A Possible Role for Caveolin as a Signaling Organizer in Olfactory Sensory Membranes*

Fast kinetics and sensitivity of olfactory signaling raise the question of whether the participating proteins may be associated in supramolecular transduction complexes. We found evidence that caveolin proteins could play an important role in organizing signaling elements in olfactory sensory neurons. Western blot analysis indicated that caveolins are highly enriched in olfactory sensory membranes, where they co-localize in detergent-insoluble complexes with key components of the signaling pathways. Furthermore, the results of immunoprecipitation experiments suggest that G proteins and effector enzyme form preassembled subcellular complexes with caveolins. Since anti-caveolin antibodies and synthetic peptides derived from the scaffolding domains of caveolin-1 and caveolin-2 effectively attenuated second messenger responses in sensory cilia preparations in a characteristic manner, the data led to the suggestion that caveolins could mediate the assembly of signaling complexes within specialized membrane microdomains of olfactory sensory neurons.

In mammals, odorants bind to heptahelical receptors on specialized sensory neurons, stimulating heterotrimeric G proteins associated with the inner face of the plasma membrane. Consequently, effector enzymes are activated, resulting in a fast and transient pulse of second messengers, which finally trigger the electric response of the cell (1)(2)(3)(4)(5). The chemosensory signal transduction process is highly sensitive (6,7) and extremely rapid (8). These findings fuel the hypothesis that the participating molecules may be organized in preassembled signaling complexes to minimize diffusion routes and to control cross-talk between different signaling pathways (9). Specialized membrane microdomains and scaffolding proteins have been proposed as structural organizers for such multimolecular signaling assemblies in various cell types (10 -12).
Simons and Ikonen propose a "raft hypothesis" for the lateral organization of lipid bilayers (12). Accordingly, assemblies of sphingolipids and cholesterol create biochemically distinct microdomains (rafts) in a glycerophospholipid environment. Such rafts may recruit and exclude specific sets of membrane proteins based on their physicochemical properties and may therefore be viewed as platforms to concentrate signaling molecules within the lipid bilayer. The properties of caveolin could enable this protein to sequester raft lipids, eventually forming and organizing large stabilized raft domains in the plasma membrane (28,29). In addition, caveolins reportedly interact with multiple components of G protein-mediated signaling pathways, including receptors, G proteins, and effector enzymes (14, 30 -33). Accordingly, caveolin-mediated clustering and dispersion of raft domains could contribute to regulating and shaping G protein-mediated signaling.
In this study we addressed the question of whether caveolins may play such a role in olfactory sensory neurons. We examined their subcellular distribution in the main olfactory epithelium (MOE) 1 and vomeronasal organ (VNO) of the rat and evaluated their possible function in the odor-induced second messenger signaling.

Methods
Isolation of Sensory Cilia from Main Olfactory Epithelia-Olfactory cilia were prepared using the calcium shock method according to Anholt (34). Briefly, after a short wash of the olfactory epithelia in ice-cold Ringer solution (120 mM NaCl, 5 mM KCl, 1.6 mM K 2 HPO 3 , 25 mM NaHCO 3 , 7.5 mM glucose, pH 7.4), the tissue was subjected to Ringer solution supplemented with 10 mM calcium chloride and gently stirred for 5 min at 4°C. Detached cilia were isolated by three sequential centrifugation steps for 5 min at 7,700 ϫ g. The supernatants were collected, and the resulting pellets were resuspended in Ringer solution including 10 mM calcium as described above. The cilia preparation was collected by a final centrifugation step of the pooled supernatants for 15 min at 27,000 ϫ g. The resulting pellet containing the cilia was resuspended in hypotonic TME buffer (10 mM Tris, 3 mM MgCl 2 , 2 mM EGTA, pH 7. 4) and stored at Ϫ70°C.
Isolation of Sensory Microvilli Fragments from Vomeronasal Organs-Microvilli fractions from rat VNO were prepared as described previously (35). Briefly, VNOs were removed from fertile female rats, washed twice in Ringer solution, and stored frozen in liquid nitrogen until further use. VNOs from 30 -60 animals were thawed on ice, minced, and subsequently stirred for 10 min at 4°C in Ringer solution containing 10 mM calcium chloride. After removing the debris by centrifugation (10 min, 3,000 ϫ g), the supernatant was collected while the pellet was resuspended in Ringer solution containing 10 mM calcium and processed as described above. The pooled supernatants were centrifuged for 30 min at 48,000 ϫ g, and the resulting pellet containing the microvilli membrane fragments was resuspended in hypotonic TME buffer and stored in aliquots at Ϫ70°C. Protein concentrations were measured by the Bradford method (36).
Isolation of Triton X-100-insoluble Membrane Domains-Detergentinsoluble membrane fractions were purified as described previously (37). Briefly, freshly dissected rat olfactory epithelia from the nasal septum and the ethmoid turbinates or from VNOs were collected and minced in TME buffer. An aliquot of 1 ml of protein solution (4 g/l) was diluted with lysis buffer (25 mM Mes, 150 mM NaCl, 1 mM EGTA, 1 mM phenylmethylsulfonyl fluoride, 10 g/ml each of aprotinin, leupeptin, and pepstatin, pH 6.5), adjusted to a final concentration of 1% (v/v) Triton X-100 or, alternatively, 60 mM octyl glycoside, and solubilized for 10 min on ice. Subsequently, samples were mixed with an equal volume of 80% sucrose in lysis puffer, placed on the bottom of a 12-ml centrifuge tube, and overlaid with a discontinuous sucrose gradient (6 ml of 30% sucrose, 4 ml of 5% sucrose, both prepared in lysis buffer). Samples were centrifuged at 200,000 ϫ g for 16 h. Fractions of 1 ml were collected from the top (fraction 1-12), whereas the pellet of the gradient was resuspended in 1 ml of lysis buffer (fraction 13). Aliquots of each fraction were stored at Ϫ70°C and subjected to SDS-PAGE.

SDS-PAGE and Western
Blot Analysis-Aliquots of sucrose gradient fractions or protein samples from sensory epithelia (38) were mixed with 5ϫ sample buffer, boiled for 2 min, and subjected to SDS-polyacrylamide gel electrophoresis using the Laemmli buffer system (39).
The separated proteins were transferred onto nitrocellulose membranes using a semidry blotting system (Amersham Pharmacia Biotech). The blot was stained with Ponceau S, dried, and stored at 4°C until further use. For Western blot analysis, nonspecific binding sites were blocked with 5% nonfat milk powder (Naturaflor, Dietmannsriel, Germany) in 10 mM Tris/HCl, pH 8.0, 150 mM NaCl, and 0.05% Tween 20 (TBST). Subsequently, the blots were incubated overnight at 4°C with specific antibodies diluted in TBST containing 3% nonfat milk powder. After three washes with TBST, a horseradish peroxidase-conjugated goat anti-rabbit IgG (1:10,000 dilution in TBST with 3% milk powder) was applied. After three washes with TBST, the ECL system was used to visualize bound antibodies.

RESULTS
To analyze the expression of caveolins in rat olfactory neurons, preparations of MOE and VNO were probed with sub-type-specific anti-caveolin antisera on Western blots. For this purpose, total MOE and VNO samples were prepared by homogenizing freshly dissected tissue in TME buffer. Sensory cilia and microvilli fragments were separated from cells by the addition of calcium chloride and purified by multiple centrifu-FIG. 2. Subcellular fractionation of olfactory sensory tissue from the MOE. Tissue from the MOE was homogenized in either lysis buffer with 1% Triton X-100 (A-C) or 60 mM octyl glycoside (D) and fractionated by discontinuous sucrose gradient centrifugation. Equal aliquots of each fraction were subjected to SDS-PAGE, and the gels were stained with Coomassie Blue (A). Fractions 1-8 represent the 5-30% sucrose layers, fractions 9 -12 represent the 40% sucrose layer, and fraction 13 represents the insoluble pellet. Equivalent gels were blotted on nitrocellulose membranes, where immunoreactivity of caveolin-1 (B) and caveolin-2 (C) was visualized with subtype-specific antibodies (1:1,000 dilution). The bulk of caveolin immunoreactivity was found in fractions 3 and 4, which exclude more than 90% of total cellular protein as determined by densitometric analysis of the Coomassie gels. In contrast, solubilization with octyl glycoside shifted caveolin-1 immunoreactivities to the low density fractions 10 -13 (D).
FIG. 3. Subcellular fractionation of olfactory sensory tissue from the VNO. Low density Triton X-100-insoluble fractions were obtained from sensory tissue of the VNO using a sucrose density gradient centrifugation procedure. Equal aliquots were collected from the top of the gradient and subjected to SDS-PAGE. Upon solubilization at 4°C with Triton X-100, endogenous caveolin-1 (A) and caveolin-2 (B), visualized with subtypespecific antibodies (1:1,000), are restricted to the low density fractions 3 and 4, whereas upon treatment with octyl glycoside, caveolin-1 is localized to the high density fractions 10 -13 (C). gation steps. Equal amounts of protein from each preparation were subjected to SDS-PAGE and blotted onto nitrocellulose membranes. The membranes were subsequently probed with caveolin-1-and caveolin-2-specific antibodies, followed by peroxidase-conjugated secondary antibodies and chemiluminescence detection.
As shown in Fig. 1, both antibodies recognized a band of about 25 kDa in all preparations, consistent with the predicted molecular masses of caveolin-1 and -2 of about 22.000 (30). Comparing the relative band intensities, we found both caveolin subtypes enriched in cilia and microvilli fractions of MOE and VNO, respectively. Centrifugation at high speed of cilia and microvilli lead to a further enrichment of immunoreactivity of both caveolin subtypes in the membrane pellet (data not shown). Antibodies directed to the muscle-specific subtype caveolin-3 did not label a polypeptide at the predicted size of this caveolin subtype (data not shown).
The data indicate that caveolin-1 and -2 are expressed in both rat olfactory organs, MOE and VNO. Furthermore, enrichment of both caveolin subtypes in sensory membranes points to a possible colocalization with components of olfactory signal transduction cascades in the appropriate subcellular compartments.
Caveolins have been found to segregate into detergent-insoluble glycolipid-enriched membrane domains (DIGs) (31). To test whether this may also be the case for caveolins expressed in olfactory tissues, we isolated low density membrane fractions from MOE and VNO using sucrose gradient ultracentrifugation. In the past, association with detergent-insoluble cytoskeleton proteins was made responsible for Triton X-100 insolubility at 4°C. The detection of multiple glycosylphosphatidylinositol-anchored proteins in DIGs, however, established Triton X-100 insolubility as the basis for isolating lipid rafts and associated proteins (44).
Olfactory epithelia from MOE and VNO were homogenized in Triton X-100-containing lysis buffer at 4°C, and after centrifugation, equal volumes of each of the 13 fractions were subjected to SDS-PAGE. Proteins were visualized by Coomassie Blue staining ( Fig. 2A), revealing a highly skewed protein distribution along the gradient. As determined by protein quantification for each fraction (data not shown), more than 90% of the total protein was found in high density fractions representing Triton X-100-soluble components (lanes 9 -13 in Fig. 2A). Only 10% of the total protein localized to the low density fractions (lanes 1-8). In contrast, when the fractions were subjected to Western blot analysis as described above, most of the caveolin-1 and -2 immunoreactivity was found concentrated in low density fractions 3-5 of the MOE (Fig. 2, B and C) and the VNO (Fig. 3, A and B). Replacing Triton X-100 in the lysis buffer by 60 mM octyl glycoside completely abolished this pattern, leaving all caveolin-1 immunoreactivity from MOE (Fig. 2D) and from VNO (Fig. 3C) in the high density fractions 10 -13. Similar results were obtained for caveolin-2 (data not shown). These results demonstrate that both caveolin subtypes show a strong tendency to associate with DIGs isolated from olfactory epithelia as described for other tissues (45).
To test whether key components of the olfactory G proteincoupled signal transduction cascades are enriched within DIG fractions from rat olfactory tissues along with caveolins, we monitored the distribution of two major olfactory G␣ subtypes from MOE, G␣ s and G␣ o (46) in sucrose density gradients. Western blot analysis with subtype-specific anti-G␣ antibodies detected some immunoreactivity for G␣ s (Fig. 4A) and G␣ o (Fig.  4B) in Triton X-100-soluble high density fractions 7-12. The most of these proteins, however, were concentrated in Triton X-100-insoluble low density fractions 4 -6. Replacing Triton X-100 by 60 mM octyl glycoside in the lysis buffer released the immunoreactivity for both proteins (data shown for G␣ s only) into the soluble high density fractions (Fig. 4C).
Similar results were obtained when preparations from VNO were used (Fig. 5). G␣ i and G␣ o subtypes have been suggested to mediate chemosensory signal transduction in this tissue (47)(48)(49). Again, the majority of both G protein ␣-subunits comigrated with the Triton X-100-insoluble low density fractions. Comparable experiments with an antibody against a conserved domain of all five known rat G␤ subtypes revealed a similar distribution (data not shown).
To test whether further downstream signaling molecules of the olfactory transduction pathways co-migrate in the low density fractions along with caveolins and G proteins, the distributions of two key effector enzymes, adenylyl cyclase III and PLC␤2 (50,51), were monitored in sucrose density gradients. Again, homogenates from MOE were prepared as described above and subjected to sucrose gradient centrifugation. After SDS-PAGE separation of each fraction, the proteins were blotted onto nitrocellulose membranes and probed with antibodies specific to rat adenylyl cyclase III (Fig. 6A) and PLC␤2 (Fig. 6B) proteins. We found significant immunoreactivity for both enzymes in Triton X-100-insoluble low density fractions 4 -6, suggesting their co-localization with caveolins and G proteins in common lipid compartments of sensory membranes.
Several studies indicate that in vivo, caveolin 1 and caveolin 2 form a hetero-oligomeric complex (for review, see Ref. 16). To explore whether caveolin-1 may interact with caveolin-2, coimmunoprecipitation experiments were performed using a caveolin-1-specific antibody conjugated to agarose; therefore, Triton X-100-insoluble low density membrane fractions iso- FIG. 7. Co-immunoprecipitation of caveolin-1 with G protein ␣-subunits and effector enzymes of olfactory signal transduction pathways. Triton X-100-insoluble low density caveolin-enriched fractions of olfactory sensory tissue of the MOE was solubilized with octyl glycoside and immunoprecipitated either with non-immune rabbit IgG (IgG) or with an anti-caveolin-1 antibody conjugated to agarose; after washing, the precipitates were extracted with Laemmli buffer and subjected to SDS-PAGE; subsequently, samples of isolated olfactory cilia (Ci) as well as immunoprecipitates obtained with the caveolin-1 antibody or control IgGs were immunoblotted with an anti-caveolin-1 antibody as a control lated by sucrose density centrifugation were solubilized with octyl glycoside and subsequently used in immunoprecipitation experiments; to assess the specificity of the procedure, protein samples were incubated either with non-immune rabbit IgG (Fig. 7, IgG), or alternatively, antibodies used in Western blotting were preincubated with the specific peptide (Fig. 7, ϩP). As demonstrated in Fig. 7, caveolin-1 was effectively precipitated by its respective antibody (panel A, Cav1) compared with the labeling of the caveolin-1 antibody observed in isolated olfactory cilia (panel A, Ci), whereas no immunostaining was detectable in control samples (panel A, IgG and Cav1 ϩ P). When caveolin-1 precipitates were probed with caveolin-2-specific antibodies, an intense immunoreactivity for caveolin-2 was observed (Fig. 7, panel B, Cav2), indicating that heterooligomeric complexes of caveolin-1 and caveolin-2 exist in the olfactory sensory epithelium.
Caveolin forms higher ordered complexes with a variety of signal transduction molecules (for review: see Ref. 16); to determine if caveolin-1 is associated with olfactory G protein ␣-subunits, caveolin-1 precipitates were probed with subtypespecific antibodies for G␣ s and G␣ o ; as demonstrated in Fig. 7, no staining for both G protein-subtypes was detected in control samples, whereas in the caveolin precipitates, a significant amount of G␣ s (panel C) and G␣ o (panel D) was detected. To explore whether effector enzymes involved in olfactory signaling are also associated with caveolin, immunoprecipitates of caveolin-1 antibodies were also assessed for adenylyl cyclase and PLC␤2 immunoreactivity; Fig. 7 shows that adenylyl cyclase (panel E) and PLC␤2 (panel F) are both co-eluted with caveolin-1, indicating that G protein ␣-subunits as well as effector enzymes of the olfactory signaling pathways form complexes with caveolin-1.
To examine the functional significance of interactions between caveolins and other olfactory signaling molecules we performed two different experimental approaches. Using antibodies specific to caveolin-1 (Fig. 8A) and caveolin-2 (Fig. 8B), we attempted to sterically protect caveolin proteins from homoand heterophilic interactions. In addition, we sought to competitively replace caveolins from potential interaction partners with synthetic peptides corresponding to the amino acid sequence of caveolin-1 (Fig. 9A) and caveolin-2 (Fig. 9B)  ing domains, which reportedly are involved in homo-and hetero-oligomerization of caveolins and in binding to other signaling proteins (20,23,52,53).
Samples of cilia preparations were preincubated with different dilutions of antibodies specific to caveolin-1 or caveolin-2 and subsequently stimulated with a mixture of odorants inducing cAMP responses (citralva, hedione, eugenol; 10 M each) (40) or odorant components eliciting IP 3 formation (lilial, lyral, ethylvanillin; 10 M each) (54). Alternatively, downstream signaling elements were activated directly with GTP␥S for G proteins and forskolin for adenylyl cyclase. The reaction was stopped by the addition of perchloric acid, and second messenger responses were quantified. Both anti-caveolin-1 (Fig. 8A) and anti-caveolin-2 (Fig. 8B) antibodies effectively inhibited cAMP formation in a concentration-dependent manner. Notably, when stimulating downstream elements of the signaling cascade directly, similar effects were observed, but the effectiveness of both antibodies was gradually reduced the further downstream the stimuli acted. The antibodies (1:1000 dilution) elicited a reduction to ϳ30% of the control cAMP response when stimulating with odorants. In contrast, when stimulating G proteins with GTP␥S, a reduction to ϳ50% of the control level was observed for both antibodies. Finally, when adenylyl cyclase III was directly activated by forskolin, the observed level of cAMP was reduced to ϳ60% of control values. A similar pattern was observed for IP 3 -mediated signaling (data not shown), although direct stimulation of phospholipase C was not possible because no selective agonist for this enzyme is available.
Preincubation of cilia with synthetic peptides representing the caveolin-1 (Fig. 9A) and caveolin-2 (Fig. 9B) scaffolding domains instead of the antibodies induced comparable effects. Concentrations of 10 M caveolin-1 scaffolding domain peptide reduced the odor-induced raise in second messenger production to 17% for cAMP and to less than 10% for IP 3 (data not shown). Experiments with caveolin-2 scaffolding domain peptides led to similar results. Preincubation with a control peptide unrelated to the scaffolding domain showed no effect on either cAMP (Fig.  9C) or IP 3 (not shown) signaling. As observed for the antibody experiments, the reduction of odorant-induced cAMP production was most pronounced. Both peptides evoked gradually smaller effects when the second messenger response was induced by GTP␥S (30 -40% of control values) or forskolin (45-65% of control values).

DISCUSSION
The results of this study suggest that caveolins may play an important role in olfactory signaling by assembling elements of the transduction machinery in microdomains of the chemosensory membranes. Immunoreactivity for caveolin-1 and caveolin-2 was found to be enriched in membrane protein fractions from cilia and microvilli preparations, the specialized chemosensory compartments of olfactory neurons in main olfactory epithelia and vomeronasal organs, respectively. Furthermore, caveolin-1 and caveolin-2 co-migrated in low buoyant density fractions along with several key components of the olfactory signal transduction cascades, including G proteins, adenylyl cyclase III, and phospholipase C. This indicates a possible co-localization of these molecules in situ in detergent-insoluble membrane domains, such as rafts or caveolae. Co-immunoprecipitation experiments further support the idea of a possible molecular association of caveolins with key olfactory signaling molecules. Most importantly, antibodies to both caveolins as well as synthetic peptides representing the scaffolding domain effectively suppressed the odor-induced second messenger formation.
It has been suggested that caveolins may act as scaffolding proteins to organize preassembled signaling complexes at the plasma membrane (30, 16), but models for molecular mechanisms underlying this function remain incomplete. Simons and co-workers (29,55) recently proposed that caveolin-1 might stabilize lipid raft microdomains. Small and highly dynamic lipid domains that are rich in cholesterol and sphingolipids are thought to form spontaneously in biological membranes based on the differential miscibility of lipids (12, 29, 56 -58). Caveolins may stabilize these domains, since they bind cholesterol (21) and oligomerize to form large assemblies of up to 600 kDa (23). Cooperatively, caveolin-1 and caveolin-2 may therefore induce the formation of large lipid domain clusters, regulating their size by intermolecular cross-linking. This possibility is particularly attractive because the oligomerizing property of caveolin-2 may be modified by phosphorylation through protein kinases (55), thus offering a potential cue for physiological regulation.
In this context, our observations are consistent with the notion that in chemosensory membranes caveolins may contribute to assemble the olfactory signal transduction machinery in microdomains, thereby shaping and regulating the odorinduced second messenger responses. Antibodies to both caveolins effectively attenuated the odor-induced second messenger formation (Fig. 8). This inhibitory effect could result from steric or competitive hindrance of a direct interaction between caveolins and other signaling molecules. The same may hold true for the attenuating effect of synthetic peptides, representing the caveolin scaffolding domains (Fig. 9). The observation that scaffolding domains in fact interact with G proteins (52) and other signaling molecules (53, 59 -63) is supportive to this idea. However, it was found that putative interaction partners bind differentially to each of the caveolins (53). Furthermore, scaffolding peptides from the two caveolins elicited different effects on G protein function and adenylyl cyclase III activity (25,64). In the view of these findings it is somewhat surprising that peptides representing the scaffolding domains of caveolin-1 and caveolin-2 equally affected the olfactory signal transduction cascades.
A solution for this controversy may be provided by a slightly different consideration. The scaffolding domain peptides, irrespective of whether they represent caveolin-1 or caveolin-2, as well as antibodies to both proteins could disturb the homo-and heterophilic interactions between caveolin proteins themselves, a view that is consistent with the role of the scaffolding domain in caveolin oligomerization (20,23,65). This could trigger the dispersion of large stable lipid domain clusters into small, dynamic microdomains, releasing the different signaling molecules into a much larger diffusion territory and, consequently, reducing the transduction efficiency. This notion is consistent with the finding that all used antibodies and peptides attenuate second messenger formation equally effectively and that scaffolding domain-derived peptides only partially affected co-precipitation of olfactory G proteins and effector enzymes (data not shown). Furthermore, it would provide a possible explanation for the observation that the inhibitory effect was more pronounced when second messenger formation was induced via odorant receptors as compared with responses induced by direct activation of G proteins and adenylyl cyclase III. It seems an intriguing question, whether caveolin-mediated coalescence and dispersion of lipid domains could play a role in regulating the olfactory signal transduction processes or G protein-coupled signaling cascades in general.