The Cellular Prion Protein PrP c Is Expressed in Human Enterocytes in Cell-Cell Junctional Domains*

The physiological function of PrP c , the cellular iso-form of prion protein, still remains unclear, although it has been established, in vitro or by using nerve cells, that it can homodimerize, bind copper, or interact with other proteins. Expression of PrP c was demonstrated as necessary for prion infection propagation. Considering the importance of the intestinal barrier in the process of oral prion infectivity, we have analyzed the expression of PrP c in enterocytes, which represent the major cell population of the intestinal epithelium. Our study, con-ducted both on normal human intestinal tissues and on the enterocytic cell line Caco-2/TC7, shows for the first time that PrP c is present in enterocytes. Interestingly, we found that this glycosylphosphatidylinositol-an-chored glycoprotein was localized in cholesterol-dependent raft domains of the upper lateral membranes of enterocytes, beneath tight junctions, in cell-cell junctional domains. We observed that PrP c , E-cadherin, and

The cellular prion protein (PrP c ) 1 was identified essentially because of its involvement in infectious degenerative encephalopathies (1). Indeed, the expression of PrP c has been demonstrated as required for prion infection propagation (2,3), and infectivity has been suggested to be the consequence of conformational modification of PrP c by the infectious prion protein scrapie (PrP Sc ) (1). Efforts were mainly directed at understand-ing infection and have essentially focused on nerve cells. The biological roles of PrP c still remain unclear. It has been established, however, that PrP c can homodimerize (4), interacts with other proteins, such as synapsin Ib, Grb2, Pint1, LRP/LR, and N-CAMs (5)(6)(7)(8)(9)(10)(11)(12), and binds copper (13). It has also been suggested (14) that PrP c triggers cell signaling through interaction with phosphorylated Fyn. In addition, it has been shown that multiple biochemical changes occurred in prion protein knockout mice. They included increased levels of NF-B and COX-IV and decreased levels of p53 and Cu,Zn superoxide dismutase activity, along with an increased neuronal sensitivity to oxidative stress in cultured cells isolated from these knockout mice (15).
Although oral transmission for prions diseases has been demonstrated, the site at which the infectious agent crosses the intestinal epithelium is still debated. After oral infection, the rapid accumulation of PrP Sc in Peyer's patches (16,17) led to the concept that M cells, which are present in the covering epithelium of lymphoid follicles, were responsible for the intestinal transfer of prion infectious particles toward the immune system (18,19). Indeed, M cells have a high activity of endocytosis and transport a wide variety of macromolecules and microorganisms to the mucosa-associated lymphoid tissue (20). However, enterocytes represent the major cell population of the intestinal epithelium (21), even over Peyer's patches (22), and PrP Sc has also been detected in the enterocytes of the villous epithelium of the small intestine of primates after oral infection (23).
While being a prerequisite for prion replication in nerve cells, rare studies have dealt with the expression of the normal cellular counterpart PrP c in the gastrointestinal tract, where replication might also occur. In these reports, the presence of high levels of PrP c was observed in crypts (24), in the mucus of some rare goblet cells (25), in the intestinal vascular endothelium (26), in afferent nerves of the lamina propria, and in the dispersed neuroendocrine system (27). However, the presence of PrP c in enterocytes has not been specifically investigated.
To determine whether enterocytes express PrP c , we used normal human intestinal tissues and the Caco-2/TC7 cell model that we have developed and that is morphologically and functionally similar to normal human enterocytes (28). Our results showed that PrP c is present in enterocytes and, interestingly, that this glycosylphosphatidylinositol (GPI)-anchored protein is targeted to junctional complexes, in the lateral membranes of adjacent cells, where it interacts with Src kinase. These results open new directions for evaluating the biological function of PrP c in epithelial cells.

EXPERIMENTAL PROCEDURES
Reagents-All chemicals were purchased from Sigma except where indicated. Mouse 12F10 (against peptide 142-160) and SAF32 (against * This work was supported in part by grants from INSERM and Ministère de la Recherche (Groupe d'Intérét Scientifiques PRION). The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18  . Rabbit L459 anti-human sucrase isomaltase polyclonal antibodies were used as described previously (28). Secondary CY2-CY3-, and CY5-labeled antibodies were purchased from Jackson ImmunoResearch (West Grove, PA). F-actin was labeled with phalloidin-fluorescein isothiocyanate.
Cell Treatments-For cholesterol depletion, cells were cultured for 48 h in Dulbecco's modified Eagle's medium /Glutamax1 medium supplemented with a lipoprotein-deficient serum and mevastatin (10 M). ␤-Cyclodextrin (2.5 mM) was added for the last 18 h of culture (31,32). The cellular junctional tightness was perturbed by treatment with EGTA (2 mM) for 5, 10, 15, and 30 min and monitored by apical biotin labeling. At each indicated time, cells were washed with cold PBS (Invitrogen) and immediately fixed for 30 min, at 4°C, with 4% paraformaldehyde (PFA) in PBS.
Confocal Fluorescence Microscopy-Cells cultured on filters or lamellae were fixed for 30 min with 4% PFA in PBS and permeabilized for 20 min with 0.1% Triton X-100 in PBS. Frozen human tissue samples were embedded in tissue-Tek (Dako) and cut with a cryo-microtome (Leica, Rueil Malmaison, France). Unlike cultured cells, human tissue sections were not permeabilized after treatment with 4% PFA. Cultured cells and human tissue sections were examined by confocal fluorescence microscopy (Zeiss LSM 510).
Cell Lysates, Sucrose Gradients, Immunoprecipitations, and Immunoblotting Analysis-For preparation of total cell lysates, Caco-2/TC7 and SH-SY5Y human neuroblastoma cells were washed twice with ice-cold PBS and lysed at 4°C for 15 min in a PBS buffer containing 1% Nonidet P-40, 0.5% deoxycholic acid, 0.1% SDS, protease inhibitors, and 125 nM phosphatase inhibitor okadaic acid. Crude cell extracts were frozen at Ϫ80°C before use. Protein concentrations were measured using the Bio-Rad DC protein-assay (Bio-Rad). For sucrose density centrifugation, 10 8 Caco-2/TC7 cells were homogenized on ice in 2 ml of a 10 mM Tris-HCl, pH 8, 150 mM NaCl buffer (TBS) containing 1% Triton X-100 plus an anti-protease mixture and anti-phosphatases (orthovanadate and ␤-glycerophosphate). The homogenate was adjusted to 40% sucrose by addition of 2 ml of 80% sucrose in TBS. The resulting 4 ml were covered with 4 ml of 30% sucrose and 4 ml of 5% sucrose and centrifuged for 3 h (39,000 rpm, 4°C) in an SW-41 rotor (Beckman Instruments, Gagny, France). Sequential 1-ml fractions were then collected from the top of the tube, and 160 l were used for electrophoresis to characterize the proteins present in each fraction. The remaining 840 l of the turbid fraction containing the floating detergent-insoluble membranes (fraction 4) were adjusted to 11 ml in TBS buffer and centrifuged in an SW-41 rotor (35,000 rpm, 1 h). The pellet was dissolved in TBS buffer containing 1% Nonidet P-40, anti-proteases, and anti-phosphatases and immunoprecipitated with anti-PrP c or anti-Src antibodies (3 g) coupled to protein A-Sepharose 4B (Amersham Biosciences). For SDS-PAGE, samples were boiled for 10 min in Laemmli buffer (2.5% SDS final concentration) and fractionated under reducing conditions in a 6% polyacrylamide gel. Proteins were transferred onto nitrocellulose membranes (Bio-Rad), probed, and then developed with ECL Western blotting reagents (Amersham Biosciences).
Immunoelectron Microscopy-Cryostat sections of polarized/differentiated Caco-2/TC7 monolayers were fixed with 4% PFA or 3% PFA plus 0.1% glutaraldehyde or acetone. After incubation with the 12F10 monoclonal antibody, gold (1 nm particles)-labeled goat anti-mouse IgGs (Amersham Biosciences) were used as secondary antibodies, and the resulting signal was enhanced by the Intense TM M silver enhancement kit (Amersham Biosciences). For double labeling, mouse 12F10 anti-PrP c monoclonal antibody and rabbit sc-18 anti-Src polyclonal antibodies were used. Gold (12-nm particles)-labeled donkey anti-mouse IgGs and gold (6 nm particles)-labeled donkey anti-rabbit IgGs (Jackson ImmunoResearch) were used as secondary antibodies. After alcoholgraded dehydration, sections were embedded in Epon, and ultrathin sections were analyzed (Jeol 100 CX II). and the distribution of PrP c in normal human intestinal tissues. Our results revealed that PrP c is present throughout the epithelium, in villi, where it is concentrated below the apical brush-border (Fig. 1, A and B), and in the crypt-villus junctions (Fig. 1C), in cells that express sucrase-isomaltase, a specific marker for the brush-border of enterocytes. PrP c was never observed in the apical membrane of the brush-border microvilli and, accordingly, did not co-localize with sucrase isomaltase (Fig. 1C). In fact, PrP c was found mostly concentrated beneath the tight junction-associated protein zonula occludens 1 (ZO1) (Fig. 1B). Moreover, the optimal cut incidence regularly obtained in crypt sections revealed that PrP c was concentrated in the upper part of the lateral membrane at the site of contact between two adjacent cells (Fig. 1C). We also found high levels of PrP c in the mucus of a few goblet cells (not shown) and in the vascular endothelium in the lamina propria (Fig. 1A), as reported in previous studies (24 -27) that did not detect PrP c in enterocytes.

PrP c Is Expressed in Human
We then analyzed the distribution of PrP c in enterocytic Caco-2/TC7 cells cultured on microporous filters. In confluent and polarized Caco-2/TC7 cells, PrP c was concentrated in the lateral membrane ( Fig. 2A and inset). Conversely, a high proportion of the protein remained intracellularly distributed in exponentially growing cells (Fig. 2B). However, in this last condition, a small proportion of PrP c was already located at the plasma membrane within the expanding cell clusters. Interestingly, the localization of PrP c at the lateral membrane appeared to be dependent on cell-cell contacts (Fig. 2B, arrows) as no PrP c labeling was observed, before confluence, on membranes not committed to cell-cell adhesion, i.e. around cell clusters (Fig. 2B, arrowheads). The specificity of the PrP c signals was assessed by using 12F10 antibody previously saturated with the 142-160 antigenic peptide (Fig. 2C). Moreover, SAF32 antibody gave similar PrP c signals at the lateral membrane in confluent Caco-2/TC7 cells (Fig. 2D). Western blot analysis, using 12F10 or SAF32 antibody, showed that a similar amount of PrP c was present in exponentially growing and confluent polarized cells (Fig. 2E). Several bands were obtained around 35 kDa, the specificity of which was controlled after blotting with 12F10 antibody previously saturated with the 142-160 antigenic peptide (Fig. 2E, middle panel). The amount of PrP c was compared by Western blot in confluent polarized Caco-2/TC7 cells and in SH-SY5Y nerve cells. Results reported in Fig. 2F showed that equal amounts of PrP c were present in both cell types. Moreover, the pattern of the isoforms detected with SAF32 antibody differed between Caco-2/TC7 and SH-SY5Y cells, most probably due to changes in the respective proportions of the glycosylated isoforms.
As observed in the normal human intestinal epithelium, PrP c was localized below ZO1 in the lateral membrane of Caco-2/TC7 cells (white arrows, Fig. 3A), in cell-cell contacts domains. PrP c and ZO1 never co-localized. By using immunoelectron microscopy (Fig. 3B), we showed that PrP c was associated with specialized domains of the lateral membrane, identified as junctional complexes with a thickening of the adjacent plasma membranes and their association with cytoskeleton. These junctional complexes were further shown to be adherens junctions as PrP c , E-cadherin, the main component of these junctions, and F-actin gave a merged signal in confocal microscopy at the major expression plane of PrP c (Fig. 3C).
Cholesterol-dependent Membrane Targeting of Mature PrP c in Caco-2/TC7 Cells-By cholesterol depletion (Fig. 4), using mevastatin and ␤-cyclodextrin (32), we further confirmed that PrP c is localized in rafts microdomains of enterocytes, as in other cell types (33). Although cholesterol depletion did not affect cell viability, it did induce a loss of the polarization of Caco-2/TC7 cells as shown by cell rounding and the delocalization of sucrase isomaltase on the whole cell membrane (not shown). As expected, ZO1, which is associated with the tight junctions known to be raft domains (34), was redistributed in the cytoplasm. In these conditions, PrP c was detected intracellularly rather than at the lateral membrane, demonstrating that the membrane targeting of PrP c requires the integrity of cholesterol-dependent raft microdomains.
Perturbation of Junctional Complexes Interferes with the Membrane Targeting of PrP c -We then analyzed whether perturbation of junctional complexes affects the membrane localization of PrP c . Loosening of cell-cell adhesion in confluent and polarized Caco-2/TC7 cells, following 2 mM EGTA-mediated Ca 2ϩ depletion (35), resulted in a progressive permeability of the monolayer to apically delivered biotin (data not shown) and, as expected (35), in the internalization of E-cadherin (Fig.  5A) without affecting the cell shape and the association of ZO1 to the tight junctions (Fig. 5B). In these conditions, PrP c was no longer detected at the lateral membrane (Fig. 5, A and B), and it was found exclusively intracellularly.
Membrane PrP c Interacts with Src Kinase-Src kinase is known to be targeted to cadherin-dependent junctions (36), and PrP c has been reported to interact with Fyn (14), a member of the Src family kinases, in neuronal cells. Therefore, we com-  pared the distribution of PrP c and Src kinase in confluent and polarized Caco-2/TC7 cells. Confocal microscopy showed a colocalization of both proteins at the cell-cell contacts (Fig. 6A). Immunoelectron microscopy analysis of the co-localization zones revealed that PrP c and Src were regularly in the immediate vicinity one of the other in these zones (Fig. 6B). The measured distance between PrP c and Src, 12.5 nm in majority, was compatible with the 10-nm thickness of the membrane, PrP c being anchored in the external leaflet while Src is anchored in the internal one. The observation of a potential interaction between PrP c and Src was further supported by biochemical data. Detergent-treated cell lysates were subjected to sucrose density gradient experiments, and the resulting fractions were analyzed for the presence of PrP c , Src, and E-cadherin. PrP c and E-cadherin were recovered both in the detergent-insoluble and detergent-soluble fractions, whereas Src was mainly found in the detergent-insoluble fraction (Fig. 6C). The detergent-insoluble fraction was then subjected to immu-noprecipitation experiments with antibodies directed against PrP c or against Src kinase. We found that PrP c and Src interact, whereas PrP c and E-cadherin did not. Indeed, a pool of Src was co-immunoprecipitated with anti-PrP c antibodies and vice versa (Fig. 6D). In contrast, E-cadherin, which is known to interact with Src (36), was co-immunoprecipitated with anti-Src antibodies but not with anti-PrP c antibodies.

DISCUSSION
This report is the first demonstration that PrP c is expressed in enterocytes (Fig. 1), which account for 80% of the intestinal epithelial cells (21). The amount of PrP c expressed in human Caco-2/TC7 enterocytes was similar to that of human SH-SY5Y nerve cells (Fig. 2). Interestingly, in the human intestine and in differentiated enterocytic Caco-2/TC7 cells, this GPI-anchored protein is targeted to junctional complexes of the lateral membranes of adjacent polarized cells. PrP c was found in cholesteroldependent rafts in adherens junctions where it co-localized with E-cadherin, the major component of these junctional complexes (Fig. 3). However, the observation that PrP c and E-cadherin were not co-immunoprecipitated excludes the possibility that they interact or participate in the same protein complex.
GPI-anchored proteins are considered targeted to the apical membrane domains of polarized cells. However, the localization of a GPI-anchored protein in junctional domains has already been observed for T-cadherin, a cell-cell binding signaling protein (37). Our finding in cells that endogenously express PrP c was further supported by Sarnataro et al. (38) who reported that a transfected PrP c was targeted to the basolateral membrane of Madin-Darby canine kidney and FRT epithelial cells. Moreover, we also observed that PrP c exhibited the same localization in normal human keratinocytes as in enterocytes, i.e. at the junctions between adjacent cells (data not shown).
Our results open questions about the role that PrP c can play when localized in epithelial cell-cell junctions. One could hypothesize that two PrP c molecules that face each other when anchored in two adjacent cell membranes (see Fig. 3B) interact and dimerize through their ␣-helices in a "head-to-tail conformation," as suggested by in vitro modeling of PrP c oligomerization (39). Contrary to most other junctional proteins involved in cell-cell contacts (40), PrP c is a GPI-anchored protein, meaning that it lacks a transmembrane and cytoplasmic tail and is unlikely to contribute to strong cell-cell interactions. Moreover, cell-cell junctional domains are enriched in signaling molecules such as the G␣ subunit and Src family kinases (37). Our results demonstrate that PrP c co-immunoprecipitates with a pool of Src kinase (Fig. 6) while E-cadherin co-immunoprecipitates with another one. Therefore, it may be hypothesized that PrP c plays a role in intercellular signaling and/or sensing of neighboring cells, through an interaction with Src kinases, rather than in adhesion per se. Because PrP c is anchored in the outer membrane and Src in the inner one, such a PrP c -dependent signaling does involve an intermediate factor(s) that has to be characterized. This intermediate partner of the PrP c -Src complex cannot be the transmembrane adhesion molecule E-cadherin since PrP c and E-cadherin were independently coimmunoprecipitated with two different pools of Src kinases (Fig. 6). The precise localization of PrP c in junctional complexes suggests that it could play a distinct role in epithelial cells and in nerve cells where this protein is present on the whole cell surface, binds to N-CAM (10), and was suggested to trigger signaling through the Src kinase Fyn activation (14). Combined with our results, this could suggest that PrP c plays a role in cell-cell adhesion signaling in epithelial cells and in cell-matrix adhesion signaling in nerve cells. Src kinases are known to be involved in both signaling pathways, being targeted to integrin-dependent cell-matrix and to cadherin-dependent cell-cell junctions where they phosphorylate substrates that induce adhesion turnover and actin remodeling (36).
Another important result is that alteration of the cell monolayer integrity (i.e. cholesterol depletion or loosening of cell junctions) leads to the intracellular sequestration of PrP c (Figs. 4 and 5), suggesting that the targeting of PrP c to junctional complexes can be modulated according to the physiological state of the epithelium. Whether similar modulation of the membrane targeting of PrP c also occurs in pathological conditions remains to be determined. Interestingly, Heppner et al. (18), using Caco-2/TC7 cells, reported that infectious scrapie prion units are more easily transferred through the cell monolayer when cells have acquired an M cell phenotype resulting from a co-culture with B lymphocytes. In view of our results, it may be hypothesized that the drastic cell remodeling that accompanies the enterocyte to M cell transition in this model, i.e. a disassembly of the brush border and a redistribution of villin to the cytoplasm (41), may have induced the intracellular sequestration of PrP c and favored intracellular interactions between PrP Sc and the endogenous PrP c . However, the presence and/or distribution of endogenous PrP c in untreated Caco-2/TC7 cells or in the "M cells" resulting from co-culture were not analyzed. These results do not preclude that a transfer of infectious particles, although less abundant than in M cells since it must involve an active transcellular transport, could also occur in the initial Caco-2/TC7 enterocytes.
In conclusion, the results reported here point out the signaling of cell-cell contacts as a putative role for the endogenous cellular prion protein PrP c in epithelial cells, through interaction with Src kinase. In addition, the human enterocytic Caco-2/TC7 cell line appears as an accurate model to investigate whether and how enterocytes participate in the passage of infectious particles through the intestinal epithelium, and whether these cells could represent a site of PrP c transconformation by the pathogenic prion agents.