Differential function of RNCAM isoforms in precise target selection of olfactory sensory neurons

Mouse olfactory sensory neurons (OSNs) are functionally specified to express one specific odorant receptor (OR) out of more than 1000 different types of receptor (Buck and Axel,1991). In the olfactory bulb (OB) of the brain, axons of OSNs expressing different ORs sort out and project with precision to defined axon convergence sites (the glomeruli) topographically(Mombaerts et al., 1996;Ressler et al., 1994;Wang et al., 1998;Vassar et al., 1994). OSN projections between the olfactory epithelium (OE) and the OB are organized in at least two additional ways. First, axons of OSNs expressing the same OR converge to glomeruli located on the medial and the lateral side of the OB,respectively. This organization gives rise to two mirror-image sensory maps within each bulb (Nagao et al.,2000). Second, OE is divided into at least four OR expression zones, each with scattered OSNs that are specified to express one OR gene from a defined zonal set of OR genes (Ressler et al., 1993; Vassar et al.,1993). Evidence that axons of OSNs located in a particular OR zone terminate in a defined zone in the OB has come from the identification of the Rb8 neural cell adhesion molecule, RNCAM (NCAM2 — Mouse Genome Informatics) (Alenius and Bohm,1997; Yoshihara et al.,1997). RNCAM is not expressed in OR zone 1 (Z1), while neurons located in OR zones 2, 3 and 4 [Z2-4, according to zone nomenclature of Ressler et al. (Ressler et al.,1993)] express RNCAM. The OB is thus divided into two main regions innervated by RNCAM-negative (Z1) and RNCAM-positive (Z2-4) axons. Owing to the capacity to mediate homophilic cell adhesion, it has been hypothesized that RNCAM regulates an initial level of selective guidance, by grouping Z2-4 axons together (Alenius and Bohm,1997; Lin and Ngai,1999; Yoshihara et al.,1997). Immunoreactivity corresponding to RNCAM disappears from axons after odor stimulation (Yoshihara et al., 1993). Thus, the function of RNCAM in Z2-4 appears to be regulated by neural activity. RNCAM is related to NCAM, DrosophilaFasciclin 2 (Fas2) and Aplysia CAM (apCAM) in terms of sequence similarity and domain structure. Alternative splicing generates isoforms with different attachments to the cell surface [i.e. by a glycosyl-phosphatidylinositol (Gpi)anchor or a transmembrane (Tm) domain]. Although subcellular localization,signaling and activity regulation are known to differ between Gpi and Tm isoforms of NCAM-related proteins (Walsh and Doherty, 1997), little is known about their combined activities within a given cell type.

To study the importance of RNCAM isoforms in olfactory map formation, we have used the olfactory marker protein (OMP) gene promoter to generate transgenic mice with ectopic (in Z1) and elevated (in Z2-4) expression of either GpiRNCAM or TmRNCAM in OSNs. Our identification of NADPH diaphorase(NADPHd) activity as a novel marker for Z1 axons and the use a genetically modified mouse (P2-IRES-taulacZ)(Mombaerts et al., 1996) have allowed us to conclude that OMP-RNCAM transgenic mice have an intact regional division of axon projections while showing errors in OR-specific axon segregation. The topographic distribution of glomeruli with axons having more than one OR identity in mice overexpressing either of the two RNCAM isoforms alone or in combination, suggest that OSNs use a dual, isoform-specific,regulatory function of RNCAM when precisely selecting target glomeruli.

To minimize risk of generating an artificial 5′ untranslated region that could interfere with translation, we cloned the second codon of RNCAM in frame with the initiation codon of OMP. An oligonucleotide that spanned from aBamHI site at -75 to start codon of the OMP gene was ligated in-frame with a sequence containing GpiRNCAM-coding sequence and 78 bp of 3′untranslated region (UTR). A SV40 polyadenylation site was introduced between the EcoRI site in the RNCAM sequence and the SmaI site in the cloning vector (pBSK, Stratagene, La Jolla, CA). This construct was used to generate a TmRNCAM transgene construct by replacing 354 bp of sequence unique to GpiRNCAM with a fragment that contained 538 bp 3′ coding region and 268 bp 3′ UTR unique to TmRNCAM. Finally, a BamHI fragment (-6000 to -75) of mouse genomic OMP was ligated toBamHI-BglII sites within the plasmids containing GpiRNCAM and TmRNCAM cDNA, respectively. After sequencing, transgenic constructs were gel purified and microinjected into pronuclei of C57/B6/CBA one-cell embryos. Two and three founder lines were generated that expressed the Tm and Gpi isoform of RNCAM, respectively. Transgenic founder mice were backcrossed to C57/B6 and the F1 generation from crosses between RNCAM transgenic mice and P2-IRES-tau-lacZ mice were used in phenotypic characterization. The littermate control mice analyzed were therefore not genetically identical to the parental P2-IRES-taulacZ(Mombaerts et al., 1996) mouse strain which might explain subtle differences in number of P2 glomeruli.

A pBAD (Invitrogen, Carlsbad, CA) vector was used to generate a Myc/His-tagged TmRNCAM protein that corresponded to amino acid residues 442-685 in the extracellular domain. Bacterial expression (strain Top10,Invitrogen, Carlsbad, CA) was induced with 1% Arabinose (Sigma-Aldrich,Sweden) at 37°C for 1 hour. Bacteria was sonicated and tagged peptide was recovered on Ni-NTA agarose (Qiagen, Germany). Rabbits were immunized with 500μg of purified protein followed by several boosting injections with 100μg of protein (Agrisera, Sweden). Antiserum was purified using a column of Top 10 bacterial extract coupled to CNBr Sepharose (Amersham Pharmacia Biotech, Sweden) and subsequently affinity purified by binding and elution from a column of tagged RNCAM protein coupled by cyanogen bromide (BioRad,Sweden). Specificity of purified anti-RNCAM antibody was confirmed by western blot analyses using homogenates from olfactory bulb and brain.

Tissue from 2-week-old mice was fixed in 4% paraformaldehyde/PBS for 30-90 minutes at room temperature and decalcified in 0.5 M EDTA at 4°C for 3-10 days. Tissue was cryoprotected by 30% sucrose at 4°C for 24 hours,embedded in OCT compound (Sakura Finetek, Torrance, CA) and cryosectioned (16μm). Retrieval of RNCAM antigen was achieved by microwaving at 650 W for 10 minutes in 10 mM citrate buffer (pH 6.0). Sections were incubated for 1 hour at room temperature in blocking solution (3% normal goat serum, 0.1 M phosphate buffer, 2 mM MgCl₂, 5 mM EGTA, 0.02% Nonidet P-40, and 0.001% sodium deoxycholate) followed by an overnight incubation at 4°C in blocking solution containing affinity-purified anti-RNCAM antibody (1:100). Sections were subsequently washed with PBS and a positive immunoreaction was visualized by using biotinylated goat anti-rabbit IgG (1:250) (Vector laboratories, Burlingame, CA) and Cy3-coupled strepavidin (1:2000) (Jackson Immuno Research laboratories, West Grove, PE) for 1.5 hours and 30 minutes,respectively. Sections were counter stained with Hoechst 33258 (Sigma-Aldrich,Sweden) and mounted with Dako fluorescent mounting media (Dako, Sweden). Microphotographs were taken using fluorescent optics on a Zeiss Axioskop microscope (Zeiss, Sweden)

Pretreatment of sections was carried out according to Breitschopf et al.(Breitschopf et al., 1992). Cryosections (10 μm) were hybridized to ³⁵S-labelled RNCAM cRNA probes; hybridization and washing conditions were as described previously(Sassoon et al., 1988). The probe that recognized both isoforms of RNCAM corresponded to a 1.4 kb fragment coding for the extracellular RNCAM domain. Isoform-specific probes corresponded to published sequences of GpiRNCAM (bp 2204-2900, Accession Number, AF016619) and TmRNCAM (bp 312-842, Accession Number, AF016620),respectively. Slides were dehydrated and processed for autoradiography using NTB-emulsion (Kodak, Sweden), and exposed for 5-14 days at 4°C. Histological staining of sections was performed with the nuclear dye Hoechst,and analysis was carried out using both fluorescent and dark-field optics on a Zeiss Axioskop microscope.

NADPH diaphorase histochemistry was carried out essentially as previously described (Dellacorte et al.,1995). Cryosections (16-40 μm) of OE and OB were mounted on slides, washed twice with 100 mM Tris HCl (pH 8.0), incubated for 60 minutes at 37°C in 1 mM β-NADPH, 0.3% Triton X-100, 1.5 mM L-Arginine, 0.8 mM NBT 100 mM Tris-HCl (pH 8.0), rinsed in PBS and counterstained with Hoechst.β-Galactosidase histochemistry was carried out on serial coronal OB cryosections (40 μm). Slides were rinsed three times for 20 minutes in 0.1 M phosphate buffer, 2 mM MgCl₂, 5 mM EGTA, 0.02% Nonidet P-40, and 0.001% sodium deoxycholate. The staining reaction was carried out in the same buffer, supplemented with 1 mg/ml X-gal, 5 mM K₃Fe(CN)₆, and 5 mM K₄ Fe(CN)₆, overnight (16 hours) at 37°C. Sections were subsequently rinsed three times in PBS,counterstained with Hoechst and mounted with Dako farmount mounting media(Dako, Sweden). Student's t-test (two tailed) was used to determine statistical significance of all morphological data sets.

Alternative splicing leads to generation of two RNCAM isoforms with different methods of cell membrane attachment: a Tm domain and a Gpi-anchor. In order to study the function of RNCAM isoforms in olfactory map formation,we generated transgenic founder lines expressing either of the two isoforms. To ensure specific expression in OSNs we used a transgene construct containing SV40 poly-adenylation sites and a 6 kb mouse genomic fragment corresponding to the region upstream of OMP coding sequence(Fig. 1A). The rat and mouse OMP promoters have been shown to drive expression efficiently of transgenes in OSNs throughout the OE (Danciger et al.,1998; Servenius et al.,1994). Two transgenic founder lines for each RNCAM isoform were analyzed in detail. In situ hybridization analyses showed that in all four lines, RNCAM expression was evident in this zone(Fig. 1B,D). As endogenous RNCAM is not expressed in the most dorsomedial OR zone (Z1), ectopic expression of transgenic RNCAM was evident in this zone(Fig. 1B-D). Demonstration of increased total (i.e. endogenous plus transgenic) levels of RNCAM mRNA splice variants in OR zones 2-4 required comparative analyses on serial tissue sections using isoform-specific probes. Ectopic levels of RNCAM in Z1 were higher than endogenous RNCAM mRNA levels detected in Z2-4, e.g. compare endogenous TmRNCAM levels (Fig. 1E,Z2) with transgenic TmRNCAM levels(Fig. 1F, Z1) and endogenous GpiRNCAM levels (Fig. 1G, Z2)with transgenic GpiRNCAM levels (Fig. 2H,Z1). From these experiments, we concluded that transgenic animals expressed RNCAM mRNA ectopically in Z1 OSNs, whereas Z2-4 OSNs demonstrated increased levels of RNCAM mRNA.

Immature OSNs are located close to the basal cell layer of OE. A probe specific for TmRNCAM hybridized to OSNs in OE layers containing both immature and mature OSNs, whereas a GpiRNCAM-specific probe hybridized predominantly to immature OSNs (Fig. 1I,J). This result indicated that RNCAM isoforms showed a different laminar distribution in OE because of post-transcriptional regulation of mRNA steady-state levels. Surprisingly, the same distribution of RNCAM isoforms was evident in transgenic mice (Fig. 1F,H). Thus, exogenous and endogenous RNCAM transcripts showed the same relative distribution in OE, regardless of the fact that the gene regulatory sequences determining transcription were different, i.e. OMP or RNCAM promoters,respectively. As cDNA of spliced transcripts was used to construct the transgenes, the result indicated that the observed difference in distribution of RNCAM mRNA isoforms was regulated by differential mRNA stability and not differential splicing. Because both transgenes contained an identical 5′UTR, the results further suggested that the determinant regulating the differential mRNA stability was located within the 3′ UTRs that differed. The turnover of an mRNA can be regulated by cis-acting elements located in the 3′ UTR such as the AU-rich elements (AREs) that are binding sites for proteins regulating mRNA turnover(Ross, 1995). Interestingly,two AREs (AUUUA and AUUUUA) were present within the 3′ UTR unique to the GpiRNCAM construct (Fig. 1A). It has been shown that the expression of Gap43 (Basp2— Mouse Genome Informatics) mRNA is regulated post-transcriptionally via AREs and that the protein is synthesized in high amount during neurite outgrowth (Beckel-Mitchener et al.,2002). In this respect, it is interesting to note that we found GpiRNCAM mRNA to be expressed in the same cell layer of OE as Gap43mRNA (Verhaagen et al., 1989). These results suggested that the GpiRNCAM isoform might function in axons of immature OSNs, while TmRNCAM in addition might have a function in more differentiated OSNs.

In tissue sections of the OB, the GpiRNCAM-specific probe generated no hybridization signal above background, while the TmRNCAM-specific probe generated a hybridization pattern indistinguishable from that reported for a probe recognizing both isoforms (Alenius and Bohm, 1997; Yoshihara et al., 1997). Thus, TmRNCAM appears to be evenly distributed along the dorsoventral and mediolateral extensions of granular, mitral and periglomerular cell layers.

We generated a polyclonal antiserum against the common extracellular domain. To analyze for cellular distribution of RNCAM isoforms, we took advantage of the fact that Z1 OSNs lack endogenous RNCAM. As expected, control mice showed no RNCAM immunoreactivity in Z1 glomeruli(Fig. 2A). For GpiRNCAM transgenic mice, intense immunoreactivity was observed in the nerve layer,whereas glomeruli showed less staining(Fig. 2B). Thus, when GpiRNCAM was expressed at levels higher than normal, this isoform preferentially localized to axons in the nerve layer. This result indicated that the immunoreactivity in Z2-4 glomeruli of control mice corresponded to TmRNCAM and not GpiRNCAM (Fig. 2G). The finding that transgenic TmRNCAM protein showed equally strong immunoreactivity in nerve and glomerular cell layers supported this notion(Fig. 2C). The differential distribution of RNCAM isoforms may relate to the fact that Gpi-anchored proteins, in contrast to many transmembrane proteins, are clustered in sphingolipidsterol rich microdomains or rafts that influence subcellular localization (Simons and Ikonen,1997). An alternative interpretation is that the levels of expression per se determine the axonal distribution of RNCAM isoforms. The finding that endogenous expression levels of TmRnCAM was higher than that of GpiRNCAM in control mice support such a notion(Fig. 1E,G). However,regardless of the mechanism causing the differential distribution observed,the results suggested that both isoform were present on axons in the nerve layer, whereas TmRNCAM was the predominant isoform on axon terminals within glomeruli.

Our analyses addressing whether gain of RNCAM function alters the broad zonal organization of projections were hampered by the fact that no marker,except RNCAM, had been identified that distinguish this division. Previous studies have reported that NADPH diaphorase (NADPHd) activity is restricted to the dorsomedial OE and OB (Davis,1991; Dellacorte et al.,1995). In situ hybridization analysis of RNCAM and NADPHd histochemistry on serial sections of OE revealed that NADPHd activity was complementary to RNCAM expression and thus restricted to Z1(Fig. 2D,E). Importantly, no RNCAM/NADPHd double-positive glomerulus was found(Fig. 2F,G). Besides the intriguing topographical restriction of NADPHd-activity, this represented a useful marker for the majority of Z1 axons and their terminals in the OB.Fig. 3 shows comparisons of NADPHd activity in whole-mount preparations, as well as serial OB sections of transgenic mice and littermate controls. These analyses revealed that control mice (Fig. 3C,F) and transgenic mice (Fig. 3A,B,D,E) had equal surface areas and spreading of the NADPHd-positive region along dorsoventral,mediolateral and rostrocaudal axes of the OB. Moreover, no NADPHd-positive glomeruli or stray axons were present in the region of the bulb that normally is innervated by RNCAM-positive axons. This indicated that topographic expression of RNCAM does not determine the division of the glomerular field into two major regions that receive input from OSNs located in Z1 and Z2-4.

As RNCAM-mediated adhesion theoretically can interfere with a number of different axon guidance decisions, we analyzed the projections from one defined OSN subpopulation (P2) in detail. The P2 OR gene is expressed by RNCAM-positive OSNs located in Z2. Axons of the P2-positive OSN subpopulation can be visualized with the aid of the P2-IRES-tau-lacZ mouse line,which was generated by gene replacement in embryonic stem cells described by Mombaerts et al. (Mombaerts et al.,1996). This mouse line has a P2 allele that is modified to express a bicistronic transcript generating normal P2 OR protein, as well as an easily visualized reporter protein (the microtubule-associated protein tau fused toβ-galactosidase). Histochemical staining for β-galactosidase activity allows for direct visualization of P2 axons. We analyzed mice carrying one copy of the P2-IRES-tau-lacZ allele together with one RNCAM transgene allele, either OMP-TmRNCAM or OMP-GpiRNCAM. P2 axons were visualized in coronal sections of nasal tissue and OB of 2-week-old mice. Transgenic and control mice had equal total numbers of morphologically similar P2 glomeruli (Fig. 4A,B;Fig. 5A). Interestingly,examination of transgenic mice revealed increased numbers of a type of glomeruli that was only partially innervated by P2 axons(Fig. 4C,D). It has been reported that intermingling of axons corresponding to two closely related ORs occurs during early postnatal development, while misrouted axons are rarely detected in adult animals (Conzelmann et al., 2001; Mombaerts et al.,1996; Potter et al.,2001; Royal and Key,1999). Royal et al. have however, reported a sporadic occurrence of semi-innervated P2 glomeruli in which P2 axons appear to terminate in a discrete subregion of the glomerulus, either with a diffuse or dense distribution (Royal and Key,1999). The increased number of such semi-innervated P2 glomeruli in transgenic mice indicated that gain of RNCAM function augmented the frequency at which such axonal targeting errors occur. We estimated the number of this type of glomeruli to be on average 0.6 per OB in 2-week-old control mice and 1.5-2 per OB in RNCAM transgenic littermates(Fig. 5B).

Axons of P2 OSNs typically innervate one glomerulus (or a few juxtaposed glomeruli) on the medial and lateral sides of each bulb. Thus, each wild-type P2-IRES-tau-lacZ mouse has at least four fully innervated glomeruli(Mombaerts et al., 1996). In our experimental system, we detected on average five fully innervated glomeruli per control mouse. It has been speculated that the variation in the exact number of glomeruli might be a reflection of genetic background differences and/or variations in the odorous environment of animal facilities(Potter et al., 2001). It is conceivable that the occurrence of semi-innervated P2 glomeruli in control mice will vary between laboratories for the same reasons. To control for genetic background and odor environment, we performed all quantitative analyses on transgenic and control littermates. As an additional control, we analyzed on OMP promoter-driven transgenic mouse line that overexpressed a cell-surface protein (neuropilin 2) unrelated to RNCAM (data not shown). Transgenic OMP-neuropilin 2 mice and littermate controls were found to have the same number of semi-innervated P2 glomeruli 0.7±0.2 per OB (SEM,n=10) and 0.6±0.2 per OB (SEM, n=8), respectively. These results showed that the increased number of targeting errors was an effect specific to RNCAM overexpression.

OR genes are subject to the phenomenon of monoallelic expression(Chess et al., 1994). Appearance of semi-innervated P2 glomeruli in mice homozygous or heterozygous for the targeted P2 allele was similar(Fig. 4D). This excluded the possibility that the unstained fraction of axons in the semi-innervated P2 glomeruli exclusively corresponded to OSNs that express P2 from the untargeted P2 allele because of allelic inactivation. Moreover, the OE in RNCAM transgenic mice had a normal structure and normal expression of NADPHd activity, endogenous RNCAM (Fig. 1E,F,G,H), neuropilin 2 and OR genes specific for Z1 (K21), Z2(P2, K20), Z3 (L45) and Z4 (A16) (data not shown). These results taken together with the predominant localization of RNCAM protein on axons indicate that the increased number of incorrectly targeted P2 glomeruli was a consequence of aberrant axon navigation within the OB, rather than disorganization of OSNs within the OE itself.

Interestingly, we detected a three- to fourfold increase of incorrectly targeted glomeruli in both TmRNCAM and GpiRNCAM transgenic mice(Fig. 5B). A close examination of the distribution of semi-innervated P2 glomeruli revealed that they were confined to different domains of the OB, depending on which isoform was overexpressed. In control mice, sporadic semi-innervated P2 glomeruli were located within domains of the glomerular field that could be represented by circles with a diameter of 200 μm being centered 100 μm caudoventral to the major P2 glomeruli (i.e. in the vicinity of the major P2 glomeruli)(Fig. 6A). The increase of semi-innervated P2 glomeruli detected in TmRNCAM transgenic mice occurred within the same domains (Fig. 6B; Fig. 7). Although semi-innervated P2 glomeruli in control mice were slightly biased to the lateral side, the reverse was true in TmRNCAM transgenic mice(Fig. 6A,B;Fig. 7). Notably, a different pattern was evident in GpiRNCAM transgenic mice. In these mice,semi-innervated P2 glomeruli were predominantly located within a domain with a diameter of approximately 200 μm, centered 340 μm caudoventral to the major lateral P2 glomeruli (Fig. 6C; Fig. 7). The result indicated that elevated levels of TmRNCAM enhance a type of axon targeting error that occurs normally close to the major P2 glomeruli, while elevated levels of GpiRNCAM result in a different type of aberrant axon navigation and targeting (se below). Moreover, TmRNCAM predominantly affected axons projecting to the medial map, while GpiRNCAM selectively affected P2 axons to target incorrect glomeruli within a domain distant to the main lateral P2 glomeruli.

The location of an OR-specific glomerulus is almost constant between individuals but some variations do occur(Schaefer et al., 2001;Strotmann et al., 2000). This local permutation in the glomerular array precluded a precise determination of whether P2 axons co-converged with axons of OSNs expressing the very same OR gene in different mice. Importantly however, semi-innervated P2 glomeruli in GpiRNCAM transgenics showed a similar location in different mice, i.e. 330±16 μm (n=22) relative to the major lateral P2 glomeruli. Using the same way of measuring revealed a distance of 705±18 μm (n=22) between major medial and lateral P2 glomeruli. These results then indicated that the positions of invariant P2 glomeruli and semi-innervated P2 glomeruli showed a similar degree of local permutation. The GpiRNCAM-specific phenotype thus appeared to be a consequence of co-convergence of P2 axons with axons corresponding to a limited set of OR identities.

The trajectories of axons forming semi-innervated P2 glomeruli in GpiRNCAM mice was followed from the glomeruli towards OE by analyzing serial sections of the OB (Fig. 4E). Some P2 axons segregated from the lateral division of P2 axons close to the major lateral P2 glomeruli. These misguided P2 axons thus bypass their target, turn posteriorly, course in an ventral direction and innervate a glomerulus with an OR specificity that is different from P2(Fig. 6C). Examination of P2 axons forming semi-innervated glomeruli in TmRNCAM transgenic and control mice revealed a different type of trajectory. Instead of bypassing the correct target glomeruli, P2 axons terminated within incorrect glomeruli on the way towards the correct target glomeruli (Fig. 6B; Fig. 8). An estimation of the location of semi-innervated P2 glomeruli by comparative NADPHd histochemistry on serial sections revealed that incorrectly targeted glomeruli were not in a region of the OB that corresponded to Z1(Fig. 8). This result indicated that axons of a defined Z2 OSN subpopulation do not co-converge with axons of Z1 OSNs that ectopically express RNCAM. Thus, the result supported the conclusion that the Z1/Z2-4 division of the glomerular field is formed independently of spatial RNCAM expression.

An increased number of semi-innervated P2 glomeruli were detected in transgenic mice overexpressing either of the two RNCAM splice variants. This suggested that the isoforms fulfill similar functions. The results also suggested that TmRNCAM and GpiRNCAM have distinct functions as convergence errors were located to different domains of the OB, depending on isoform overexpressed. Interestingly, mating TmRNCAM and GpiRNCAM mice generated offspring with a phenotype similar to that of TmRNCAM transgenic mice. Thus,TmRNCAM/GpiRNCAM double transgenic mice showed an increased number of semi-innervated P2 glomeruli within a domain in the vicinity of major P2 glomeruli (Fig. 6D). Moreover,the increase was biased toward the medial P2 glomerulus, similar to the situation in transgenic mice overexpressing TmRNCAM only(Fig. 7). This result indicated that TmRNCAM inhibits formation of GpiRNCAM-specific targeting errors distal to the lateral P2 glomerulus. TmRNCAM, on the other hand, exhibits a GpiRNCAM-independent function that results in targeting errors proximal to major P2 glomeruli.

RNCAM is a homophilic cell adhesion molecule related to NCAM and invertebrate orthologs such as apCAM and Fas2. We show that RNCAM-negative axons and their terminals within glomeruli in the OB can be visualized by NADPHd histochemistry. This novel marker for Z1 axons has allowed us to address the function of zonally restricted expression of RNCAM in OE. Transgenic mice with ectopic RNCAM expression in Z1 and control mice show an indistinguishable topography of NADPHd-positive and NADPHd-negative OB regions. This finding indicates that Z1 axons project to the dorsomedial OB and avoid the ventrolateral part, irrespective of whether they express RNCAM or not. Conversely, we show that axons of a defined RNCAM-positive OSN subpopulation (P2) are not attracted to the NADPHd-positive (Z1) region of the OB in RNCAM transgenic mice. This indicates that RNCAM does not determine formation of a broad regional division of projections from OE to OB. A candidate molecule with such functions is the semaphorin receptor neuropilin 2, which was recently shown to mediate zonal segregation, but not axonal convergence, of olfactory sensory neurons located in the vomeronasal organ(Cloutier et al., 2002). Supporting the notion that other genes may regulate the formation of zone-specific projections in the main olfactory system is the finding that neuropilin 2 mRNA is expressed in a step gradient, with highest relative levels in Z4 and lowest in Z1 (Norlin et al., 2001).

If RNCAM does not determine formation of a broad regional division of projections, it is possible that the topographic distribution of RNCAM expression has evolved to play a role in axon convergence and/or structural plasticity for neurons in Z2-4. In fact, previous studies of OSN-expressing tagged ORs corresponding to a RNCAM-negative Z1 OR (M72) and a RNCAM-positive Z2 OR (P2) have provided evidence for heterogeneity in glomerular formation during development and sensitivity to targeted deletion of an olfactory cyclic nucleotide-gated channel subunit (Potter et al., 2001; Zheng et al.,2000). To determine if RNCAM function influences axon convergence,we have used the P2-IRES-tau-lacZ allele to probe for navigational errors of P2 axons. One finding is that increased expression of RNCAM in OSNs results in an increased number of glomeruli that are innervated by axons that represent more than one OR identity. This indicates that normal level of RNCAM is important for the formation of highly refined axon projections of OSNs, to which RNCAM expression is normally confined.

Alternative splicing generates two RNCAM isoforms that share an identical extracellular domain that mediates homophilic cell adhesion(Yoshihara et al., 1997). Different functions of RNCAM isoforms can be envisioned in light of the fact that TmRNCAM has a cytoplasmic domain that potentially could mediate signaling and/or coupling to the cytoskeleton (Walsh and Doherty, 1997). Like other Gpianchored proteins, GpiRNCAM may also affect intracellular signaling and not only represent adhesiveness as such (Davy et al., 1999). Interestingly, we find that overexpressing TmRNCAM increases the number of targeting errors of P2 axons within domains in the immediate vicinity of their original glomerular targets, while overexpression of GpiRNCAM results in semi-innervated P2 glomeruli located distant from the lateral P2 glomerulus. Moreover, TmRNCAM suppresses the GpiRNCAM-specific phenotype, while the reverse is not true for the TmRNCAM-specific phenotype. We interpret these results to suggest that topographic targeting of OR-defined axon subclasses is influenced by two distinct RNCAM-dependent mechanisms.

We find both isoforms of RNCAM expressed on axons in the nerve layer and the distribution of alternatively spliced transcripts in OE suggests that layers containing a high proportion of maturing OSNs express both isoforms. It is thus plausible that genetic interaction of GpiRNCAM and TmRNCAM is a consequence of their co-localization on axons in the nerve layer during axonal outgrowth and navigation. Misrouted P2 axons in GpiRNCAM transgenic mice bypass the correct glomerular target and turn caudally before defasciculating at an incorrect choice point, i.e. at a glomerulus with an OR identity different from P2. The resolution of the experimental system allows us to determine that incorrectly targeted glomeruli show a similar location in different individuals. This phenotype indicates that elevated levels of GpiRNCAM reduce the specificity of OR-specific target selection. Genetic analyses in Drosophila have shown that alteration in the degree of Fas2-mediated axon-axon attraction directly influences the ability of motor axons to respond to target-derived attractants and repellents(Fambrough and Goodman, 1996;Winberg et al., 1998). For example, the combined function of Fas2 and semaphorins is involved in controlling selective defasciculation important for target selection during the generation of neuromuscular connectivity(Winberg et al., 1998;Yu et al., 2000). The possibility that the mechanism of target selection in Drosophila is phylogenetically conserved, in order to be applicable for OR-specific axon segregation in mammals, is appealing given that neuropilin 2 expression correlates with OR zones (Cloutier et al.,2002; Norlin et al.,2001). A GpiRNCAM-mediated change in the balance between RNCAM and neuropilin 2-mediated axon-axon attraction and repulsion, respectively, may thus increase the probability of defasciculation of P2 axons in an region of the OB that corresponds to an incorrect OR zone.

Errant axons in TmRNCAM transgenic mice, unlike those in GpiRNCAM transgenic mice, terminate within incorrect glomeruli on the way towards the correct target glomeruli. Moreover, targeting errors were located close to medial and lateral P2 glomeruli within domains in which sporadic semi-innervated glomeruli are found also in control mice. Previous studies have shown that OSNs expressing related ORs belonging to the same zonal set tend to form glomeruli in close proximity to each other(Conzelmann et al., 2001;Malnic et al., 1999;Tsuboi et al., 1999). Analyses of tagged OR genes indicate that axons of highly related OR identities segregate from each other during a short postnatal phase and that only a few misprojections can be detected in glomeruli close to the correct target glomerulus at 1 week of age (Conzelmann et al., 2001). It is thus conceivable that regulated levels of TmRNCAM influences precision of targeting within a domain of glomeruli with highly related OR identities or, alternatively, glomeruli that belong to the same zonal set of OR identities.

The finding that increased levels of TmRNCAM enhance one type, and suppress another type, of OR-specific targeting error is intriguing in light of the fact that presence of TmRNCAM on the axonal surface appears to be regulated. Yoshihara et. al. have reported that odor stimulation causes disappearance of immunoreactivity corresponding to an epitope (R4B12) of the rabbit ortholog of RNCAM (Yoshihara et al.,1993). Diminished RNCAM immunoreactivity is limited to axon terminals within glomeruli. The fact that we find TmRNCAM to be the predominant isoform in glomeruli opens up the possibility that TmRNCAM and not GpiRNCAM is regulated as a consequence of odor exposure. Others have shown that increased intracellular cAMP results in selective internalization of the Tm form of apCAM because of phosphorylation of a sequence (PEST) in the intracellular domain (Bailey et al.,1997). As TmRNCAM contains a PEST sequence(Yoshihara et al., 1997), it is tempting to speculate that formation and/or maintenance of precise OR-specific connections is influenced by dynamic changes of surface-expressed TmRNCAM levels. Supporting this notion is the fact that the intensity of RNCAM immunoreactivity varies slightly between individual glomeruli (data not shown).

Axons of OSNs expressing a given OR take two routes that project to glomeruli located in either the medial or the lateral OB. An equal representation of OR-specific glomeruli and OR-zones in both the lateral and medial division suggests that each bulb has two symmetrical sensory maps(Nagao et al., 2000). Interestingly, the GpiRNCAM-specific phenotype is restricted to the lateral mirror map, while OSNs contributing to the medial mirror map was more sensitive to the effect of TmRNCAM. Our result thus suggests that the two mirror maps in each bulb are not formed or maintained with identical mechanisms. We do not find evidence for an uneven distribution of RNCAM mRNA splice variants or RNCAM immunoreactivity in glomeruli along dorsoventral,mediolateral and rostrocaudal aspects of OE or OB. Medially and laterally projecting P2 OSNs may be intrinsically different. Alternatively, differential sensitivity to the effects of the two RNCAM isoforms can be a consequence of interactions with signals having medial and lateral distribution. To address such issue it will be important to identify molecular cues that divide OSN projections in two mirror image maps, as well as proteins that directly interact with RNCAM.

Taken together, we have uncovered three important functional characteristics of RNCAM. First, RNCAM does not determine the division of OSN projections into Z1 and Z2-4, respectively. Second, normal regulated levels of RNCAM are important for accurate OR-specific axon segregation of Z2-4 OSNs. Finally, we identify specific activities of RNCAM splice variants that influence formation of precise topographic axon connections in the vicinity and distant relative to the correct target.

We thank Dr Peter Mombaerts for kindly providing P2-IRES-taulacZmice, Dr Frank Margolis for OMP antibody, Dr Bo Servenius for mouse genomic OMP clone and Dr A. Berghard for helpful comments on the manuscript. This work was supported by grants from the Swedish Natural Science Research Council(B5101-1250/2001), Petrus & Augusta Hedlund's Fund, the Swedish Society for Medical Research and the Magnus Bergvalls Fund.