Cellular Prion Protein and Caveolin-1 Interaction in a Neuronal Cell Line Precedes Fyn/Erk 1/2 Signal Transduction

It has been reported that cellular prion protein (PrPc) is enriched in caveolae or caveolae-like domains with caveolin-1 (Cav-1) participating to signal transduction events by Fyn kinase recruitment. By using the Glutathione-S-transferase (GST)-fusion proteins assay, we observed that PrPc strongly interacts in vitro with Cav-1. Thus, we ascertained the PrPc caveolar localization in a hypothalamic neuronal cell line (GN11), by confocal microscopy analysis, flotation on density gradient, and coimmunoprecipitation experiments. Following the anti-PrPc antibody-mediated stimulation of live GN11 cells, we observed that PrPc clustered on plasma membrane domains rich in Cav-1 in which Fyn kinase converged to be activated. After these events, a signaling cascade through p42/44 MAP kinase (Erk 1/2) was triggered, suggesting that following translocations from rafts to caveolae or caveolaelike domains PrPc could interact with Cav-1 and induce signal transduction events.


INTRODUCTION
Multiproteic caveolar complexes represent a sophisticated membrane organization involved in signal transduction. Their efficiency is linked to the insertion of proteins in a restricted membrane area (50-100 nm) where the generation of a signal has a vectorial and oriented characteristic and allows the recruitment of low-abundance proteins in order to activate signaling pathways which, in neural cells, seem to control differentiation and cell survival [1][2][3][4]. Caveolae are a subclass of membrane microdomains distinguishable by their shape (they are flask-like invaginations) and by the presence of membrane proteins of the caveolin family. Caveolin-1 (Cav-1) is a small 22 kDa highly versatile protein capable of organizing several caveolar functions. Lisanti and coworkers have precisely mapped the molecule defining two sites involved in the binding of caveolar constituents: a hydrophobic region (aa 82-101) called scaffolding domain (SD), and a more hydrophilic motif present in the C-terminal region indicated as CID motif [5].
Cellular prion protein (PrPc) is a secreted protein anchored to cell surface via a GPI anchor and believed to function as a cell surface receptor [6] or ligand [7,8]. PrPc is characterized by an amino terminal unstructured highly flexible region characterized by the presence of multiple octapeptide repeats highly conserved during evolution that are binding sites for copper ions [9]. In neuroblastoma cells lacking caveolae, PrPc has been isolated in detergent-insoluble complexes denominated "caveolae-like domains" (CLDs) and it has been hypothesized that PrPc conversion in its pathological conformer PrP scrapie (PrPsc) occurs in this subcellular compartment [10,11]. Recent data obtained by electron microscopy in CHO cells clearly confirmed that PrPc is internalized by caveolae [12]. Moreover, it has been observed that Cav-1 is coimmunoprecipitated by using PrPc antibody and that Cav-1 mediates the recruitment and the activation of Fyn kinase after anti-PrPc antibody-mediated stimulation [13,14]. Evidence supporting a role of PrPc in regulating cell proliferation, differentiation, and survival has been collected [15].
Fyn kinase is a member of Src family kinase involved in signal transduction events. It has been reported that Fyn kinase during signal transduction events is noncovalently associated with glycosylphosphatidylinositol (GPI)-anchored proteins [16][17][18] and that Fyn kinase, after the palmitoylation of its Cys3, is included in caveolae [19]. Moreover, it has been shown that, following antibody-mediated cross linking, GPI-anchored proteins lead to signal transduction events in T cells, B cells, monocytes, and granulocytes [20] and that are sequestered in caveolae [21].
Findings reported here demonstrate that PrPc and Cav-1 interact in vitro and colocalize in GN11 cells, a hypothalamic neuronal cell line that highly expresses Cav-1 gene. Moreover, we examined the role played by caveolae and PrPc in signal transduction by transfecting GN11 cells with a novel PrPcexpressing vector showing a high transfection efficiency, in order to compare Fyn and Erk 1/2 kinases activity in wildtype and PrPc-overexpressing cells. Our results highlight the key role of caveolae as sophisticated microenvironments in which PrPc clusters to generate signal transduction pathways.

Plasmid construction
Haemagglutinin-tagged PrP (PrP-HA) plasmid was prepared as follows: the coding region (23-254) of mouse PrPc was prepared, as described in Negro et al [34], with two Mets at positions 108 and 111 (L108M, V111M) to provide the epitope specific for commercial Mab 3F4. Then, the coding region was cloned into the mammalian expression vector pRK7HA (a kind gift of Dr Elisabetta Ciani, Bologna, Italy), downstream of the HA-tag sequence, between BamHI and EcoRI restriction sites. The region coding for mouse PrP signal peptide  was then cloned into the recombinant vector, upstream of the HA epitope sequence, between AgeI and HindIII restriction sites. Expression vector for human Dpl was prepared as in [34].

GST-binding assay
Recombinant full-length forms of morPrP (23-231) and hur-Dpl  were generated in, and purified from, Escherichia coli as described in Negro et al [34]. To provide the epitope specific for Mab 3F4, morPrP carried two Mets at positions 108 and 111 (L108M, V111M). Murine Cav-1 forms were recombinantly obtained as GST-fusion proteins as described in [35]. For binding assays, 0.2 μM of Glutathione S-Transferase (GST) or GST-Cav fusion proteins (prebound to glutathione-Sepharose beads) were incubated overnight with equimolar amounts of morPrP or hurDpl (in a final volume of 0.25 ml), under continuous shaking at 4 • C. Proteins bound to glutathione-Sepharose beads were eluted, washed, and immunoblotted to detect the presence of PrP or Dpl. In competition experiments, equimolar amounts of either Ab · Tg or C-20 antibodies were present during the entire incubation period.

Determination of the dissociation constant for the binding GST-Cav fusion proteins and morPrP
Different concentrations of morPrP (0.03-0.7 μM) were incubated with 0.2 μM GST-Cav fusion proteins (GST-Cav 61-101, GST-Cav 135-178, and GST-Cav FL 1-178) in the same conditions described above. Proteins bound to glutathione-Sepharose beads were eluted and exposed to western blotting assay. The blots, stained with anti-PrP Mab 3F4, were analyzed in transmitted light by ImageMaster VDS software (Pharmacia Biotech, Sweden) to measure the intensity of the bands. PrP concentration was plotted as a function of band intensity expressed in arbitrary units.

Western blot
For western blot analysis, denatured proteins were first separated by a 12% SDS-PAGE, electrophoretically transferred onto nitrocellulose and then, depending on the experiment, incubated with the specific antibodies. Immunolabeling was visualized by ECL procedure (Amersham, UK) and the band intensities were measured by ImageMaster VDS software (Pharmacia Biotech, Sweden).

Immunocytochemistry
PrPc-overexpressing GN11 cells grown on coverslips were fixed with 4% paraformaldehyde and incubated with Mab 3F4 or were stimulated by antibody-mediated ligation and fixed. Cells were permeabilized with Triton X-100 (0.1% in phosphate buffer saline (PBS)) and saturated in PBS-BSA 3% for 30 min. Then, cells were incubated with anti-Cav-1 or anti-phospho Fyn Pabs (for double fluorescence staining experiments) or with anti-Fyn kinase Pab (for triple fluorescence staining experiments). After washings, cells were incubated with anti-mouse FITC-conjugated and with antirabbit Cy5-(for double staining) or Cy3-(for triple staining) conjugated secondary antibodies. For triple immunofluorescence staining experiments, anti-Cav-1 FITC-conjugated antibody was added. Finally, slides were mounted in glycerol-PBS medium containing 30 mg/ml DABCO (Sigma, USA). Evaluation of antibody specificity was carried out either by omitting primary antibody or by using unspecific sera.

Confocal microscopy
The confocal imaging was performed on a Radiance 2000 confocal laser scanning microscope (BioRad), equipped with a Nikon 40x, 1.4 NA objective, and with a Krypton and a Red Diode lasers, to excite FITC (green) and Cy5 (red) fluorescence simultaneously. Colocalization was evaluated on medial optical sections using LaserPix software (BioRad) [38]. Briefly, the two-dimensional scatter plot diagram of each cell was analyzed to evaluate the spatial colocalization of the fluorochromes. For each scatter plot diagram, pixels with highly colocalized fluorochromes, that is, with intensity values greater than 150 gray levels (on a scale from 0 to 255) for both detectors were selected to calculate the colocalization maps and create a binary image.

Interaction between PrP and Cav-1
To evaluate the interaction between Cav-1 and PrPc, we carried out in vitro binding experiments as previously described in [35], by using four different GST-Cav1 fusion proteins and a control GST-nonfusion protein to challenge mor-PrP. Bound PrP molecules were then immunodetected by western blot analysis. Results clearly show that GST-CavFL (full lenght, aa 1-178), GST-CavSD (scaffolding domain, aa 61-101), and GST-CavCT (C-terminal region, aa 135-178) very efficiently bind morPrP. No morPrP was bound when GST alone and GST-CavNT (N-terminal region, aa 1-81) were challenged, indicating that the binding was specific (Figure 1(a), 1).
We hypothesized that the unstructured PrP N-terminal region [41] could represent an important binding site for  (2) were then visualized by western blotting with Mab 3F4 and polyclonal antibody Q55, respectively. In 3, the binding assay was carried out in the presence of polyclonal antibodies (0.2 μM) to PrP, namely, Ab · Tg recognizing the octapeptide region, and C-20 directed against the C-terminus. In the first lane of (a) 1, 2, 3, 50 ng morPrP or hurDpl were added as standard, while in the second lane of (1) and (2)  Cav-1. This region is characterized by the presence of an octapeptide sequence PHGGGWGQ repeated five times in tandem and involved in the binding of copper ions [9]. To demonstrate our hypothesis, we first evaluated whether Dpl protein binds to Cav-1. Dpl is a protein which shows high sequence homology with the C-terminus of PrPc but lacks Mattia Toni et al 5 the N-terminal octapeptide region [42,43]. To this end, we probed the hurDpl under the same conditions used for mor-PrP binding to Cav-1. However, contrary to morPrP, hurDpl showed no interaction with Cav-1 (Figure 1(a), 2). More details about the PrPc N-terminal sites involved in the binding to Cav-1 were obtained using a polyclonal antibody Ab · Tg. This antibody specifically recognized two nonadjacent octapeptide repeats (57-65 and 81-89) since a substitution G/S was sufficient to abrogate recognition. The use of Ab · Tg antibody almost completely blocked the morPrP binding to GST-Cav SD (61-101), while an antibody directed against a PrP C-terminal epitope had a much lower effect (Figure 1(a), 3). Taken together, these results indicate that the octapeptide repeats could represent a PrP binding site for Cav-1.
The results of experiments aimed at evaluating the PrP/Cav-1 binding affinity constants are shown in Figure 1(b). Amounts of morPrP ranging from 0.03 μM to 0.7 μM were incubated with different GST-caveolin fusion proteins, as described under Material and Methods. Bands intensity in western blots was quantified by densitometry. Plots representing the binding of GST-CavFL (1-178), GST-CavSD (61-101), and GST-CavCT (135-178) are shown in panels B1, B2, and B3, respectively. Evaluation of dissociation constant (Kd) by reciprocal plots showed that Kd ranged from 8 to 9 ×10 8 M −1 for all the three fusion proteins. Since no binding was observed using GST alone or GST CavNT fusion protein, it appears that unspecific binding is not occurring in our experimental conditions.
To determine if the interaction between PrP and Cav-1 observed in vitro occurred also in a physiological environment and if this interaction had any effects on signal transduction, we used, as cellular model, a cell line of immortalized luteinizing hormone-releasing hormone (LHRH) neurons that naturally expresses high levels of Cav-1 [44] and that was prone to be transiently transfected with high efficiency with a plasmid encoding the HA tagged morPrP protein.
In order to obtain a neuronal environment closer to that in which PrPc operates physiologically, we treated GN11 cells with TPA, an inducer of cell differentiation (Figure 2(b), 1-4). Analysis of PrPc-overexpressing GN11 cells treated with TPA has shown an increase in the co-localization level between PrPc and Cav-1 (Figure 2(b), 3 and 4). Since in vitro experiments (Figure 1(a), 2) have shown an incapacity of Cav-1 to bind Dpl and since Dpl is a surface protein exposed to the exoplasmic space through a GPI-anchor like PrPc [45], confocal microscopy analyses were performed on GN11 cells overexpressing the Dpl protein differentiated with TPA ( Figure 2(b), [5][6][7][8]. When Dpl was cross-linked using a specific antibody, colocalization with Cav-1 at plasma membrane level was undetectable (Figure 2(b), 7-8) confirming a different behavior between PrPc and Dpl and suggesting that Doppel, although located through a GPI anchor in the same microdomains of PrPc, could be unable to participate in the transduction of survival signals, thus favoring conditions slowly leading to apoptosis. For example, failure of Doppel to bind Cav-1 could impair Fyn kinase recruitment due to PrP c activation [13], thus altering the normal fate of cell differentiation occurring during brain development. Our results may also help to explain why the reintroduction of a wildtype Prnp gene, in PrP c 0/0 mice expressing Doppel protein in the brain, rescues them from neurodegeneration [46]. In fact, PrP c , having a higher affinity for transducing machinery, may displace Doppel protein and trigger survival signals. Evidence supporting a role of PrP c in regulating cell proliferation, differentiation, and survival has been recently collected by Satoh and coworkers [47], who compared the gene expression profile in fibroblast cell lines derived from PrP c -deficient mice (Prnp 0/0) and from PrP wild-type mice. They observed that the disruption of Prnp gene resulted in an aberrant regulation of a battery of genes important for cell proliferation, differentiation, and survival, including those located in the Ras and Rac signaling pathways. Moreover, Kuwahara and coworkers [48] observed that the reintroduction of wild-type Prnp gene prevented apoptosis in hippocampal cells derived from Prnp 0/0 mice and cultured under serum-free conditions.
To further confirm the PrPc/Cav-1 colocalization, experiments of flotation in density gradient were performed. Extracting cells with Triton X-100 at 4 • C and fractionating the insoluble material by OptiPrep or sucrose-gradient centrifugation are considered a stringent criterion to establish whether a protein is present in detergent-insoluble glycolipid-rich domains (DIGs) and colocalizes with Cav-1 [49]. Nevertheless, it is difficult to obtain a separation between rafts and caveolae. We applied this procedure to GN11 cells, transiently transfected with Prnp or Prnd, and evaluated PrPc, Cav-1, and Dpl distribution by western blot. Results obtained (Figure 3(a)) show a marked colocalization between PrPc and Cav-1 in low-density gradient fractions (Figure 3(a), 1 and 2, resp.). Interestingly, Dpl protein (Figure 3(a), 3), even if it is a GPI-linked protein too, shows a wide distribution in most fractions of the gradient that does not correlate with Cav-1 distribution.
Mouillet-Richard et al [13] detected the presence of Cav-1 in immunoprecipitated samples obtained by using an anti-PrPc antibody, particularly when neuronal cells were induced to differentiate. We proposed to confirm these results in GN11 neuronal cells transiently transfected with murine Prnpgene. Reciprocal coimmunoprecipitation experiments in HA tagged-PrPc-overexpressing GN11 cells were performed to further confirm PrPc-Cav1 interaction (Figure 3(b)). We found that anti-HA-tag antibodies coimmunoprecipitate Cav-1 (Figure 3(b), lane 2) and, reciprocally, that anti-Cav-1 antibodies coimmunoprecipitate PrP (Figure 3(b), lanes 3 and 4) in cells stimulated by anti-PrPc antibody-(30 min) and by secondary antibody-(30 min) mediated ligation. After deglycosylation treatment (Figure 3(b), lane 4), the smear typical of PrPc was resolved in a single band of 27 kDa. Interestingly, the sensitivity of western blot detection did not permit to reveal Cav-1 in coimmunoprecipitates of unstimulated cells (lane 1) confirming the importance of antibody stimulation in PrPc/Cav-1 interaction.
To follow the fate of PrPc/Cav-1 complexes, a timecourse of PrPc and Cav-1 distributions after anti-PrPc antibody-mediated stimulation was analyzed by confocal microscopy in living GN11 cells overexpressing PrPc. In Figure 4 it is shown that the PrPc/Cav-1 colocalized signal, at plasma membrane level, was evident after 60 min of anti-PrPc antibody-mediated stimulation (A-D). However, after 90 min, PrPc was prominent in the intracellular space and only a few spots of colocalized signals were present at plasma membrane level (E-H). After 150 min, PrPc became almost exclusively intracellular (I) while Cav-1 distribution appeared unchanged (J), and colocalization signals were un-detectable (K), (L). It is likely that when caveolae are detached from surface membrane the dissociation of PrPc/Cav-1 complexes takes place. However, western blotting analysis of PrPc at different times (M, N, and O) showed that no degradation of PrPc, even after 2 h, was evident. Moreover, the extent of PrPc glycosylation was unchanged.

Involvement of PrPc in signal transduction
Since we hypothesized a role of PrPc clustering and of PrPc-Cv 1 colocalization in signal transduction, we performed confocal microscopy experiments to assess if Fyn kinase responds, in someway, to an anti-PrP antibody-mediated stimulation ( Figure 4). Biedi and collaborators demonstrated the recruitment of Fyn kinase from cytoplasm to caveolae in fibroblasts overexpressing wild-type IGF-IR stimulated by IGF-I [50]. In a similar way, after the stimulation of PrPc-overexpressing GN11 cells with anti-PrPc antibody (30 min) and with secondary antibody (30 min) to induce PrPc patching, we observed the shift of Fyn kinase from the cytoplasm towards the membrane (in Figure 5(c), compare the Fyn distribution in a transfected cell subjected to the stimulation (arrowhead) with an untransfected cell not subjected to the stimulation (arrow)). After the antibody-mediated stimulation, we revealed a high  colocalization between PrP and Cav-1 (panel D) and between PrP and Fyn kinase (panel E) on the plasma membrane. It was previously reported that PrPc and Fyn kinase do not colocalize in neurons; however in those experiments cells were not stimulated by anti-PrPc antibody [51].
To evaluate if the Fyn kinase recruited in the clusters of PrPc/Cav-1 was activated, we performed confocal immunofluorescence experiments by using a specific antibody that recognizes the active form of Fyn kinase phosphorylated on Tyr 416 ( Figure 6). In GN11 cells that do not express PrPc, the signal of active Fyn kinase is low and wide dispersed in the cytoplasm (data not shown). On the contrary, in PrP-overexpressing GN11 cells we detected a low activation of Fyn kinase on plasma membrane ( Figure 6, B and F). Following anti-PrPc antibody-mediated stimulation (20 min), an increased level of active Fyn kinase located in the correspondence of PrPc clusters was detected (J, N, arrowheads indicate PrPc clusters). After 40 min of antibody stimulation, the signal of active Fyn kinase was weaker and after 60 min it was no more detectable (data not shown). Taken together, these data show that following anti-PrPc antibody mediated stimulation, PrPc patches on the plasma membrane in clusters in which Fyn kinase is recruited and activated.
It has been reported that activated Fyn kinase can transduce a signal cascade through Erk 1/2 kinase [52]. To evaluate if in GN11 cells the activation of PrPc by antibody-mediated ligation triggers a signal transduction pathway through Erk 1/2, we perform the experiments reported in Figure 7  Erk 1/2 signaling pathway. Our data, obtained in a neuronal model expressing Cav-1, confirm that PrPc triggers a Fyn kinase and Erk 1/2 signaling pathway as previously reported [15,31,32,53].
To obtain this activation, PrPc and Fyn kinase need to co-localize in caveolae: PrPc clusters on the plasma membrane following the antibody (or other putative physiological ligands) binding, and Fyn kinase moves from cytoplasm to caveolae.
In order to demonstrate the essential role of Cav-1 in the PrPc-Fyn-Erk 1/2 signaling pathway, we down-regulated or misallocated caveolin 1 by using anti-Cav-1 antisense oligonucleotides and Filipin III, respectively. The aim was to obtain an experimental condition in which the absence of Cav-1 could prevent the activation of the transduction cascade triggered by PrPc clustering. However, these treatments in GN11 cells did not permit to obtain clear results as they caused a strong increase of Erk 1/2 phosphorylation level (data not shown). The ability of Cav-1 to directly inhibits Erk 1/2 activation has been reported in other experimental models [25,26]. Moreover, the hyperactivation of Erk 1/2 signaling was reported in Cav-1 (−/−) null mouse cells and tissues [28,29,54], and in NIH-3T3 cells in which Cav-1 was downregulated by antisense [27]. The inhibitory effect of Cav-1 on Erk 1/2 may explain why, in our cellular model, it is so hard to demonstrate the direct involvement of Cav-1 in signal transduction events triggered by PrPc stimulation.
The data reported here confirm the involvement of PrPc in signal transduction events through Fyn and Erk 1/2 kinase and demonstrate caveolae to be the physical place in which this signal cascade switches on. In vitro binding experiments's results suggest that a physical interaction between PrPc and Cav-1 could be involved in these signaling events, even if this is unlikely considering the nowadaysaccepted location of PrPc and Cav-1 on the opposite side of the plasma membrane.