LRP1 controls biosynthetic and endocytic trafficking of neuronal prion protein

PrP^C is endocytosed by mechanisms that differ according to the cell type studied, presumably reflecting differences in the expression of partner proteins that determine the trafficking of PrP^C that lacks a cytoplasmic domain into which endocytic trafficking motifs could be embedded. On N2a cells, PrP^C is endocytosed constitutively by means of coated pits (Shyng et al., 1994; Sunyach et al., 2003); whereas cells that do not endogenously express PrP^C but transfected with a vector encoding PrP^C can require 100 μM Cu²⁺ to initiate coated pit endocytosis (Taylor and Hooper, 2007) or else slowly internalise PrP^C by non-coated pit mechanisms (Peters et al., 2003).

The cell type of prime relevance to mammalian prion biology is the neuron. PrP^C occupies lipid rafts that, on neurons, are distinctive, differing both chemically and spatially from those containing the major neuronal glycosylphosphatidylinositol (GPI)-anchored protein, Thy-1 (Brügger et al., 2004; Madore et al., 1999; Sunyach et al., 2003). PrP^C exits its rafts to traverse detergent-soluble membrane as it enters coated pits for endocytosis and then recycles back to the surface via perinuclear sorting compartments, the entire cycle taking 6-10 minutes (Morris et al., 2006; Sunyach et al., 2003).

PrP^C has two distinct domains (Fig. 1): a GPI-anchored, α-helical C-terminal domain (amino acids 126-231) and an unstructured N-terminal domain (residues 23-125) that is necessary and sufficient for coated pit endocytosis (Shyng et al., 1995a; Sunyach et al., 2003), with the immediate N-terminal basic motif (NH₂-KKRPKP-) being essential for this process (Sunyach et al., 2003).

Binding through a basic motif to low-density lipoprotein receptor-related protein 1 (LRP1) is the mechanism by which the GPI-anchored urokinase plasminogen activator receptor is internalised by means of coated pits (Horn et al., 1998). This prompted us to investigate whether LRP1 is the endocytic partner for neuronal PrP^C (Morris et al., 2006). Recently, Taylor and Hooper (Taylor and Hooper, 2007) reported that LRP1 is required for the Cu²⁺-dependent endocytosis of exogenous PrP^C on SH-SY5Y cells.

LRP1 is a massive protein (∼600 kDa) that is proteolytically nicked during biosynthesis to give two stably associated polypeptides: an 85-kDa membrane-spanning C-terminal fragment and a 515-kDa extracellular N-terminal chain (Fig. 1). LRP1 is expressed abundantly on neurons (Bu et al., 1994; Moestrup et al., 1992), where its fundamental role is the uptake of cholesterol and fatty acids by these cells whose vast membrane processes require the import of glial-derived sterol and lipid for growth and synapse formation (Mauch et al., 2001). In addition, LRP1 binds to >30 ligands extracellularly, and its cytoplasmic domain binds to endocytic and scaffold adaptors that link the receptor to other membrane proteins, including Alzheimer's precursor protein (APP) (Herz and Chen, 2006; Zerbinatti et al., 2004). The LRP1 knockout mouse is embryonic lethal, and even a partial knockout in adult brain results in slow lethality (May et al., 2004).

LRP family members have a specific chaperone, receptor-associated protein (RAP), that masks their ligand-binding sites during biosynthesis. Recombinant RAP added externally to cells inhibits the binding and endocytosis of LRP ligands (Iadonato et al., 1993; Moestrup et al., 1993b).

PrP^C has also been ascribed numerous binding partners (Linden et al., 2007). Any attempt to add to its list of molecular associates needs to demonstrate both specificity and biological relevance. One reported receptor for PrP is the laminin precursor protein (confusingly also abbreviated `LRP') (Gauczynski et al., 2001) that functions both as a cytoplasmic ribosomal subunit and a broad-specificity adhesive protein on the cell surface (Kazmin et al., 2003).

We report here that LRP1 associates with endogenously expressed PrP^C on the neuronal cell surface to internalise it and, furthermore, that LRP1 binds to PrP^C in biosynthetic compartments to assist its trafficking to the neuronal surface.

Sensory neurons are in this study used because in vivo they uniquely have axons and no dendrites; thus, lacking any synaptic input, they are not denervated upon being dissociated and so survive, remain differentiated and regenerate their axons when introduced into culture from adult tissue (Scott, 1977). In dissociated culture, they extend axons over laminin-coated substrates and flattened satellite cells, while their large cell bodies float, usually without contacting other cells, into the medium above. Fig. 2A,B shows immunolabelling for PrP^C at the level of the substrate, and the zone 6 μm above this, where the large neuronal cell bodies are suitable for analysis of the endocytic trafficking of fluorescently labelled surface proteins.

Fig. 2C is an immunofluorescent view of the relative distribution of LRP1 and PrP^C on the surface of a cultured adult sensory neuron. Partial colocalisation of the two proteins was found, with 40% of PrP^C overlapping LRP1 (Table 1; conversely, 43.2±14.7% of surface LRP1 colocalised with PrP^C). Somewhat higher colocalisation (54.2±21.6%) was seen between PrP^C and the lipid `raft' marker ganglioside GM1 labelled with the cholera toxin B chain (Ctx; Fig. 2D). Antibodies against other neuronally expressed LRP family members (LRP1b, LRP6, LRP8 and VLDLR) failed to react with sensory neurons assessed by immunoblotting or immunohistochemistry.

To gain a more dynamic view of the proximity of the two proteins during endocytosis, surface PrP^C was prelabelled at 10°C with monovalent 2S Alexa-Fluor-594–Fab to PrP^C; the cells were then washed to remove unbound Fab and placed at 37°C. The pulse-labelled PrP^C was endocytosed rapidly over the next 2 minutes (Fig. 2E), a period during which labelled PrP^C remaining on the cell surface colocalised increasingly with LRP1 (Table 1).

The proximity of PrP^C to LRP1 during endocytosis was analysed at higher resolution by examining, by transmission electron microscopy (EM), neurons that were prelabelled for PrP^C with 5 nm Fab gold, and for LRP1 by its ligand, activated α₂macroglobulin (α₂M^*), attached to 10 nm gold. Also, for comparison with PrP^C, the transferrin (Tf) receptor was labelled with Tf coupled to 10 nm gold. After endocytosis for 1.5 minutes at 37°C, within the restricted confines of coated pits, 66.2% of PrP^C label was within 50 nm of α₂M^* (Fig. 2F), but only 24.3% was within 50 nm of the Tf label. The converse selective association of the ligands with PrP^C also held: 84.7% of α₂M^* was within 50 nm of PrP^C, compared with only 35.4% of Tf. Thus, within this morphologically identifiable endocytic compartment, PrP^C was more closely associated with an LRP1 ligand than with a ligand for an independently endocytosed receptor.

To assess inhibition of endocytosis, neurons were prelabelled for surface PrP^C with 2S Alexa 594-Fab; as a positive control, their Tf receptors were labelled with Alexa-Fluor-488–Tf. Preincubation with recombinant RAP inhibited endocytosis of PrP^C but not Tf (Fig. 3A,B; Table 2).

To circumvent long-term disruption of neuronal metabolism by stable inhibition of LRP1 expression, transient knockdown of LRP1 was achieved using the antennapedia peptide penetratin-1 to deliver one of three siRNAs specific for LRP1 (siRNA^LRP1.1, siRNA^LRP1.2, siRNA^LRP1.3) or a control (siRNA^Con). The siRNA^LRP1.1 produced rapid knockdown of LRP1 protein (Fig. 3E-H) that was followed by several-fold overexpression by 72 hours (data not shown). For inhibition of endocytosis, 90 minutes of treatment of sensory neurons with the siRNA-penetratin conjugate was used, as each of the three siRNAs directed against LRP1 lowered expression of LRP1 protein by 30-50% within this period without affecting the level of surface PrP^C, as determined by quantitative analysis of immunofluorescent binding to surface PrP^C and total cellular LRP1 (Fig. 3E,F). For example, in an experiment in which siRNA^LRP1.1 lowered expression of LRP1 to 64.0% of the control level (P=0.003), the level of surface PrP^C on the two sets of cells was identical (P=0.98; supplementary material Table S2c). With this brief downregulation of LRP1, all three siRNA^LRP1 constructs significantly inhibited the endocytosis of PrP^C but not Tf (Table 2; Fig. 3C,D).

Exposure of neurons to siRNA^LRP1.1 for 2-4 hours reduced LRP1 levels by ∼75% (Fig. 3G,H). This was accompanied by a marked reduction in surface PrP^C, to 43.9±8.1% of siRNA^Con-treated cells within 4 hours, as determined by immunolabelling of live cells at 10°C (Fig. 4A-D; P<0.001; supplementary material Table S3a). This treatment increased the surface level of GPI-anchored Thy-1 (Fig. 4E,F) by 2.2±1.2 fold compared with controls (P=0.04; supplementary material Table S3b). The increased Thy-1 was evident in immunoblots, as was an even larger increase in APP (Fig. 5).

Sustained reduction of LRP1 levels did not affect the ability of the residual surface LRP1 to be endocytosed, as shown by the complete internalisation within 2 minutes of much reduced levels of its ligand, α₂M^* (Fig. 4G,H). Similarly, the reduced level of surface PrP^C was >80% endocytosed within 2 minutes (Fig. 4I,J), giving a reduced amount of PrP^C endocytosed without any discernible reduction in its rate of endocytosis.

However, siRNA^LRP1 inhibition for 4 hours was accompanied not only by a loss in surface PrP^C but also by a 3.5±0.9 (P<0.01) fold increase in intracellular PrP^C (Fig. 4A-D). Intracellular Thy-1 was not elevated (Fig. 4E,F).

Most of the increased intracellular PrP^C colocalised with early biosynthetic compartments (Table 3; Fig. 6). The chaperone BiP was used as a marker of the endoplasmic reticulum (ER); α-mannosidase II as a marker for the mid-Golgi, and syntaxin 6 as a marker for the trans-Golgi. The levels of the latter two did not differ between siRNA^Con and siRNA^LRP1.1-treated cells (P>0.5), but BiP levels were elevated 4.4±0.9 fold (P<0.05), suggesting that an ER stress response had been induced. However, the proportion of intracellular PrP^C that colocalised with BiP was nearly halved in the siRNA^LRP1.1-treated cells (Table 3), so this compartment contained 1.9 fold the normal PrP^C level. The mid- and trans-Golgi retained their normal share of intracellular PrP^C (and so contained 3.5 times their normal PrP^C).

To test whether the accumulation of PrP^C along biosynthetic pathways is due to an ER stress response, neurons were heat-shocked at 39°C for 4 hours. BiP levels were elevated 2.5 fold compared with their basal levels at 37°C (P=0.015). There was a minor, insignificant decrease in both surface and intracellular PrP^C levels (supplementary material Fig. S2 and supplementary material Tables S5a and S5b).

Transgenic overexpression of PrP^C, in lowering the ratio of LRP1:PrP^C, should affect PrP^C trafficking in the same way as downregulation of LRP1 if the latter is an obligate trafficking partner. Sensory neurons from PrP^C-overexpressing Tg20 mice were found to have 14-fold higher PrP^C levels than genetically wild-type (WT) neurons (similar to the increased expression reported in brain) (Fischer et al., 1996) and a much smaller (1.8 fold) elevation in their LRP1 level. As a result, Tg20 neurons have an eightfold higher than normal ratio of PrP^C to LRP1 (supplementary material Table S6a).

The excess PrP^C was evident in the time taken by Tg20 sensory neurons to endocytose their surface-labelled PrP^C. WT neurons internalised >80% of surface-labelled PrP^C within 2 minutes at 37°C, irrespective of the level of surface PrP^C they expressed (which varied over a tenfold range; Fig. 7A, left panel). Tg20 PrP^C-overexpressing neurons internalised only a small proportion of their surface-labelled PrP^C at this stage, with a strong inverse correlation between the level of PrP^C expressed and the proportion endocytosed (Fig. 7A, right panel). Even after 6 minutes at 37°C, the highest PrP^C-expressing Tg20 neurons still had ∼40% left on their surface (Fig. 7A,B; Table 4). This endocytosis of PrP^C was strongly inhibited by siRNA^LRP1.1 (Fig. 7B,C; Table 4). Thus, at high levels of overexpression of PrP^C, the expression level of LRP1 becomes rate limiting for PrP^C.

Transgenic overexpression of PrP^C also selectively increased intracellular PrP^C, evident by comparing the ratio of the intensity of immunofluorescent labelling of internal to surface PrP^C in optical cross-sections of neurons (Fig. 7D,E). This ratio was 1.4±1.3 for WT cells, compared with 5.0±1.5 for Tg20 neurons (Table 5), indicating a selective intracellular accumulation of PrP^C at the expense of surface levels in the overexpressing neurons. This mirrors the accumulation of intracellular PrP^C in biosynthetic compartments seen following LRP1 downregulation by prolonged siRNA^LRP1-treatment, including the accumulation of excess PrP^C primarily in post-ER compartments (Table 3). In the case of Tg20 neurons, however, there is no elevation in BiP levels (P=0.47; supplementary material Table S6b).

To assess biochemically whether PrP^C could interact with LRP1 on the cell surface and in intracellular compartments of normal, untreated sensory neurons, LRP1 was sequentially immunoprecipitated from the surface, and then from internal compartments, of these cells (Fig. 8A). A subset of surface PrP^C co-immunoprecipitated with LRP1 (Fig. 8A, lane 2 compared with lane 1). There was no co-immunoprecipitation of Thy-1, the major neuronal GPI-anchored surface protein with PrP^C or LRP1 (lanes 1′and 2′; total Thy-1 is in the unbound fractions, lanes 5′ and 6′).

The major proportion of PrP^C on the sensory neurons was immunoprecipitated from internal compartments (Fig. 8A, lane 3 compared with lane 1, with unbound PrP^C shown in lane 5). A substantial proportion of internal PrP^C co-immunoprecipitated with LRP1 (Fig. 8A, lane 4 compared with lane 3), although a major part did not (unbound fraction in lane 6).

To identify whether any of the PrP^C associated with LRP1 was in the ER, samples immunoprecipitated by antibodies against either PrP^C or LRP1 were subjected to control or endoglycosidase H (endoH) digestion and then electrophoresed and immunoblotted for PrP^C. It is difficult to discern an endoH-sensitive band within total PrP^C (Fig. 8B, lanes 1 and 2), whereas, in the proportion of PrP^C co-immunoprecipitated with LRP1, there was an increase in fully deglycosylated PrP^C after endoH treatment (Fig. 8B, lanes 3 and 4), suggesting an ER origin of part of the PrP^C bound to LRP1 in normal sensory neurons.

As the N-terminal domain of PrP^C is flexible and does not interact structurally with the C-terminal domain (Hornemann et al., 2004), the interaction with purified human placental LRP1 of the N-terminal residues 23-107 of PrP^C, isolated as a GST fusion protein (GST-PrP^23-107), was determined by surface plasmon resonance (SPR). The N-terminal domain bound with high affinity to LRP1 (Fig. 9A). The binding affinity was calculated from the estimated association (K_a=2.75×10⁴ M^–1s^–1) and dissociation (K_d=5.44×10^–4 s^–1) rate constants as K_D=K_d/K_a=20 nM (Fig. 9A). Similar affinities were measured in the range of 70–350 nM ligand. This binding was strongly inhibited by chelating Ca²⁺ or lowering the pH to 6.0 (data not shown), as expected for an LRP1 ligand (Fisher et al., 2006; Herz et al., 1988; Moestrup et al., 1993a). Of particular interest was the ability of PrP^C to compete with RAP. RAP at 1 μM rapidly saturated LRP1; with RAP prebound and its concentration held at 1 μM, further addition of GST-PrP^23-107 at 100 nM bound to LRP1 to 45% of its control level (Fig. 9B). This demonstrates a remarkable ability of the N-terminal domain of PrP to bind to LRP1 in the presence of its biosynthetic chaperone.

The rapid and constitutive endocytosis of PrP^C on sensory neurons was inhibited by the downregulation of LRP1 by three specific siRNAs; physical association was found by co-immunoprecipitation of PrP^C and LRP1 on the neuronal surface, confirming fluorescence and EM immunocytochemical colocalisation that increased during the process of endocytosis; and the N-terminal domain of PrP^C, responsible for its endocytosis (Shyng et al., 1995a; Sunyach et al., 2003), bound with high affinity to LRP1. In addition, RAP inhibited PrP^C endocytosis, although unexpectedly high concentrations were needed; 80 nM is normally sufficient to inhibit binding of ligands to LRP1 [e.g. this concentration inhibits 90% of the binding of prion fibrils to sensory neurons (C.J.P., unpublished)], yet, in its presence, 60% of surface PrP^C was still internalised. Even at 1 μM RAP, 30% of PrP^C was still endocytosed within 2 minutes, in comparison with the 55% inhibition of GST-PrP^23-107 by this concentration of RAP. Inhibition of PrP^C by RAP occurs but is atypical for an LRP ligand. Overall, however, converging lines of evidence argue that LRP1 is the key endocytic partner of PrP^C on these neurons.

This conclusion is perhaps not unexpected, given the role of LRP1 in mediating the endocytosis of the GPI-anchored plasminogen activator receptor (Horn et al., 1998). Furthermore, Taylor and Hooper (Taylor and Hooper, 2007) reported that siRNA^LRP1 knockdown of LRP1 inhibits the 100 μM Cu²⁺-dependent endocytosis of exogenous transfection-expressed PrP^C on SH-SY5Y cells. These authors showed reduced internalisation of PrP^C in the presence of siRNA^LRP1 compared with untreated cells, without testing whether their long-term siRNA^LRP1 treatment reduced surface expression of PrP^C. In the light of our data, it would seem premature to conclude that it is the Cu²⁺-dependent endocytosis of PrP^C, rather than its surface level, that is reduced by lowering LRP1 levels in SH-SY5Y cells.

Given the diverse roles of LRP1 in controlling neuronal lipid metabolism and surface proteins (Herz and Chen, 2006; May and Herz, 2003), separating direct from indirect effects of modulating LRP1 expression is not trivial. Not only did sustained reduction of LRP1 halve the level of surface PrP^C, it elevated Thy-1 and particularly APP levels. The latter was expected as LRP1 controls the degradation of APP (Pietrzik et al., 2004; Zerbinatti et al., 2004), but the former, not so. There is no known interaction between LRP1 and Thy-1 on the cell surface (they do not co-immunoprecipitate and, as Thy-1 is not rapidly endocytosed, they are most unlikely to interact). Possibly GPI-anchored proteins compete for a rate-limiting step in transport from the trans-Golgi to the cell surface, so that Thy-1 would benefit from lowered PrP^C traffic.

We have previously noted that the surface environment of PrP^C on neurons is complex, with PrP^C leaving its detergent-resistant domains to traverse normal (detergent soluble) membrane for endocytosis (Brügger et al., 2004; Madore et al., 1999; Sunyach et al., 2003). This complexity is evident in the colocalisation of 50% of PrP^C with the lipid `raft' marker ganglioside GM1 (compare the ∼70% distribution of PrP^C in detergent-resistant membrane) (Madore et al., 1999; Sunyach et al., 2003), and the increasing colocalisation of PrP^C with LRP1 as endocytosis progresses. We have been unable to determine whether LRP1 is within detergent-soluble or -insoluble membranes on the surface of sensory neurons because it is too readily proteolysed in detergent extracts from these cells. However, we can find no LRP1 at all within detergent-insoluble membrane fractions from brain membranes (E.G.M.V., unpublished). On myoblasts and Chinese hamster ovary (CHO) cells, LRP1 is similarly located not within detergent-resistant membranes, although it is reported to transit briefly detergent-resistant, before entering detergent-soluble, membranes for endocytosis on mouse embryonic fibroblasts (Wu and Gonias, 2005). We suspect that LRP1 does not physically enter neuronal rafts, although it might not need to in order to bind to raft-associated PrP^C as it is so large it could readily extend over the 40-100 nm diameter (Kusumi et al., 2005) of membrane rafts.

The size of LRP1 might lead to quite a different role for the PrP^C-LRP1 interaction in vivo from the endocytic one described here. LRP1 is so massive it could multiply-span the normal intercellular spaces of adult brain. It is more than four times larger than N-cadherin that homotypically binds to itself across the ∼20 nm gap of central nervous system (CNS) synaptic clefts (Bruses, 2006; Zuber et al., 2005). In the closely packed confines of the CNS, LRP1 could bind to PrP^C on an adjacent, as readily as its own, cell surface. This would produce the opposite function to that studied in our isolated cell system – intercellular binding between PrP^C and LRP1 presumably would prevent the endocytic trafficking of both, retaining each on the apposed surfaces until the binding of other soluble ligands (e.g. astrocyte-derived ApoE) displaced PrP^C to allow LRP1 to internalise the newly bound cargo.

This study specifically concentrated upon isolated neuronal cell bodies, devoid of attachment to other cells. Nonetheless, occasionally two neuronal cell bodies were in contact, and at such points distinct immunolabelled PrP^C remained on the surface after 2 minutes at 37°C (data not shown). The possibility that intercellular interactions between PrP^C and LRP1 stabilise both molecules to their apposed surfaces needs to be further investigated with cells in contact. Intercellular ligand binding functions have been proposed for PrP^C (e.g. Mange et al., 2002; Shmerling et al., 1998). Intercellular restriction of LRP1 trafficking by PrP^C might be particularly important at synapses, where the intracellular domain of LRP1 can organise the distribution of postsynaptic receptors through linking adaptors such as PSD95 (Herz and Chen, 2006).

Intercellular binding by LRP1-PrP^C, in restricting endocytic trafficking of both, would presumably increase the half-life of LRP1 on sensory neurons from the ∼1 hour implied by our siRNA inhibition towards the 3 hours measured on HepG2 cells (Melman et al., 2002).

Members of the LRP family of receptors, each with multiple high-affinity binding sites for secreted and surface membrane proteins, form massive complexes with their ligands in the ER that prevent export to the Golgi unless their binding sites are masked by RAP (Willnow et al., 1996). This client-specific chaperone binds to nascent LRP1 as it emerges into the ER lumen (Bu and Rennke, 1996) and remains bound until LRP1 reaches the trans-Golgi (Lee et al., 2006), where the acidic (pH 5.8) environment dissociates RAP (which is retrieved by ERD2 to the ER) (Bu et al., 1997), leaving activated LRP1 to proceed to the cell surface.

LRP1 co-immunoprecipitates endoH-sensitive PrP^C from normal sensory neurons, indicating a direct interaction between these proteins in the ER. PrP^C must therefore be able to bind to LRP1 in the presence of RAP in biosynthetic compartments, a property demonstrated by SPR by the ability of GST-PrP^23-107 to bind at 45% normal capacity to LRP1 in the presence of 1 μM RAP. This, we believe, is the only reported case of a ligand binding to LRP1 in the presence of RAP. Moestrup and colleagues have extensively studied by SPR the binding kinetics of LRP family ligands. RAP has been used as a standard control and has in all previous cases completely inhibited, at 1 μM, all ligand binding (e.g. Moestrup et al., 1998; Nykjaer et al., 2001).

What is unexpected is that this interaction enhances the progress of PrP^C through biosynthetic compartments. LRP1 has been observed to promote the biosynthetic trafficking of β1 integrin to the cell surface, although direct contact between the receptor and the integrin could not be demonstrated (Salicioni et al., 2004). With levels of LRP1 reduced, high levels of PrP^C accumulated in biosynthetic compartments up to, and including, the trans-Golgi. Sustained reduction of LRP1 levels also induced high BiP levels, indicative of a stress response. However, intracellular accumulation of PrP^C was not observed by heat-shock alone, and was found with transgenic neurons with highly elevated levels of PrP^C but not LRP1 (or BiP). Taken together, these data argue for a functional role for LRP1 in expediting the biosynthetic trafficking of PrP^C.

A block in PrP^C biosynthetic trafficking at an endoH-insensitive stage occurs following deletion of N-terminal residues (23-50) (Gilch et al., 2004; Nunziante et al., 2003), following suramin-induced aggregation (Gilch et al., 2004), and with mutant forms of PrP^C that are retained in an endoH-insensitive compartment in neurons (Stewart and Harris, 2005). The LRP1-dependent block in transport seen here by lowering LRP1 levels might be a general restriction point in PrP^C biosynthesis.

We do not currently have the tools to answer this question definitively, but, from the work described here and ongoing studies, it seems highly likely that LRP1 is the central organiser, but not the sole functional component, of a larger complex that traffics PrP^C during biosynthesis and at the cell surface. We find that dissociating LRP1-PrP^C complexes is different depending on the cell source (RER microsomes versus cell surface; E.G.M.V., unpublished). Other binding partners might act as co-receptors for cellular PrP^C, decreasing its overall affinity for LRP1 sufficiently for RAP (and, presumably, other more physiologically relevant surface ligands such as ApoE) to displace PrP^C from the receptor.

Strong candidates for such a role are heparan sulphate proteoglycans (HSPGs). These bind to basic motifs in the N-terminus of PrP^C (Warner et al., 2002), affect its rate of endocytosis (Shyng et al., 1995b) and the formation of PrP^Sc (Ben-Zaken et al., 2003; Deleault et al., 2005). HSPGs have complex interactions with the binding and internalisation of LRP1 ligands such as thrombospondin (Wang et al., 2004) and lipoproteins (Wilsie and Orlando, 2003) and might similarly modulate the association-dissociation kinetics of LRP1-PrP^C complexes. We have noted that enzymatic removal of HSPG on sensory neurons decreases the rate of endocytosis of PrP^C (Sunyach et al., 2003).

LRP1 performs two roles in internalising proteins: for nutrient uptake (particularly lipid/cholesterol), and, as a surface scavenger, removing spent debris such as protease-inhibitor complexes from the extracellular space (May and Herz, 2003). PrP^C could function as part of the LRP1 scavenger receptor complex as its N-terminal domain has multiple binding motifs (Caughey et al., 2006) and its hydrophobic sequence (residues 112-130), which is exposed to the aqueous environment, could bind to denatured proteins as PrP^C rapidly traffics across the neuronal surface. A scavenger receptor role is also suggested by the pattern of PrP^C expression in vivo, where it is most abundant on cells that might need enhanced levels of protection from the extracellular environment: on neurons, which cannot divide and live as long as the body they control, and on cells that constantly monitor danger in the environment – normal and follicular dendritic cells, intraepithelial lymphocytes and neuroendocrine cells (Ford et al., 2002a; Ford et al., 2002b).

Members of the LRP family could play multiple roles in prion infection. Infectious prion fibrils bind with picomolar affinity to very-low-density lipoproteins (Safar et al., 2006) and so might be cleared from extracellular space by LDL receptors. With neuronal LRP1, the crucial questions are whether LRP1 also binds to infectious prion fibrils and, if so, whether the same molecule of LRP1 (which binds most ligands to both cluster 2 and 4) (May and Herz, 2003) will bind to both the cellular and infectious forms of PrP? Then, given the massive size of LRP1, would such binding of the template and substrate in the prion conversion reaction bring the two together or keep them apart?

A cDNA fragment encoding residues 23-107 of PrP^C (Sunyach et al., 2003) with a C-terminal Cys residue (added to provide an anchor for conjugation) was cloned into the pGEX-4T1 vector (GE Healthcare). GST fusion protein was purified on glutathione-sepharose beads in the presence of 1 mM DTT. The Cys residue was blocked with 5 mM iodoacetamide in PBS to prevent disulphide bond formation, confirmed by recovery in non-reducing PAGE of the monomeric fusion protein without higher polymers.

Monomeric affinity-purified Fab against mouse PrP residues 142-162 (referred to as `2S antibody') was coupled to Alexa-Fluor-488 or -594 (Molecular Probes; 1–3 moles of dye per mole of Fab) or to 5 nm gold (British Biocell) as described previously (Sunyach et al., 2003). To detect PrP on immunoblots or for immunocytochemistry of fixed cells, SAF32 and SAF83 (SPI-Bio Massy, France) monoclonal antibodies against PrP were used, directly coupled to fluorochromes or gold for immunocytochemistry. Mouse monoclonals to syntaxin 6 (Abcam) and α-mannosidase II (Covance), the LRP1 ligand α₂M^* (Research Diagnostics) and cholera toxin B chain (Sigma) were coupled to fluorochromes, as described above. Fe²⁺-loaded transferrin (Tf) coupled to Alexa-Fluor-488 or Texas Red was from Molecular Probes; both ligands were used at 50 μg/ml.

For detection of LRP1, rabbit antibodies against the entire LRP1 (RRR) (Bu et al., 1995), or its cytoplasmic domain (MMMM) (Zerbinatti et al., 2004), were used; LRP1b was detected with an anti-epitope antibody (Cam et al., 2004). Rabbit antibodies against other neuronal LRP family members (LRP6, LRP8, VLDLR) were from Santa Cruz, and against BiP were from Abcam (ab21685). Rabbit antibodies were detected with our own affinity-purified, species-specific horse anti-rabbit IgG conjugated to Alexa fluorochromes, as described above.

To study the binding of GST-PrP^23-107 to LRP1 by SPR on a Biacore 3000 (Biacore, Sweden), CM5 sensor chips were activated with a 1:1 mixture of 0.2 M N-ethyl-N′-(3-dimethylaminopropyl) carbodiimide and 0.05 M N-hydroxysuccimide in water. LRP1, purified from human placenta (Moestrup et al., 1993a), was immobilised at a concentration 15 μg/ml in 10 mM sodium acetate pH 3.0, to a receptor density of 20-30 fmol LRP/mm². Remaining binding sites were blocked with 1 M ethanolamine pH 8.5. A control flow cell was made by performing the activation and blocking procedure only. Samples were dissolved in 10 mM HEPES, 150 mM NaCl, 1.5 mM CaCl₂, 1 mM EGTA, pH 7.4 plus 0.005% Tween 20. Sample and running buffers were identical. Samples were injected at a flow rate of 5 μl/minute. Regeneration of the sensor chip after each analysis cycle was performed with 1.6 M glycine-HCl buffer, pH 3.0. The Biacore response is expressed in relative response units (RU), this being the difference in response between the protein and control flow channel. Kinetic parameters were determined by BIAevaluation 4.1 software using a Langmuir 1:1 binding model and simultaneous fitting of all curves in the concentration range.

Neurons were used after 3-7 days in culture established from the dorsal root ganglia of 4-8-week-old 129Sv, or Tg20 transgenic (Fischer et al., 1996), mice (Mus musculus) or (for biochemistry) rats (Rattus norvegicus) (Sunyach et al., 2003). Genetically normal rat and mouse neurons endocytosed their endogenous PrP^C at the same rate and were inhibited by the siRNA^LRP1 used with similar efficiency.

For siRNA delivery by penetratin-1 (Davidson et al., 2004; Dom et al., 2003), all extracellular DNA was removed by treating cultures overnight with RNAse-free DNAse (5 units/ml; Qiagen) and including the enzyme in the incubation with siRNA-penetratin.

Endocytosis assays were performed as described previously (Sunyach et al., 2003). Briefly, cells were prelabelled with ligands for 30 minutes at 10-12°C in B-27 supplement-free medium, washed once and transferred to a 37°C incubator for endocytosis, normally for 0-2 minutes. At timed intervals, samples were fixed with 1.5% PFA for 15 minutes, chromatin stained with 4′,6 diamidino-2-phenylindole (DAPI), and mounted in mowiol. Cells were observed with a Zeiss Axiovert 100 microscope with a piezo Z-axis drive on a 63× apochromat 1.4 NA oil objective illuminated with an EXFO X-cite 120 metal-halide lamp and images collected with a Hamamatsu Orca-ER camera. Exposure time (50-500 mseconds) and excitation intensity were set so that points of highest light intensity were within 20-60% saturation for the camera. Images (normally 30) taken every 100 nm through a middle section of the cell body [to obtain multiple data points per wavelength in the Z axis (Heintzmann, 2005)] were transferred to Volocity (Improvision; Warwick, UK), subtractively deconvolved against their seven neighbouring images, assembled into a 3D stack and fluorescence measured using Volocity. Background was set at 10% above the level that resolved the measured fluorescence into multiple discrete objects (rather than record the whole cell as a single fluorescent object). Surface objects were defined by drawing a line just below the surface of each cell.

For EM, gold was labeled with Fab, Tf or α₂M^* and incubated with cells and processed for EM as described previously (Sunyach et al., 2003). Fields of both PrP^C-α₂M^* and PrP^C-Tf double labeling were viewed to count >500 particles of 5 nm gold labeling for PrP^C, and the proximity to them of the 10 nm label for the other two ligands determined. Label for the LRP1 ligand α₂M^* was present at 80%, and that for Tf at 60%, the abundance of label for PrP^C. The relative density of the α₂M^* and Tf labels was similar (respectively 80% and 60% the density of PrP^C label), as was the fraction of label for all three within coated pits (11-18%).

To assess surface immunolabelling, where the relative distribution of surface proteins was crucial (Fig. 2C,D; Table 1), neurons were kept at 37°C as they were washed in PBS and fixed for 15 minutes with 0.5% PFA, then immunolabelled at 4°C. Where the amount rather than surface distribution of surface protein was crucial, living cells were immunolabeled at 10°C, then fixed in 0.5% PFA for 15 minutes. If internal protein was also to be measured, cells were permeabilised at 4°C in Brij-96–PBS for 30 minutes, then internal protein was immunolabelled. Cells were DAPI stained, fixed again in 1.5% PFA-PBS for 1-16 hours before being mounted in mowiol. If primarily internal proteins were to be labeled (e.g. Fig. 6), cells were fixed for 30 minutes with 1.5% PFA at room temperature, then permeabilised for 30 minutes at 4°C in 0.2% Triton-X-100–PBS before immunolabelling. Mowiol moutant was made in PBS containing 1/7 by volume mowiol 4-88, 2/7 glycerol, 0.1% NaN₃, 0.1 mg/ml DABCO [1,4 diazabizyclo(2.2.2)octane] and 0.01 mg/ml n-propyl gallate (all from Sigma).

Recombinant RAP (80 nM or 1 μM) (Andersen et al., 2001) was preincubated with cells for 30–60 minutes at 37°C before the PrP internalisation assay was performed. The siRNA^LRP1.1 sequence was adapted for the mouse from the human sequence that we have found to be effective in suppressing LRP1 in human smooth muscle cells (Li et al., 2003); sense strand Th-5′-GCAGUUUGCCUGCAGAGAGdTdT-3′, antisense strand 5′UCUCGCAGGCAAACUGCdTdT-3′). The control siRNA^Con corresponded to the reverse of the siRNA^LRP1 sequence. NCBI Blastn showed that the siRNA^LRP1.1 would only target mRNA encoding Lrp1 and that there would be no target for siRNA^Con. In addition, siRNA^LRP1.2 (sense strand Th-5′-CGUUGGUUAUGCACAUGAAdTdT-3′, antisense strand 5′UUCAUGUGCAUAACCAACGdTdT-3′), and siRNA^LRP1.3 (sense strand Th-5′-GCCGGGUGUACAAAUGUAAdTdT-3′, antisense strand 5′UUACAUUUGUACACCCGGCdTdT-3′) were predesigned and homology tested using proprietary algorithms by the manufacturer (Qiagen). The appropriate thio-modified double-stranded RNAs (Dharmacon and Qiagen) were coupled to penetratin-1 (Quantum Biotechnologies) following Davidson and colleagues (Davidson et al., 2004) and added to cells at a final concentration of 250 nM.

For sequential immunoprecipitation of surface then intracellular antigen, antibody [SAF32 (1 μg) for PrP^C; RRR (3 μg) for LRP1; Ox7 (1 μg) for Thy-1] was added to 1.5 ml of culture medium and neurons held at 10°C for 30 minutes with gentle rocking before being washed twice in ice-cold PBS and recovered from the plate by aspiration. Cells were lysed in 0.5% Brij 96 plus 0.5% sodium deoxycholate plus 1 μM BSA. Nuclei were pelleted at 2400 g for 5 minutes, and protein A Dynal beads (10% v/v) were incubated with the supernatant for 30 minutes at 4°C, then washed with lysis buffer, and antigen eluted using Laemmli SDS-PAGE sample buffer.

Internal antigen was then immunoprecipitated by incubating the unbound fraction with primary antibodies followed by protein A Dynal beads and eluted with 0.1% SDS in 10 mM Tris pH 6.8 for digestion with endoglycosidase H (endoH).

For hydrolysis of core glycan chains with endoH, samples were reduced with 10 mM DTT for 5 minutes at 100°C, diluted twofold in 0.5 M citrate buffer pH 5.5 plus 0.1 mM EDTA and endoH (Roche) added at 0.1 U/ml for 60 minutes at 37°C. The reaction was stopped by boiling samples in Laemmli sample buffer.

We thank Carol Troy (Columbia, NY) for advice on coupling and use of penetratin-siRNA, Charles Weissmann for the Tg20 mice, Ken Brady (Centre for Ultrastructural Imaging, King's College London, UK) for expert assistance with the EM, and Maria Woolston for positive controls for antibodies against the LRP family; we also thank Helen Cox, Paul Felts and Susan Standring for many useful discussions, and Improvision for continuing support. This work was supported by BBSRC grants 18/BS516350 and BB/C506680/1. We also acknowledge the reviewers for their incisive, constructive comments that have much improved this paper. This work was supported by BBSRC grants 18/BS516350 and BB/C506680/1, and an MRC doctoral training studentship to C.J.P.