The GPI-anchoring of PrP

Abstract

The cellular prion protein (PrP^C) is an N-glycosylated GPI-anchored protein usually present in lipid rafts with numerous putative functions. When it changes its conformation to a pathological isoform (then referred to as PrP^Sc), it is an essential part of the prion, the agent causing fatal and transmissible neurodegenerative prion diseases. There is growing evidence that toxicity and neuronal damage on the one hand and propagation/infectivity on the other hand are two distinct processes of the disease and that the GPI-anchor attachment of PrP^C and PrP^Sc plays an important role in protein localization and in neurotoxicity. Here we review how the signal sequence of the GPI-anchor matters in PrP^C localization, how an altered cellular localization of PrP^C or differences in GPI-anchor composition can affect prion infection, and we discuss through which mechanisms changes on the anchorage of PrP^C can modify the disease process.

Keywords: :

• GPI-anchor
• lipid rafts
• prion diseases
• prion protein
• processing
• PrP
• signaling

PrP, a GPI-Anchored Protein

The prion protein (PrP^C) acquires two post-translational modifications along the secretory pathway: the addition of up to two N-glycan moieties at positions 181 and 197 (human amino acid numbering) and a glycosylphosphatidylinositol (GPI)-anchor at position 230 which attaches the protein to the outer leaflet of the plasma membrane (for review see ref. 1). Through a yet not fully understood mechanism, PrP^C, which is highly expressed in neurons, can change its conformation into an abnormal protease-resistant, potentially infectious isoform (PrP^Sc). The latter is an essential part of the agent responsible for transmissible spongiform encephalopathies (TSE), the prion, and is still GPI-anchored.^2,3 In this review, to simplify matters, we will use the term PrP^Sc as a general term for disease-associated PrP species independently of their biochemical properties or aggregation status, although it should be noted that different terms exist that imply conceptual differences.

The GPI-anchorage is highly conserved among eukaryotes and used by 10–20% of all membrane proteins in eukaryotic cells. Biosynthetically it is a complex attachment involving more than 20 genes implicated in ten reactions in the endoplasmatic reticulum (ER) which are not yet fully understood.^4,5 GPI-anchored proteins (GPI-APs) have a wide range of functions ranging from receptors and cell surface hydrolases to adhesion molecules among others^6,7 but there is no clear correlation between GPI-APs and distinct biological functions. Therefore it is hypothesized that the GPI-anchor bestows some advantages over other types of more simple attachments. For instance, GPI-APs are more mobile within the plane of the lipid bilayer and occupy less space than proteins with a transmembrane domain, allowing the proteins to be more tightly packed together. Moreover, GPI-APs also have the capacity of modifying the lipid composition of the membrane itself⁸ and GPI-anchors may also be necessary to complete the formation of the functional structure of the protein as described for Thy-1, another GPI-AP highly expressed in neurons.^9,10 One intriguing capacity of GPI-APs is their ability to be spontaneously inserted into membranes in vitro and in vivo, a phenomenon known as “cell surface painting,”^11,12 which has also been observed for PrP^C.¹³

All GPI-APs contain a signal sequence at the N-terminus that directs the protein to the ER during synthesis at the ribosome. Once it is translocated to the ER, a second hydrophobic signal sequence (SS-GPI) present at the C-terminus of the protein is removed by a transamidase in a fast reaction of about a minute and a pre-formed GPI-anchor is attached.¹⁴ All GPI-anchors contain a common structure composed of an ethanolamine phosphate in an amide linkage to the C-terminal carboxyl group of the protein, a highly conserved glycan core of three mannose residues, glucosamine and a phosphatidylinositol (PI) group, and phospholipid tails that attach the GPI-anchor to the cell membrane. The addition of side-branching sugars and modifications of the lipid moiety during the passage through the secretory pathway to the plasma membrane depends on the type of cell and organism, thus leading to a large structural diversity. Of note, PrP^C is one of only three proteins (the others being porcine membrane dipeptidase¹⁵ and CD59¹⁶) known to date that contain a sialic acid in their GPI-anchor structures¹⁷ (for a review see ref. 18).

The Signal Sequence of GPI (SS-GPI) as a Sorting Signal

Interestingly, further lipid remodeling of the GPI-anchor occurs in the Golgi apparatus where the GPI-APs associate to specialized microdomains known as lipid rafts or detergent resistant membranes (DRMs). These domains are described as nanodomains highly enriched in sphingolipids and sterols that contain specific proteins and that can coalesce by lipid-lipid, protein-lipid, and protein-protein interactions¹⁹ forming a platform for signal transduction through cellular receptors.^20-23 In fact, several GPI-APs, which are attached to the outer leaflet of the plasma membrane, can signal through tyrosine kinases after raft clustering,²⁴ probably mediated by transmembrane proteins that are also found in lipid rafts.²⁵

It has been proposed that the GPI-anchor is a sorting signal for selective targeting to lipid rafts and, in the majority of instances, a sorting signal for being apically sorted in polarized cells and to axons in neurons.^26-29 However, some exceptions to this rule have been described and, in fact, PrP^C is one of these exceptions, being sorted mainly basolaterally in Madin-Darby canine kidney cells (MDCK), which represent a well-established model of polarized cells.³⁰ Concerning signals that direct PrP^C to the basolateral membrane, Paladino et al. proposed that the SS-GPI could play a role in its sorting. Thus, when the green fluorescent protein (GFP) is attached to the SS-GPI of PrP^C, it is sorted basolaterally whereas when it is attached to the SS-GPI of the folate receptor, it is directed to the apical side in MDCK cells. The same authors also suggested that the differential mechanism for apical vs. basolateral sorting of GPI-APs in lipid rafts could be the ability of the apically sorted GPI-APs to form high molecular mass complexes (HMM) in the Golgi apparatus,³¹ which could depend on the SS-GPI.³² As shown in Figure 1, our studies also revealed that, when the SS-GPI of PrP^C is replaced by the one of Thy-1, the resulting PrP^CGPIThy-1 relocates from the basolateral to the apical side in MDCK cells.³³PrP^C and Thy-1 both reside in lipid rafts but share different lipid environments within these domains speaking in favor of structurally heterogeneous raft domains^34,35 which could have a role in sorting through HMM formation.³⁶ The localization of PrP^C in neurons is more controversial. Here it can be found enriched in either axons or cell bodies and dendrites.^35,37,38 How a signal peptide, that is removed in less than a minute after the protein is translocated into the ER, can affect the sorting of GPI-APs and maybe the GPI-anchor structure and/or the lipid environment, is a question that remains unanswered. However, other factors apart from membrane composition can play a role. It is known that the SS-GPI markedly influences the efficiency of the transamidation process, which in turn may influence membrane localization.^11,39 In a more daring concept it has also been proposed that different signal sequences can actually direct the addition of different types of GPI-anchors.^40,41 However, experimental proof of such a mechanism is still required.

Figure 1. PrP^C and PrP^CGPIThy-1 expression in MDCK cells. (A) Confocal microscopy showing Z-stacks taken from top to bottom. As published before³³PrP^C (in green) is basolaterally (b) sorted in fully polarized MDCK cells whereas changing the SS-GPI of PrP^C for the one of Thy-1 preferentially directs the PrP^CGPIThy-1 to the apical (a) compartment. Scale bar is 10μm. Originally published in PLoS One. (B) Schematic representation of the differential sorting of wild type (WT) PrP^C and PrP^CGPIThy-1 shown in A.

It is important to note that three point mutations (M232R, M232T, and P238S) located at the SS-GPI of PrP^C lead to familial prion disease in humans.⁴² It has been proposed that the two first mutations are related to an increased transmembrane orientation which could explain neurotoxicity.⁴³ More recently, Guizzunti et al. used a chimeric protein expressing EGFP with the SS-GPI of PrP linked to a Myc tag to demonstrate that the P238S mutation in the SS-GPI of PrP prevents degradation of the SS-GPI by the proteasome. As a consequence the cleaved SS-GPI peptide tends to accumulate in the ER. With this elegant approach, the authors monitored for the first time the fate of a GPI-anchor signal peptide and also provided an hypothesis of how mutations in the SS-GPI of PrP^C might contribute to the development of prion disease.⁴⁴

The GPI-Anchor of PrP^C and its Role in Prion Disease

The type of membrane attachment or modifications in the GPI-anchor of PrP^C has an impact in the pathophysiology of prion disease. The site of conversion of PrP^C to PrP^Sc has been proposed to be at the plasma membrane and/or in the endocytic compartment^45-47 with the requirement that PrP^C and PrP^Sc are attached to the same raft domain at the membrane in cell-free conversion reactions.⁴⁸ A recent study using tagged PrP^C, which is able to be converted to PrP^Sc, pinpointed de novo formation of PrP^Sc at the plasma membrane to lipid raft-enriched domains. Conspicuously, this is a very fast process taking about a minute. PrP^Sc is then rapidly internalized into early endosomes.^49,50 These studies imply that any perturbation of lipid rafts will have an effect on prion conversion. For instance, depletion of cholesterol, either by inhibiting its biosynthesis⁵¹ or by addition of amphotericin B or filipin treatment (antibiotics known to sequester cholesterol), has been shown to decrease or inhibit PrP^Sc formation.^52,53 On the other hand, when PrP^C is recombinantly expressed as a transmembrane protein (CD4PrP^C), which is then targeted out of lipid rafts, PrP^Sc formation is also abolished.⁵⁴ Moreover, although GPI-anchorless PrP^C can be efficiently converted to PrP^Sc in a cell-free conversion assay,⁵⁵ cells expressing GPI-anchorless PrP^C do not support persistent infection.⁵⁶ Thus, in vitro, the presence of PrP^C at the cell surface is a prerequisite for efficient conversion to PrP^Sc.⁵⁷

Taken together, these reports indicate that PrP^C localization in lipid raft-enriched domains is of high importance for prion conversion. Interestingly, only when the amyloidogenic yeast protein Sup35 is attached to a GPI-anchor, it can be propagated to neighboring cells and these cells can then sustain and propagate the aggregates for many passages. This may raise the idea that prion diseases, in contrast to other protein misfolding diseases such as Alzheimer disease, are naturally transmissible because PrP^C is the only amyloidogenic protein carrying a GPI-anchor.⁵⁸

Recent reports highlighted the idea that prion infection can be modified not only by the presence or absence of its GPI-anchor or by disrupting lipid rafts but also by altering the structure/composition of the GPI-anchor itself. Nisbet et al. showed that by exchanging amino acids at the C-terminal moiety within the SS-GPI of murine PrP^C (MoPrP) by the ones of rabbit PrP^C (RbPrP, rabbits are naturally relatively resistant to prion infection⁵⁹), the resulting MoPrP-RbGPI becomes resistant to PrP^Sc conversion.⁶⁰

Interestingly, studies coming from the group of Bate and Williams showed that modifications in the lipid part of the GPI-anchor or of the sialic acid residue lead to alterations in cholesterol at the plasma membrane and to reduced PrP^Sc loads and infectivity (these experiments are discussed below in the section “Signaling through PrP^C”).

In vivo, the scenario becomes more complex. Neuropathologically, prion diseases are characterized by the presence of diffuse or (more rarely) amyloidic depositions of PrP^Sc, neuropil vacuolation (spongiosis) as well as abundant astrogliosis and microglial activation. On the ultrastructural level, many of the lesions that are specific for TSE and co-localize with PrP^Sc are related to lesions of the plasma membrane.⁶¹

Although the presence of PrP^C is mandatory in order to develop prion diseases,⁶² in inducible knockout mice infected with prions, the pathological changes in the brain are reverted once PrP^C production is shut off, in spite of further PrP^Sc accumulation.⁶³ This is very interesting because there is growing evidence that in prion diseases there are two mechanistically different, yet related phenomena, the conversion of PrP^C to PrP^Sc (prion propagation) and the neurotoxic signaling.⁶⁴ The latter is probably dependent on membrane attachment of PrP^C via its GPI-anchor and on the localization of PrP^C within lipid rafts as discussed later.

To study the issue of GPI-anchor attachment in vivo, Chesebro and colleagues generated mice expressing GPI-anchorless PrP^C and infected them with mouse-adapted prions. They showed that GPI-anchorless PrP^C is neither present in lipid rafts nor at the plasma membrane but that it is able to be converted to PrP^Sc, which presents with abundant thioflavin S-positive PrP^Sc amyloid plaques, especially around blood vessels. In these mice, infectivity was preserved but clinical onset of disease was delayed. When the animals died of disease, neurodegeneration was attributed to direct damage by amyloidic PrP^Sc accumulation either by disrupting the brain tissue or by causing vascular damage.⁶⁵Interestingly, the mice displayed a conspicuous lack of spongiosis. Spongiotic lesions can occur in a variety of neurodegenerative conditions but the type of vacuoles seen in TSE is very specific and genuine. The vacuole membrane is fragmented and contains membranes and granular osmiophilic granules.⁶¹These changes are independent of PrP^Sc deposition. The lack of spongiotic changes in the GPI-anchorless mice was attributed either to a decreased toxicity of amyloidic anchorless PrP^Sc or to the absence of the GPI-anchor and thus to the lack of neurotoxic events. Our own results also point in this direction as we have generated transgenic animals expressing PrP^C with the SS-GPI of Thy1 (PrP^CGPIThy-1) and we have observed that there is a relocalization of PrP^CGPIThy-1 to the axons. Moreover, after prion-infection, there is a delayed onset to clinical disease accompanied by an altered neuropathology, different prion strain characteristics, and decreased amounts of PrP^Sc (Puig et al., unpublished results).

Taken all of this into consideration, an important question arises: what are the mechanisms through which the alteration of membrane composition or alterations in the GPI-anchorage lead to changes in prion conversion and neurotoxicity? Regarding prion conversion, for GPI-anchorless PrP^C it is clear that conversion is enhanced (as has also been shown in cell culture models⁴⁸) but neurotoxicity is decreased. In other instances, such as in our PrP^CGPIThy-1 transgenic mice, when PrP^C is localized in another cellular compartment, it is possible that direct interaction between PrP^C and PrP^Sc is reduced at the plasma membrane and that conversion is thus slowed down. For the conversion of PrP^C to PrP^Sc a cofactor is considered necessary (for review see ref. 66). Many proteins have been proposed to interact with PrP^C but in vitro-data suggest that glypican-1, an heparan sulfate proteoglycan (HSPG) resident in lipid rafts, actually acts as a scaffolding protein that keeps PrP^C and PrP^Sc in lipid rafts and facilitates prion conversion.⁶⁷ Therefore, meeting specific partners/cofactors in the appropriate lipid raft domain may be of importance for efficient PrP^Sc formation. One can also hypothesize that the endocytic pathway, necessary for the prion replication, is also altered. In fact, the mechanism of internalization of PrP^C is different compared with other GPI-anchored proteins, which are usually endocytosed via a clathrin-independent pathway, although this largely depends on the cell type.⁶⁸PrP^C leaves raft domains in order to be endocytosed via clathrin-coated pits and is rapidly endocytosed in primary neurons. Thy-1, in contrast, is internalized with a much slower rate and via non-coated vesicles.⁶⁹ The low-density lipoprotein receptor-related protein 1 (LRP1) is thought to be the transmembrane protein that supports PrP^C to be internalized via clathrin-coated pits.^70,71 Other mechanisms implicating cooperation between lipid rafts and clathrin have also been proposed.⁷² One might speculate that by changing either the properties of the GPI-anchor or the membrane composition, endocytosis might be altered with regard to rates and routes of internalization.

With regard to toxicity and neurodegeneration, the signaling pathways that are linked to PrP^C could likewise be modified by alterations in the lipid composition of the GPI-anchor as well as by proteolysis of PrP^C. These issues will be briefly discussed in the following sections.

Signaling Through PrP^C

Many reviews are dedicated to the role of signaling through PrP^C in cell adhesion and in neurite extension/maintenance.^73,74 Here we will only focus on how alterations of lipid rafts or changes in the GPI-anchor could lead to disturbances in signaling implicated in prion diseases and, briefly, other proteinopathies.

As pointed out above, in vitro-studies from the group of Bate and Williams show that manipulation of the GPI-anchor composition leads to modulation of prion infection, implicating phospholipase A₂ (cPLA₂) as the toxic signaling pathway.^75,76 cPLA₂ is involved in the generation of signaling molecules through the arachidonic acid pathway but also in membrane trafficking (for a review see ref.77). On the one hand, clustering of PrP^C, either mediated through antibodies or by addition of PrP^Sc, increases the cholesterol content in membranes and activates cPLA₂ signaling, which then triggers synaptic damage. cPLA₂ activation and synaptic damage could not be observed after clustering of Thy-1, indicating a specific effect of PrP^C. The authors inferred that it is the clustering of the GPI-anchors, either caused by antibody cross-linking or PrP^Sc aggregation, that leads to neurotoxic effects. Importantly, when either the sialic acid or (one or both) acyl-chains are removed from the GPI-anchor of PrP^C, or when the acylation of the GPI-anchor of PrP^Sc is altered, these neurotoxic effects are not observed. The authors proposed that the membrane composition is altered in the presence of either monoacylated or desialylated PrP^C and that cPLA₂ can consequently not be activated.⁷⁸A similar scenario is observed when PrP^C is administered to prion-infected neuronal cell lines and integrates itself into the outer leaflet of the plasma membrane. This leads to increased amounts of cholesterol content in membranes and increased activation of cPLA₂, together with an increased production of PrP^Sc. On the contrary, when monoacylated PrP^C (PrP^C-G-lyso-PI) is introduced, these effects are abolished. PrP^C-G-lyso-PI is not present in lipid rafts, is not converted to PrP^Sc, and seems to displace cPLA₂ from PrP^Sc-containing rafts, thus no longer permitting toxic signaling. Moreover, a decrease in PrP^Sc production is observed suggesting either that PrP^C-G-lyso-PI is an inefficient template for conversion or that it competes with PrP^C for partners involved in endocytosis, thus altering the trafficking and limiting the interactions between PrP^C and PrP^Sc.^79,80

Other partners for signal transduction via PrP^C have been described. For instance, PrP^C has been involved in the inhibition of excitotoxicity mediated by the N-methyl-D-aspartate receptor (NMDAR).⁸¹ It has been hypothesized that this regulation is executed either through direct interaction of PrP^C with NMDAR subunits (PrP^C co-immunoprecipitates with the NR2D subunit of the NMDAR) or via protein kinases⁸² such as the Src tyrosine kinase family member Fyn.⁸³ Of note, NMDARs localize in lipid raft domains but can also be present in non-raft domains after phosphorylation.⁸⁴ Moreover, cholesterol depletion reduces localization of NMDARs in lipid rafts and leads to neuroprotection.⁸⁵ On the other hand, Mouillet-Richard et al. first described Fyn as being activated by crosslinking of PrP^C.⁸⁶ Since then, this pathway has been implicated in many of the functions that are attributed to PrP^C, for instance its role as a neuronal receptor for toxic amyloid β (Aβ) oligomers.⁸⁷ Binding of Aβ oligomers derived from patients with Alzheimer disease activates the Fyn pathway leading to NR2B phosphorylation and altered NMDAR localization.^88-90 However, clustering of PrP^C by crosslinking failed to activate this pathway, suggesting that binding of β-sheet-rich oligomers leads to a conformational change in PrP^C that is necessary to activate this pathway. It has been suggested that the concept of PrP^C as a receptor for toxic oligomers can be generalized for other β-sheet-rich oligomers found in different neurodegenerative conditions (including PrP^Sc in prion diseases) and that the toxicity following this interaction is mediated by NMDAR.⁹¹ The same authors highlighted the importance of PrP^C localization in lipid rafts to mediate toxic signaling, since a transmembrane form of PrP (CD4PrP) did not mediate any toxic effects. Other kinases that are altered as a consequence of PrP^Sc binding to PrP^C include ERK1/2, JNK, and the CREB transcription factor.^92,93 A schematic draw synthesizing the issues discussed here is shown in Figure 2.

Figure 2. Simplified scheme of toxic signaling through PrP^C. On the one hand (left) it has been described that PrP^C can bind Aβ oligomers at the post-synaptic density (PSD), either derived from recombinant sources or from brains of Alzheimer disease patients, and elicit toxicity through Fyn activation, NMDAR phosphorylation and altered localization.^88,90 Phosphorylation of NMDAR subunits would then lead to dendritic spine loss and excitotoxicty. As long as PrP^C is attached to the outer leaflet of the membrane and Fyn to the inner, it has been proposed that a transmembrane protein should act as a scaffold. LRP1¹¹⁸ and mGluR5¹¹⁹ have been suggested as the interacting transmembrane partners after Aβ oligomer binding to PrP^C. Resenberger et al. also showed that this signaling cascade can apply for various β-sheet rich conformers.⁹¹ In the case of PrP^Sc binding, apart from Fyn activation, other signaling targets, such as ERK1/2, p38 and JNK, are seen to be activated in neuronal cell lines.⁹² In addition, it might be hypothesized that β-sheet rich oligomer-associated NMDAR activation per se induces MAP kinase pathways. On the other hand (right) experiments by Bate and Williams demonstrated that after either crosslinking of PrP^C with antibodies or by applying PrP^Sc to cortical primary neurons, there is an increase in membrane cholesterol content that leads to a toxic pathway implicating cPLA₂ and arachidonic acid metabolites.⁷⁸ Reference to original studies is given by the numbers in brackets.

Anchorage and Localization of PrP^C May Influence its Proteolytic Processing

Proteolytic processing reflects an additional aspect of prion protein biology that is likely to be influenced by modifications of the membrane lipid composition or in the GPI-anchor attachment resulting in alterations in PrP^C localization. In recent years it has become increasingly obvious that evolutionary conserved cleavage events, especially the α-cleavage in the center of PrP^C94 and the shedding in close proximity to its GPI-anchor,^95,96 not only regulate PrP^C membrane homeostasis and reduce amounts of full-length PrP^C and, thus, the mediator of neurotoxic signaling and the substrate for conformational prion conversion. In addition, these cleavages produce biologically active fragments^97,98 (reviewed in ref. 99). Of note, the released soluble fragments, N1 (for α-cleavage) and shed PrP as well as their artificial correlates, have recently been linked with neuroprotection by blocking both the access of toxic oligomeric protein species to the cell as well as the conversion process in the context of prion diseases.^91,100-107 Thus, it is not surprising that these cleavage events are currently discussed as therapeutic targets with regard to prion diseases and other neurodegenerative proteinopathies.

The identity of the protease(s) responsible for the α-cleavage remains a matter of controversy (reviewed in refs. 108 and 109) and a recent study suggests a more complex scenario than initially expected, involving different members of the ADAM family of metalloproteases.¹¹⁰ In contrast, the role of ADAM10 as the major PrP^C sheddase has been established by recent studies.^{95,96,107,111}

While knowledge on the concrete localization of the two cleavage events is limited^54,111-113 both processing steps seem to depend on PrP^C membrane attachment. As a matter of implicitness, for a protein to be proteolytically cleaved it has to meet and interact with the relevant protease. Therefore, it would not be surprising if processes modifying the GPI-anchor attachment of PrP^C, resulting in altered structure or sorting/localization, or disturbing the integrity of lipid rafts may influence the substrate-protease interaction and alter the neuroprotective processing of PrP^C. In fact, it has been shown for other ADAM10 substrates mainly present in lipid rafts, that depletion of these microdomains significantly increased the proteolytic processing due to increased interaction with the protease (as active ADAM10 is preferentially found outside of rafts).^114-117 Moreover, attachment of a GPI-anchor to ADAM10 was shown to target the protease to lipid rafts, which resulted in altered processing of the amyloid precursor protein.¹¹⁶ With regard to PrP^C more research is necessary to clarify whether manipulations of this kind might offer a treatment option for prion diseases and other proteinopathies.

Concluding Remarks

Increasing evidence locates PrP^C as a key player in mediating neurotoxic signaling not only in prion diseases but in other neurodegenerative diseases such as Alzheimer disease. Understanding the importance of membrane localization of PrP^C in lipid rafts as well as how the anchorage of PrP^C can influence disease progression is of outstanding importance in order to develop further therapeutic strategies.

Disclosure of Potential Conflicts of Interest

No potential conflicts of interest were disclosed.

Acknowledgments

This work is supported by grants of the Werner-Otto-Stiftung (to H.A.) and the Deutsche Forschungsgemeinschaft (DFG, GRK1459 and FOR885).

10.4161/pri.27892