-mediated signaling is important for pathfinding of zns-2-positive pioneer axons from unipolar neurons to exit the olfactory placode and that impaired pathfinding of the pioneer axons might be the reason why OSN axons subsequently fail to project to the OB in the absence of Cxcr4b function.

The restricted expression of Cxcl12a along the path of olfactory axon navigation suggests that Cxcl12a might guide growing olfactory axons to the OB. To examine whether Cxcl12a plays a role as an attractive cue or a permissive substrate for olfactory axons, we heat shocked hsp:cxcl12a;omp:yfp double-transgenic embryos at 12 and 19 hpf, just before the olfactory placode assembly and just before the emergence of pioneer axons, respectively, and analyzed axon projection at 3.5 dpf. As described above, we found that ubiquitous expression of Cxcl12a at 12, but not 19, hpf perturbed the olfactory placode assembly (Fig. 5B,C). However, neither at 12 nor 19 hpf did ubiquitous expression of Cxcl12a induce significant defects in olfactory axon projection at 3.5 dpf(Fig. 5E,F). These results suggest that Cxcl12a does not guide olfactory axons as an attractive cue but rather creates a permissive environment for navigating olfactory axons. Importantly, these results also demonstrate that axon guidance defects in the absence of Cxcr4b-mediated chemokine signaling cannot be attributed solely to the mispositioning of olfactory neuron cell bodies.

Olfactory placode assembly and subsequent OSN axon projection are two crucial aspects in the formation of functional neural circuitry of the primary olfactory system. Here, we found that these events depend on Cxcl12a chemokine signaling through its receptor Cxcr4b in zebrafish. Both loss and gain of function for the Cxcl12a/Cxcr4b signaling perturbed placode assembly of olfactory neuron precursors into a compact cluster. Furthermore, loss of function frequently resulted in a defect in pathfinding of pioneer axons,leading to a complete loss of OSN axon projection to the OB. These findings reveal a dual role of chemokine signaling in controlling cell migration and axon pathfinding in the zebrafish olfactory system.

Olfactory neuron precursors initially lie in the olfactory placodal field,the region stretching along the lateral edge of the anterior neural plate(Whitlock and Westerfield,2000; Whitlock,2004). As development progresses, they converge to form the olfactory placode. The region expressing cxcr4b in the anterior head corresponds to the olfactory placodal field. Previous studies have demonstrated that Cxcr4b-mediated signaling guides migration of various types of cells in different ways in zebrafish, including primordial germ cells(Doitsidou et al., 2002; Knaut et al., 2003), posterior lateral line primordia (David et al.,2002) and trigeminal sensory neurons(Knaut et al., 2005). The cxcr4b-expressing germ cells and posterior lateral line primordia move along their migratory route, maintaining direct contacts with the cxcl12a expression domains. By contrast, most cxcr4b-expressing trigeminal sensory neurons are initially located at a distance from the cxcl12a expression domains, and then move toward the Cxcl12a sources. These observations suggest that Cxcl12/Cxcr4 signaling may act in either short or long range, depending on the situations of individual cell types. In the case of olfactory neuron precursors, they are closely adjacent to the cxcl12a expression domain until the olfactory placode is formed, suggesting a short-range action of Cxcl12a in this process.

In the absence of Cxcr4b-mediated signaling, olfactory neuron precursors dispersed anteriorly, resulting in ventrally displaced olfactory neurons. Hence, chemokine signaling is required for correct positioning of olfactory neurons. Moreover, global and localized misexpression of Cxcl12a perturbed olfactory placode assembly, suggesting that Cxcl12a/Cxcr4b signaling is not merely permissive, but acts as a guidance cue to position the olfactory neurons. This notion is consistent with the observation that spatial restriction of the cxcl12a expression domain occurs synchronously with convergence of the olfactory placodal field. There is a significant anterior tissue movement during early segmental stages(Karlstrom and Kane, 1996),and this movement should result in ventral cell streaming at the most anterior part of the developing head. The appearance of ventrally displaced olfactory neurons in the absence of Cxcr4b signaling raises the possibility that Cxcl12a provides olfactory neuron precursors with a retention signal to withstand the anterior/ventral movement of neighboring cells. Our observation that ectopic Cxcl12a sources in the ventral region, but not in the dorsal region, can recruit olfactory neurons to positions close to the ectopic Cxcl12a sources could also be explained by the competition between the endogenous Cxcl12a source and the anterior/ventral streaming of cells expressing exogenous Cxcl12a. Such a mechanism, by which Cxcl12 might act to retain cxcr4b-expressing cells in the correct position, has been implicated in trigeminal sensory ganglion assembly(Knaut et al., 2005). In this case, cxcl12 expression domains are located posteriorly adjacent to the final position of ganglia, opposite to the direction of morphogenetic movements. By contrast, the cxcl12a expression domain in the anterior head was positioned medially to the array of olfactory neuron precursors,although a similar mispositioning phenotype was observed in both olfactory neurons and trigeminal sensory neurons upon disruption of Cxcr4b signaling. Direct in vivo observations on dynamic behaviors of olfactory neuron precursors and neighboring cells in wild-type and ody mutant embryos may clarify the mode of Cxcl12a action in more detail.

In ody mutants, ∼50% of the embryos exhibited defects in OSN axon projection to the OB. The simplest interpretation is that Cxcl12a may act as a chemoattractant for olfactory axons, because cxcl12a is expressed in the placode-telencephalon border where the axons cross to enter the telencephalon and also in the anterior tip of the telencephalon toward which the axons grow. However, Cxcl12 does not appear to exhibit obvious attractive activity when assayed on several classes of vertebrate neurons in vitro (Chalasani et al., 2003). Secondly, axon guidance defects might be an indirect consequence of loss of Cxcr4b signaling. In the absence of Cxcr4b signaling, olfactory neurons are mispositioned and outgrowing axons are challenged with a new micro-environment, and failure to interpret this new environment correctly may cause axons to navigate to new targets. However, such an indirect effect of loss of Cxcr4b signaling on axon guidance is unlikely, because ubiquitous misexpression of Cxcl12a at 12 hpf, which induces mispositioning of olfactory neurons, did not cause the ody-like OSN axon pathfinding defects(Fig. 5B,E). As Cxcl12 has been shown to reduce the growth cone responsiveness to chemorepellents in vitro(Chalasani et al., 2003) and in vivo (Chalasani et al., 2007),the third possibility is that exposure of olfactory axons to Cxcl12a at the placode-telencephalon border might reduce the growth cone responsiveness to chemorepellent factors that exist around the path of olfactory axon navigation, allowing them to exit the olfactory placode and to grow toward the OB. This permissive model is consistent with our observation that forced ubiquitous expression of Cxcl12a at either 12 or 19 hpf induces no significant defect in OSN axon projection to the OB(Fig. 5). A previous study(Yoshida et al., 2002) has demonstrated that alteration of protein kinase A (PKA) activity in olfactory axons is important for pathfinding and has proposed that the interaction of growing axons with the placode-telencephalon border may downregulate PKA signaling, modulating the response of growth cone to guidance cues. Cxcl12a hence appears to be a promising candidate to fulfill this function via Cxcr4b,which signals through Gi proteins, thereby decreasing the intracellular cyclic AMP level (Tran and Miller,2003).

The defects in OSN axon projection in ody mutants are variable and appear to occur in an all-or-none manner. Once OSN axons can reach the OB, the pattern of proto-glomerular targeting by OSNs in ody mutants is indistinguishable from that in wild type. This all-or-none phenotype could be explained by the notion that pioneer axons provide a scaffold essential for subsequently projecting OSN axons(Whitlock and Westerfield,1998; Miyasaka et al.,2005). Indeed, the pathfinding by pioneer axons was frequently impaired in ody mutants. In some cases, a portion of zns-2-positive pioneer axons failed to exit the olfactory placode, but the remaining ones from the same placode could reach the OB at 1.5 dpf. Such defects with an intermediate severity are rarely observed in OSN axon projection at 3.5 dpf. These findings imply that there might be a threshold in the number of OB-reaching pioneer axons that is required for establishment of a sound projection of OSN axons. Our observation that mature OSNs did not express cxcr4b also suggests that Cxcr4b-mediated signaling influences the early growing axons from either unipolar neurons or newly differentiated immature OSNs.

Cxcl12/Cxcr4 signaling has recently been shown to regulate retinal axon projection in zebrafish (Li et al.,2005) and motor axon projection in mouse(Lieberam et al., 2005). In cxcr4b mutant zebrafish, some ganglion cell axons fail to exit the eye but extend in aberrant directions within the retina. Similarly, in the absence of Cxcr4 signaling in mice, some axons of a set of motoneurons, termed vMNs, aberrantly exit through the dorsal exit point of the spinal cord instead of correctly choosing the ventral exit point. In both cases, the initial phase of axon pathfinding is perturbed similarly to the case for olfactory axons deprived of Cxcr4b function. These findings suggest that Cxcl12/Cxcr4 signaling might be used as a general molecular tool to create a favorable environment and to shape an initial trajectory for various types of early growing axons in vertebrates.

We thank members of the Yoshihara laboratory for fish care and discussion,Nobuaki Tamamaki (Kumamoto University) for rabbit ant-GFP antibody and John Y. Kuwada (University of Michigan) for pHSP70/4-EGFP plasmid. H.K. thanks Alexander F. Schier (Harvard University) for support. This work was supported in part by a Grant-in-Aid for Scientific Research (B) to Y.Y. from the Ministry of Education, Culture, Sports, Science, and Technology of Japan. H.K. was supported by a long-term fellowship from the Human Frontier Science Program (HFSP). The transgenic fish line,Tg(OMP^2k:gap-YFP)^rw032a, can be obtained upon request under the support of Zebrafish National Bioresource Project of Japan.