Junctional expression of the prion protein PrP^C by brain endothelial cells: a role in trans-endothelial migration of human monocytes

Prion protein (PrP^C) is highly expressed in brain, particularly by neurons, and in several other tissues such as lymphoid organs and intestine. Conversion of PrP^C to the scrapie, protease-resistant isoform PrP^Sc, followed by its accumulation as amyloid fibrils, is required for the pathogenesis of prion diseases, including bovine spongiform encephalopathy and Creutzfeldt-Jakob disease in humans (Dormont, 2002; Prusiner, 1982). Because direct interaction between PrP^Sc and PrP^C is believed to be important for conformational conversion of PrP^C and disease progression, unraveling the unclear physiological function of PrP^C would help understand how loss or alteration of function might contribute to prion pathology.

PrP^C is a glycosylphosphatidylinositol (GPI)-anchored glycoprotein that, like most GPI proteins, is found in membrane raft microdomains. Several signaling molecules, such as transmembrane receptors and Src-family kinases, are also found in these microdomains, suggesting that PrP^C might be part of a signaling complex at the cell surface. This hypothesis was first strengthened by the observation that PrP^C was functionally coupled to the activation of Src-family kinases and ERK pathways in neuronal cell lines (Mouillet-Richard et al., 2000; Schneider et al., 2003). In addition, other signaling and/or adhesion molecules were also proposed as potential cellular partners of PrP^C: laminin (Graner et al., 2000), laminin receptor precursor (Rieger et al., 1997), heparan sulfate proteoglycans (Hundt et al., 2001; Snow et al., 1990), neural cell adhesion molecule (Schmitt-Ulms et al., 2001), neuropilin (West et al., 2005) and stress-inducible protein-1 (Zanata et al., 2002). In addition, the copper-binding activity of PrP^C and its role in copper metabolism has long been recognized, and this activity has been proposed to participate directly in neuronal resistance to oxidative stress: loss of function of PrP^C in prion diseases would therefore be associated with extensive oxidative neuronal damage (Brown and Sassoon, 2002).

Within the central nervous system, PrP^C is expressed not only in neurons, but also in glial cells (Moser et al., 1995), and the contribution of each cell type to prion replication and pathogenesis has recently been evaluated (Jeffrey et al., 2004; Prinz et al., 2004). In addition, overexpression of PrP^C was detected in vivo in brain endothelial cells in a rat model of cerebral ischemia (Shyu et al., 2005). This recent observation confirmed and extended earlier reports that PrP^C was expressed by brain endothelial cells (Deli et al., 2000), as well as by minor subtypes of peripheral microvascular (intestinal and renal capillaries) (Lemaire-Vieille et al., 2000) and macrovascular endothelial cells (umbilical vein and aorta) (Simak et al., 2002; Starke et al., 2002). However, the physiological relevance of these observations still remains elusive.

Brain endothelial cells display a unique phenotype compared with peripheral endothelial cells, characterized by the presence of intercellular tight junctions and the polarized expression of numerous membrane transporters (Engelhardt, 2003). They constitute the blood-brain barrier (BBB), which actively contributes to cerebral homeostasis and strictly controls leukocyte migration and pathogen invasion from the blood to the brain (Ballabh et al., 2004; Nassif et al., 2002). In the present study, we provide more information about the subcellular localization of PrP^C and its putative function in brain endothelial cells: we report that PrP^C is expressed by vascular endothelium in mouse brain sections and is essentially present at endothelial cell-cell junctions in freshly isolated brain microvessels, as well as in cultured brain endothelial cells of various origins (mouse, rat, human). We provide evidence that this junctional localization is probably dependent upon homophilic interactions between PrP^C molecules on adjacent cells. Moreover, we unexpectedly observed a role of PrP^C in the transmigration of human monocytes through human brain endothelial cells.

To investigate the expression of PrP^C in brain endothelial cells, we first conducted an immunohistological analysis of mouse brain sections. Blood vessels in mouse striatum, identified by the junctional expression of the tight-junction-associated zonula occludens protein-1 (ZO-1), were clearly found to express PrP^C (Fig. 1). As a control for staining specificity, only cerebral blood vessels of wild-type (WT) mice, but not of PrP^C-deficient knockout (KO) mice, were positively stained with anti-PrP^C antibodies, whereas ZO-1 staining was positive in both mouse strains. Interestingly, both microvessels (Fig. 1, insets) and larger vessels were stained with anti-PrP^C antibodies in WT brain sections.

To confirm that PrP^C cerebrovascular expression was localized to the endothelium, we performed an immunohistological analysis of freshly isolated brain microvessels from two-week-old rats as previously described (Perrière et al., 2005). By immunofluorescence confocal microscopy, expression of PrP^C was detected at cell-cell contacts between adjacent endothelial cells, similar to the junctional proteins ZO-1 and platelet endothelial cell adhesion molecule-1 (PECAM-1) (Fig. 2A). The preferred junctional expression of PrP^C by brain endothelium was further demonstrated using primary cultures of rat brain endothelial cells, using anti-ZO-1 and anti-PECAM-1 antibodies as controls for junctional staining (Fig. 2B). Similar observations were made using the rat (RBE4) and human (hCMEC/D3) brain endothelial cell lines (Fig. 3A), which we and others previously characterized extensively as in vitro models of brain endothelium (Friedrich et al., 2003; Reichel et al., 2002; Roux and Couraud, 2005; Roux et al., 1994; Weksler et al., 2005). Moreover, RT-PCR analysis unambiguously identified PrP^C transcripts in both brain endothelial cell lines, using primary rat astrocytes and human astrocytoma (U373) cells as positive controls (Fig. 3B) (Castelnau et al., 1994). Taken together, these observations show that brain endothelial cells express PrP^C and suggest that PrP^C might be part of the endothelial junctional complexes, like PECAM-1 and ZO-1 (Bazzoni and Dejana, 2004), which contribute to the integrity of endothelial monolayers.

Whereas GPI-anchored proteins are well known to be highly expressed in cholesterol-rich, raft membrane microdomains (also called caveolae-like microdomains) at the apical surface of polarized cells, some junctional proteins, including PECAM-1 (Gratzinger et al., 2003) and ZO-1, but not E-cadherin (Nusrat et al., 2000), were more recently reported to be expressed in raft microdomains. We compared the subcellular localization of the GPI-anchored protein PrP^C and the junctional protein PECAM-1 in RBE4 brain endothelial cells by cell fractionation on a 5-30% discontinuous sucrose gradient. As shown in Fig. 4A, raft microdomains were collected in the low-sucrose-density fractions (fractions 4-6), identified by the expression of caveolin-1, the prototype marker of raft/caveolae-like microdomains in endothelial cells; by contrast, the majority of membrane proteins, like the cadherin-associated junctional protein β-catenin, were recovered in the high-sucrose-density fraction (fraction 10). Co-expression of PrP^C and PECAM-1 was detected in the caveolin-1-rich fraction 4 (Fig. 4A). PrP^C appeared as a typical smear owing to the PNGase-sensitive complex pattern of N-glycosylation of its mature form, which was similarly detected in whole cell lysate (Fig. 4B) (Rudd et al., 2002). In addition, PrP^C was also detected in the high-sucrose-density fraction 10, with a distinct banding pattern that probably corresponds to differentially glycosylated forms of PrP^C (Fig. 3A) (Sarnataro et al., 2002). These results suggest that most of the mature form of PrP^C is co-localized with PECAM-1 in raft/caveolae-like microdomains at the cell-cell junctions of brain endothelial cells in culture.

To investigate whether PrP^C might be involved in interactions between adjacent cells, non-confluent RBE4 cell cultures were analyzed for PrP^C localization by indirect immunofluorescence and compared with confluent cultures (Fig. 3A). Whereas PrP^C was diffusely expressed over the entire cell surface in isolated cells, translocation of PrP^C to cell-cell contacts was clearly observed between two individual cells (Fig. 5A, arrow). A similar observation was made regarding PECAM-1 expression (Fig. 5B, arrow), which is stabilized at cell-cell contacts by homophilic interactions between adjacent endothelial cells (Sun et al., 1996). To establish whether PrP^C junctional expression might similarly be dependent upon homophilic interactions between adjacent cells, brain endothelial cells were isolated from wild-type (WT) and PrP^C-deficient (KO) mice, grown to confluence in mixed cultures and analyzed for PrP^C distribution. Adherens junction complexes could form at the interface between WT and KO endothelial cells, as shown by junctional expression of the cadherin-associated protein β-catenin in endothelial cells of both origins (Fig. 6A). In contrast to β-catenin, PrP^C was only observed at cell-cell contacts between two WT cells (Fig. 6B, arrow), but not at contacts between WT and KO cells (Fig. 6B, arrowhead). Altogether, these results strongly suggest that PrP^C is involved in homophilic interactions between adjacent cells and that these interactions are required for its junctional localization in brain endothelial cells.

It is well established that homophilic interactions between PECAM-1 molecules expressed at the surface of endothelial cells and monocytes directly contribute to the trans-endothelial migration of monocytes (Muller et al., 1993). We therefore investigated whether PrP^C might also be involved in this process, using the human brain endothelial cell line hCMEC/D3 (Weksler et al., 2005) and U937 human monocytic cells. By FACS analysis (Fig. 7), we first confirmed that U937 cells express PrP^C, PECAM-1 and VLA-4, the latter being the integrin receptor for vascular cell adhesion molecule-1 (VCAM-1) involved in monocyte adhesion to endothelial cells.

We previously documented that hCMEC/D3 cells express, either constitutively or upon treatment by the inflammatory cytokines tumor necrosis factor α (TNF-α) and interferon γ (IFN-γ), several surface molecules (PECAM-1, ICAM-1 and VCAM-1) known to be involved in adhesion and/or trans-endothelial migration of leukocytes (Weksler et al., 2005). Here, we present evidence that adhesion of U937 monocytic cells to hCMEC/D3 cells was strongly induced by endothelial pre-treatment with TNF-α and IFN-γ (Fig. 8A), and was drastically inhibited (89±2%) in the presence of anti-VLA-4 blocking antibodies (Fig. 8B). By contrast, antibodies to junctional proteins not involved in leukocyte adhesion (e.g. PECAM-1) (Allport et al., 2000), had no inhibitory effect compared with irrelevant anti-transferrin receptor (CD71) antibodies. In these conditions, none of the four anti-PrP^C antibodies tested showed any inhibitory activity (Fig. 8B).

We then observed that trans-endothelial migration of U937 cells through hCMEC/D3 monolayers was, as expected, dependent upon TNF-α and IFN-γ pre-treatment of endothelial cells and the presence of a concentration gradient of the chemokine stromal-derived factor-1α (SDF-1α) (Fig. 9A). Unexpectedly, all four anti-PrP^C antibodies tested (SAF32, SAF34, SAF61 and 6H4 antibodies), when pre-incubated with U937 and hCMEC/D3 cells for 1 hour prior to the assay, inhibited transmigration of U937 cells to a similar extent as blocking anti-PECAM-1 antibodies (49±8%) used as control: 46±10%, 66±7%, 59±8% and 62±2% inhibition was observed respectively in the presence of SAF32, SAF34, SAF61 and 6H4 (Fig. 9B). In a control experiment, anti-PrP^C (SAF32, SAF34) or anti-CD71 antibodies did not affect U937 cell migration through endothelial-cell-free Transwell filters coated in fibronectin and collagen, indicating that neither viability nor migration capacity of U937 cells was affected by antibody pretreatment (Fig. 9C).

We further assessed the contribution of PrP^C to leukocyte transendothelial migration using freshly isolated human monocytes known to express PrP^C constitutively (Dürig et al., 2000). Transmigration of freshly isolated monocytes through hCMEC/D3 monolayers was observed at a higher level under a gradient of monocyte chemoattractant protein-1 (MCP-1) (5.5±0.3%) than under a gradient of SDF-1α (1.3±0.3%) (data not shown), as previously reported (Bleul et al., 1996; Cambien et al., 2001). This process was significantly inhibited following pre-incubation of both cell types with the anti-PrP^C antibodies SAF 34 (38±4%) as well as by the control anti-PECAM-1 antibodies (68±3%) (Fig. 10A), thus confirming our data regarding U937 trans-endothelial migration. No additive effect between anti-PrP^C and anti-PECAM-1 antibodies was detected (data not shown). As reported in Fig. 11, transmigration inhibition was also observed when only one cell type was pretreated with anti-PrP^C antibodies or with anti-PECAM-1 antibodies, further suggesting that PrP^C expressed on both cell types is involved in trans-endothelial migration of monocytes. As control, antibody pre-treatments did not affect the viability of the monocytes or their migration capacity through endothelial-cell-free Transwell filters coated in fibronectin and collagen (Fig. 10B). These results provide the first indication that PrP^C contributes to the trans-endothelial migration of monocytes.

We present evidence that brain endothelial cells express PrP^C at intercellular junctions, using freshly isolated rat brain microvessels, primary cultures of mouse and rat brain endothelial cells, and rat (RBE4) and human (hCMEC/D3) brain endothelial cell lines. We show that PrP^C is co-localized with PECAM-1 in raft/caveolae-like membrane microdomains, is largely restricted to cell-cell contacts and is probably involved in PrP^C-PrP^C homophilic interactions between adjacent cells. Moreover, we believe that the most important result of the present study is the identification of a previously unsuspected functional role of PrP^C in monocyte transmigration through human brain endothelial cells.

Like other GPI-anchored proteins, PrP^C has been localized to raft/caveolae-like membrane microdomains in most cell types, including neurons and cells of the immune system (Massimino et al., 2002). However, in contrast to the majority of GPI-anchored proteins, which are sorted to the apical membrane of polarized cells, raft-associated PrP^C was shown to be restricted to the basolateral membrane (Sarnataro et al., 2002) and/or intercellular junctions of polarized epithelial cells (Morel et al., 2004). Our results extend these observations to brain endothelial cells, which constitute a polarized physiological barrier in vivo known as the BBB. Our observation that PrP^C is co-localized with PECAM-1 at intercellular junctions and in raft/caveolae-like membrane microdomains of brain endothelial cells is supported by a recent report that PECAM-1 is expressed in endothelial raft microdomains (Gratzinger et al., 2003). Together with the previous demonstration that tight junction proteins are also found in raft/caveolae-like membrane microdomains in epithelial cells (Nusrat et al., 2000), our results suggest that PrP^C is localized in junctional microdomains that are involved in the regulation of cell-cell adhesion in epithelial and brain endothelial cells.

Localization of PrP^C to cell-cell contacts in brain endothelial cells was shown here to be dependent on the expression of PrP^C by two adjacent cells. This conclusion is strengthened by two parallel observations: (1) in non-confluent cultures, PrP^C expression was diffuse at the cellular surface, but concentrated at occasional cell-cell contacts; and (2) in confluent mixed cultures of PrP^C-expressing (WT) and PrP^C-deficient (KO) brain endothelial cells, PrP^C was strictly localized to contacts between two WT cells, although adhesion complexes did form between WT and KO endothelial cells, as shown by β-catenin junctional expression. Taken together, these results strongly suggest that PrP^C localization to cell-cell contacts in brain endothelial cells is controlled by homophilic interactions between adjacent cells. As previously proposed for epithelial junctions (Morel et al., 2004), PrP^C-PrP^C interaction on two adjacent cells might involve a `head-to-tail conformation' similar to that described for PrP^C oligomerization by in vitro modeling (Knaus et al., 2001). In addition, our observations are highly reminiscent of those previously reported with several junctional adhesion molecules, such as PECAM-1, VE-cadherin or JAM-A expressed by endothelial cells (Bazzoni et al., 2000; Lampugnani et al., 1995; Sun et al., 1996), further supporting our hypothesis of a role for PrP^C as an intercellular adhesion molecule in endothelial cells.

However, over the past ten years, alternative hypotheses have been proposed about the putative physiological function of PrP^C. In particular, there is increasing evidence supporting a functional role of PrP^C in copper binding and metabolism, and in the modulation of anti-oxidant enzyme Cu/Zn superoxide dismutase activity (Brown and Sassoon, 2002). In addition, a role has also been proposed for PrP^C in cellular adhesion to extracellular matrix and in maintenance of integrity of the physiological intestinal barrier (Graner et al., 2000; Mattei et al., 2004; Morel et al., 2004; Rieger et al., 1997).

In the present study, we report an unsuspected role of PrP^C in the trans-endothelial migration of U937 human monocytic cells as well as freshly isolated human monocytes. Indeed, anti-PrP^C antibodies, like blocking anti-PECAM-1 antibodies used as control, were shown to prevent the trans-endothelial migration of U937 monocytic cells or monocytes. By contrast, neither anti-PrP^C antibodies nor anti-PECAM-1 antibodies affected the adhesion of monocytic cells. It is interesting to note that the present study actually points to several similarities between PrP^C and PECAM-1 localization and function: both proteins (1) co-localize to raft/caveolae-like membrane microdomains; (2) are concentrated at intercellular junctions of endothelial cells; and (3) are involved in monocyte transmigration through brain endothelium, inasmuch as their engagement with specific antibodies prevents monocyte transmigration. These data strongly suggest the involvement of PrP^C homophilic interactions between adjacent endothelial cells, as well as between endothelial cells and monocytes, which is again similar to PECAM-1 (Newman, 1997). On the basis of these parallel observations between PrP^C and PECAM-1, it could be proposed that PrP^C might interact with PECAM-1, as it does with other membrane proteins such as neural cell adhesion molecule (Schmitt-Ulms et al., 2001) or laminin receptor precursor (Rieger et al., 1997), and that anti-PrP^C antibodies might indirectly prevent monocyte migration by targeting a junctional PrP^C–PECAM-1 complex. However, we failed to demonstrate such an interaction by co-immunoprecipitation assays (data not shown), strengthening the conclusion that PrP^C might be directly involved in controlling the trans-endothelial migration of monocytes.

Several recent studies on leukocyte trans-endothelial migration described two distinct mechanisms for this process, involving either a transcellular or a paracellular route, which are proposed to be preferentially used by peripheral blood mononuclear cells and polymorphonuclear leukocytes, respectively (Barreiro et al., 2002; Carman and Springer, 2004). Because the transcellular pathway was very recently shown to depend on raft/caveolae-like membrane microdomains (Millan et al., 2006), a contribution of PrP^C to this process might be suggested. However, our data with U937 monocytic cells and freshly isolated monocytes, together with the proposition that monocytes preferentially transmigrate through a PECAM-1-dependent paracellular route, particularly upon chemokine treatment (Mamdouh et al., 2003; Muller et al., 1993; Nieminen et al., 2006), further suggest the involvement of PrP^C in a paracellular process. Moreover, our observation that anti-PrP^C and anti-PECAM-1 antibodies have no additive effect suggests that PrP^C and PECAM-1 might be involved in the same step of monocyte trans-endothelial migration. Additional experiments will be required to discriminate between these hypotheses and to unravel the precise mechanism of PrP^C contribution to monocyte transmigration.

In conclusion, the results of the present study identify PrP^C as a junctional protein in brain endothelial cells, probably involved in homophilic interactions between adjacent cells, and strongly support a role for PrP^C in the trans-endothelial migration of monocytes.

All reagents were purchased from Sigma unless otherwise stated. Dispase II was obtained from Roche Molecular Biochemicals. Collagenase type 2 was purchased from Worthington Biochemical Corporation. Endothelial Basal Medium (EBM-2) and EGM-2 BulletKit were obtained from Cambrex. RPMI 1640, αMEM and F10 media, fetal calf serum (FCS), basic fibroblast growth factor (bFGF, human, recombinant), were purchased from Invitrogen, and bovine-plasma-derived serum (BPDS) was purchased from First Link. Rat tail collagen type I was purchased from BD Biosciences, and collagen type IV from Nunc Labttek Gibco. Peptide: N-Glycosidase F (PNGase F) was from Ozyme. Tumor necrosis factor α (TNF-α), interferon γ (IFN-γ), stromal-derived factor-1α (SDF-1α, CXCL12) and monocyte chemoattractant protein-1 (MCP-1, CCL-2) were from Euromedex.

Mouse monoclonal antibodies specific to PrP^C (SAF32, IgG2b; SAF34, IgG2b; SAF61, IgG2a) were purchased from SPI-BIO, mouse monoclonal antibody 6H4 (IgG1) was from Prionics. Antibodies to zonula occludens-1 (ZO-1; rabbit polyclonal) were purchased from Zymed, antibodies to β-catenin (rabbit polyclonal) and to VLA-4 (CD49d; mouse monoclonal, IgG1) were from Euromedex, mouse monoclonal antibody HEC-7 anti-PECAM-1 (IgG1) was from Perbio Science France, and mouse monoclonal antibody 3A12 anti-PECAM-1 (IgG1) was from BD Biosciences. Antibodies to platelet endothelial cell adhesion molecule-1 (PECAM-1; M20, goat polyclonal) and to caveolin-1 (mouse monoclonal, IgG1) were from TEBU. Antibodies to transferrin receptor (CD71; mouse monoclonal, IgG1) were from R&D Systems. Peroxidase-conjugated anti-mouse or anti-rabbit IgG antibodies and ECL reagents were from Amersham, Cy2- or Cy3-conjugated anti-mouse or anti-rabbit antibodies from Jackson ImmunoResearch Laboratories, and Alexa Fluor488-conjugated anti-mouse antibodies were from Invitrogen.

Rats (OFA) and wild-type (WT) mice (C57BL/6J) were purchased from Charles River laboratories, PrP^–/– mice (C57BL/6Jx129) (Bueler et al., 1992) were from CDTA. PrP^–/– mice were inbred for at least nine generations onto the C57BL/6J background and used in experiments with age-matched wild-type C57BL/6J mice as control.

All animals were treated according to protocols evaluated and approved by the local ethical committee of INSERM, France. Isolation of brain microvessels and primary culture of endothelial cells were performed as previously described (Perrière et al., 2005) for rat and, with minor modifications, for mouse. Briefly, 8-10 8-week-old mice were used for brain microvessel isolation. Mouse cerebral cortices free of meninges were digested in a mixture of collagenase (270 U/ml), dispase (0.1%) and DNAse (10 U/ml) for 30 minutes at 37°C, centrifuged and then incubated for another 10 minutes at 37°C with the same mixture. Endothelial cells were grown in EBM-2 medium supplemented with 20% plasma-derived bovine serum, Hepes (1 mM), glutamine (2 mM), bFGF (2 ng/ml for rat cells, 5 ng/ml for mouse cells), antibiotics (penicillin 100 U/ml, streptomycin 100 μg/ml) and hydrocortisone (1.4 μM); in addition, the medium was supplemented with puromycin (4 μg/ml for rat cells or 1 μg/ml for mouse cells) for the first 4 days in culture as previously described (Perriere et al., 2005). Microvessels were seeded on Thermanox coverslips coated with collagen type IV (0.1 μg/ml) for direct immunofluorescence assay. Microvessels were either fixed 1 hour after seeding or maintained in culture until a confluent monolayer of endothelial cells was obtained.

This study used the RBE4 rat brain endothelial cell line that has been extensively characterized by us and others (Durieu-Trautmann et al., 1994; Etienne-Manneville et al., 2000; Roux et al., 1994) and the hCMEC/D3 human brain endothelial cell line we recently described (Weksler et al., 2005). Briefly, RBE4 cells were grown on plates coated with type I collagen in αMEM/F10 medium supplemented with 10% FCS, bFGF and G418; and hCMEC/D3 cells were grown on plates coated with type I collagen in EGM-2 medium. The U937 human monocytic cell line was obtained from P. Lutz (IPBS-CNRS, Toulouse, France) and grown in RPMI 1640 medium supplemented with 10% FCS.

Peripheral blood mononuclear cells (PBMCs) were isolated from buffy coat units, obtained from consenting donors, by gradient centrifugation on Ficoll-Paque (Amersham) using standard protocols. Monocytes were then isolated by magnetic negative cell sorting (Dynal Monocyte Negative Isolation Kit from Invitrogen), according to the manufacturer's protocol, after a 2-hour adhesion step on uncoated flasks at 37°C, 5% CO₂ in RPMI 1640 medium supplemented with 10% FCS.

U373-MG human astrocytoma cells were obtained from the ATCC repository (LGC Promochem) and grown on DMEM with 4.5 g glucose (Invitrogen) supplemented with 10% FCS. Primary cultures of rat astrocytes were prepared as previously described (Perriere et al., 2005) and grown on DMEM with 4.5 g glucose supplemented with 10% FCS.

Brain tissue was obtained from 5-month-old wild-type (WT) or PrP^–/– C57BL/6J mice. According to protocols evaluated and approved by the local ethical committee of INSERM, France, the mice were deeply anesthetized with pentobarbital and perfused transcardially with 4% paraformaldehyde in 0.1 M Na₂HPO₄/NaH₂PO₄ buffer, pH 7.5. After perfusion, the brains were collected, postfixed and cut with a vibratome. Sections (30 μm) were then processed for immunohistochemistry. After H₂O₂ treatment, non-specific antibody binding was blocked with 3% normal goat serum and incubated with primary antibodies directed to ZO-1 (5 μg/ml) or PrP^C: SAF 32 (1 μg/ml) plus SAF 61 (1 μg/ml) for 16 hours at 4°C. After washing with Tris buffer saline (TBS), sections were then incubated for 45 minutes at 25°C with Cy3-conjugated anti-rabbit and Alexa Fluor 488-conjugated anti-mouse antibodies (1/400) in 3% normal goat serum. Negative controls were obtained by omission of primary antibodies. Sections were then rinsed three times in TBS and mounted under coverslips using Vectashield with DAPI (Vector Laboratories). Immunofluorescence images were collected with a confocal fluorescence microscope (LEICA TCS SP2).

Immunofluorescence analyses were conducted as previously described (Cazaubon et al., 1997) with brain microvessels, primary cultures of brain endothelial cells and endothelial cell lines seeded on Thermanox coverslips. In all cases, cells were fixed with paraformaldehyde (4%) in phosphate-buffered saline (PBS) for 10 minutes, protected with glycine 0.1 M for 15 minutes and blocked with PBS containing bovine serum albumin (BSA, 2%) for 1 hour. Alternatively, for detection of intracellular antigens, cells were permeabilized with 0.05% saponin for 1 hour. The cells were then incubated for 16 hours at 4°C with monoclonal antibodies specific to PrP^C (2 μg/ml), PECAM-1 (2 μg/ml) or ZO-1 (1 μg/ml). Cy2- or Cy3-conjugated anti-rabbit or anti-mouse antibodies were used as secondary antibodies. Negative controls were obtained by omission of primary antibodies. Immunofluorescence images were collected with a confocal fluorescence microscope (LEICA TCS SP2).

Total RNA was isolated from confluent monolayers of RBE4, hCMEC/D3 and U373 cells, as well as primary cultures of rat astrocytes, using Trizol (Invitrogen) according to the manufacturer's instructions. PCR amplification was carried out from 10 μg total RNA after reverse transcription using SuperScript II Reverse Transcriptase (Invitrogen) according to standard protocols. Primers for PCR amplification of rat PrP^C were 5′-GTGCACGACTGTGTCAAT-3′ (forward) and 5′-CTCCTCATCCCACGATCAGG-3′ (reverse), yielding an expected product size of 244 bp. Primers for PCR amplification of human PrP^C were 5′-GTGCACGACTGCGTCAAT-3′ (forward) and 5′-CCTTCCTCATCCCACTATCAGG-3′ (reverse), yielding an expected product size of 243 bp. Standardization was performed using primers 5′-CCTGCTGGATTACATTAAAGCGCTG-3′ (forward) and 5′-CCTGAAGTGCTCATTATAGTCAAGG-3′ (reverse) for rat hypoxanthine-guanine phosphoribosyl transferase (HPRT) or primers 5′-GGAGAAGGCTGGGGC-3′ (forward) and 5′-GATGGCATGGACTGTGG-3′ (reverse) for human glyceraldehyde-3-phosphate dehydrogenase (GAPDH) PCR amplifications. All primers were synthesized by Eurogentec. PCR conditions were as follows: denaturation at 94°C for 5 minutes; 30 cycles of 94°C for 1 minute; 50°C (rat) or 60°C (human) for 1 minute; 72°C for 1 minute; and an extension step of 72°C for 7 minutes.

Fluorescence-assisted cell sorting (FACS) assay was performed following standard protocols. Briefly, U937 cells were washed, saturated with 2% BSA and incubated for 10 minutes at 25°C with the indicated antibodies at 2 μg/ml, followed by FITC-conjugated anti-mouse secondary antibody; omission of first antibodies was used as control. Acquisition was performed on an Epics XL cytometer (BD Biosciences).

RBE4 or hCMEC/D3 cells grown to confluence were used to prepare caveolae fractions as previously described (Teixeira et al., 1999). Briefly, cells were washed in ice-cold PBS and scraped into 2 ml of 500 mM sodium carbonate, pH 11.0. Homogenization was carried out using a loose-fitting Dounce homogenizer (20 strokes) and a sonicator (three bursts of 20 seconds). The homogenate was then adjusted to 40% sucrose by addition of 2 ml 80% sucrose solution prepared in 25 mM MES pH 6.5, 0.15 M NaCl (MBS buffer), placed in the bottom of an ultracentifuge tube and a 5-30% discontinuous sucrose gradient was formed above. After centrifugation at 200,000 g for 16-20 hours in an SW41 rotor (Beckman Instruments), a light-scattering band confined to the 5-30% sucrose interface was observed that contained caveolin-1. Gradient fractions were analyzed by western blotting as previously described (Cazaubon et al., 1997), using the following antibodies: anti-caveolin-1 (0.5 μg/ml), anti-PrP^C (1 μg/ml), anti-PECAM-1 (1 μg/ml) and anti-β-catenin (1 μg/ml).

RBE4 cells were scrapped onto 500 μl of 1% NP40 detergent lysis buffer containing 10 mM Tris-HCl, 150 mM NaCl and protease inhibitors (leupeptine, aprotinin and pepstatin at 2 μg/ml each). Lysates were then centrifuged to remove nuclei and cell debris, and digested with PNGase F according to manufacturer's instructions. Briefly, cleared lysates (500 μl) were incubated with denaturing buffer for 10 minutes at 100°C, followed by incubation with 2500 units of enzyme in appropriate buffer for 1 hour at 37°C. As a control, lysates were incubated in parallel without enzyme. Samples (10 μl) were analysed by western blotting, as described above, using anti-PrP^C antibodies.

U937 human monocytic cells were fluorescently labeled with 10 μM 5-chloromethylfluorescein diacetate (CMFDA) for 30 minutes at 37°C and were subsequently washed in EBM-2 medium with 0.1% BSA. Fluorescently labeled U937 cells (10⁵ per well) were incubated with non-stimulated human brain endothelial cells hCMEC/D3 or hCMEC/D3 cells pre-treated for 24 hours with IFN-γ (200 U/ml) plus TNF-α (200 U/ml), and were allowed to adhere for 30 minutes at 37°C. When indicated, U937 cells and hCMEC/D3 monolayers were pre-treated for 1 hour at 37°C with various blocking or control antibodies (20 μg/ml) before the adhesion assay. After the incubation, non-adherent U937 cells were removed by washing with PBS and the adherent cells were hypotonically lysed. The proportion of adherent U937 cells was determined by quantification of the amount of fluorescence released using a FUSION fluorescent plate reader (Perkin Elmer) with an excitation wavelength of 492 nm and an emission wavelength of 517 nm.

hCMEC/D3 cells (10⁵ cells/cm²) were seeded onto type I collagen-coated 6.5 mm Transwell culture inserts with a pore size of 3 μm (Corning) and grown for 3 days in 5% CO₂ at 37°C. Prior to the assays, monolayers were stimulated with 200 U/ml IFN-γ plus 200 U/ml TNF-α for 24 hours and washed twice with EBM-2 medium with 0.1% BSA.

U937 monocytic cells or freshly isolated human monocytes were fluorescently labeled with 10 μM of CMFDA at 37°C for respectively 30 minutes or 45 minutes. Fluorescently labeled U937 cells (10⁶ cells/ml in 100 μl of EBM-2 medium) or freshly isolated monocytes (2.5×10⁵ cells/ml in 100 μl of RPMI medium) were added to the apical chamber. A chemotactic gradient was created by addition to the lower chamber of SDF-1α (100 ng/ml) or MCP-1 (25 ng/ml), respectively. U937 cells or monocytes were allowed to migrate at 37°C for 16-18 hours or 3 hours, respectively.

When indicated, U937 cells, monocytes and/or hCMEC/D3 monolayers were preincubated for 1 hour at 37°C with various blocking or control antibodies (20 μg/ml) before the migration assay. Transmigrated, fluorescently labeled U937 cells or monocytes were then recovered from the lower chambers and counted by flow cytometry, using an Epics XL cytometer (BD Biosciences).

Data are expressed as means ± s.e.m. Student's t or Mann-Whitney U tests were performed for statistical data analysis. For statistical significance (P value), see legend of respective figure.

We thank F. Roux for helpful discussions, P. Bourdoncle for technical assistance with confocal microscopy, and S. Bourdoulous, C. Nahmias and C. Rampon for critical reading of the manuscript. This work was supported by the GIS `Infection à Prion' programme, the Centre National de la Recherche Scientifique, the Institut National de la Santé et de la Recherche Médicale, the Université René Descartes-Paris 5 and the Ministère de la Jeunesse, de l'Education nationale et de la Recherche. P.V. was supported by a grant from the Fundação para a Ciência e Tecnologia, co-financed by the POCI2010 program (Portugal) and the European Social Fund.