Clustering of sialylated glycosylphosphatidylinositol anchors mediates PrP-induced activation of cytoplasmic phospholipase A₂ and synapse damage

Abstract

Precisely how the accumulation of PrP^Sc causes the neuronal degeneration that leads to the clinical symptoms of prion diseases is poorly understood. Our recent paper showed that the clustering of specific glycosylphosphatidylinositol (GPI) anchors attached to PrP proteins triggered synapse damage in cultured neurons. First, we demonstrated that small, soluble PrP^Sc oligomers caused synapse damage via a GPI-dependent process. Our hypothesis, that the clustering of specific GPIs caused synapse damage, was supported by observations that cross-linkage of PrP^C, either chemically or by monoclonal antibodies, also triggered synapse damage. Synapse damage was preceded by an increase in the cholesterol content of synapses and activation of cytoplasmic phospholipase A₂ (cPLA₂). The presence of a terminal sialic acid moiety, a rare modification of mammalian GPI anchors, was essential in the activation of cPLA₂ and synapse damage induced by cross-linked PrP^C. We conclude that the sialic acid modifies local membrane microenvironments (rafts) surrounding clustered PrP molecules resulting in aberrant activation of cPLA₂ and synapse damage. A recent observation, that toxic amyloid-β assemblies cross-link PrP^C, suggests that synapse damage in prion and Alzheimer diseases is mediated via a common molecular mechanism, and raises the possibility that the pharmacological modification of GPI anchors might constitute a novel therapeutic approach to these diseases.

The key event in the prion diseases is the conversion of a normal host protein (PrP^C) into a disease-associated isoform (PrP^Sc).¹ The accumulation of PrP^Sc within the brain leads to the loss of synapses during the early stages of prion diseases.² Our interest in the molecular mechanisms of synapse damage was stimulated by observations that the loss of synapses is a better correlate of dementia than the death of neurons³ and that synapse damage can be reversible if the injurious stimulus is removed. Synapse loss can be detected through reductions in levels of the pre-synaptic membrane protein synaptophysin and, for example, the loss of synaptophysin in the brain closely correlates with clinical symptoms in Alzheimer disease.⁴ Here, synaptic density in neuronal cultures was measured using an ELISA for synaptophysin. The loss of synaptophysin in neurons was accompanied by the loss of other synaptic proteins, including synaptobrevin and synapsin-1, and while such an approach has its limitations, taken together we believe that these observations are a good indication of synapse damage in vitro. Our recent paper⁵ describes three major observations: first that small, soluble oligomers of PrP^Sc caused the activation of cytoplasmic phospholipase A₂ (cPLA₂) and triggered synapse damage in cultured neurons via a glycosylphosphatidylinositol (GPI)-dependent process; second that the cross-linkage of PrP^C had similar effects to PrP^Sc oligomers on cell signaling and synapse damage; and finally that cell signaling and synapse damage was dependent upon the presence of sialic acid in the GPI anchor attached to PrP proteins.

There is increasing interest in the role of GPI anchors in complex biological functions, including the regulation of membrane composition, protein trafficking and cell signaling.⁶ To examine the role of the GPI anchor that links PrP^C to cell membranes,⁷ it was digested with phospholipases (A₂ and C) or neuraminidase to create monoacylated, deacylated or desialylated forms of PrP^C respectively, modifications that could not be achieved by genetic manipulation methods. Previously, the role of GPI anchors in the biological activity of PrP^Sc had been difficult to determine as aggregated forms of PrP^Sc were resistant to digestion with phospholipases. This problem was overcome by combining a shaking technique developed to amplify small amounts of PrP^Sc for detection,⁸ with small PrP^Sc oligomers released from scrapie-infected GT1 cells (ScGT1 cells). The PrP^Sc oligomers in these supernatants were more potent at triggering synapse damage than the synthetic PrP peptides that are commonly used in toxicity studies. Thus, the synaptophysin content of neurons was reduced by 50% (EC₅₀) by 100 pg/ml of cell-derived PrP^Sc whereas the EC₅₀ of synthetic human PrP82–146 was 200 ng/ml; a thousand fold higher. This is consistent with similar observations that have been made between the toxicity of synthetic amyloid-β₄₂ peptides and cell-derived amyloid-β₄₂ oligomers⁹ used in Alzheimer disease research.

Critically the synapse damage induced by PrP^Sc oligomers was lost after digestion with phospholipases, enzymes that remove acyl chains from the GPI anchor without affecting protein structure. The self-aggregation of PrP^Sc results in the clustering of GPI anchors at high densities within cell membranes and as cell activation by some GPI-anchored proteins is associated with the protein clustering^10,11 we hypothesized that the synapse damage induced by PrP^Sc was mediated by the increased density of GPI anchors in cell membranes. To test this hypothesis we sought to replicate the effects of PrP^Sc by cross-linkage of PrP^C, a process that mimics the aggregation of GPI anchors without affecting the structure of PrP^C. The addition of cross-linked PrP^C caused synapse damage, whereas higher concentrations of mock-treated PrP^C had no significant effect. We also established that synapse damage induced by cross-linked PrP^C was dependent upon acylation of the GPI anchor. The cross-linked PrP^C did not share other properties of PrP^Sc, it remained sensitive to protease digestion and did not induce the production of protease-resistant PrP when incubated with cells capable of PrP^Sc formation. It was notable that higher concentrations of cross-linked PrP^C reduced neuronal viability as measured by thiazolyl blue tetrazolium (MTT).

The observation that cross-linked Thy-1 preparations did not cause synapse damage suggested that the composition of the clustered GPIs was critical for biological activity. The original investigations into the GPI attached to PrP noted the presence of sialic acid, a rare modification of mammalian GPIs.¹² In our study, the biotinylated lectins used to probe the glycan composition of GPIs showed that the GPI attached to neuron-derived PrP^C contained sialic acid. Furthermore, the removal of sialic acid from the GPI of PrP^C via neuraminidase digestion resulted in cross-linked PrP^C preparations that did not cause synapse damage. Consistent with observations that GPI composition was cell type specific¹³ the GPI attached to PrP^C derived from glial cells was found not to contain sialic acid and we consequently we showed that preparations of cross-linked glia-derived PrP^C did not cause synapse damage. Collectively these results indicate that the sialic acid moiety in the GPI anchor is critical in mediating synapse damage. The factors that determine the composition of GPI anchors are not well understood. We recently found that not all PrP^C produced by some neuroblastoma cell lines contained sialic acid and therefore hypothesize that PrP^Sc produced by these cells would cause less synapse damage than PrP^Sc produced in ScGT1 cells. Another factor that may affect toxicity is the homogeneity of PrP^Sc aggregates. We reported that cross-linkage of PrP^C with heterologous proteins (such as bovine serum albumin, Thy-1, monoacylated PrP^C or deacylated PrP^C) significantly reduced the extent of synapse damage when compared with cross-linked PrP^C homo-oligomers and concluded that both the composition and the density of GPIs are critical factors leading to synapse damage. These observations could help explain one of the conundrums in prion research, namely the lack of a direct correlation between pathology and the amounts of PrP^Sc in different prion disease models as some PrP^Sc oligomers may be more toxic than others.

Our study also investigated the mechanisms by which clustered GPIs might cause synapse damage. The presence of GPI anchors targets proteins including PrP^C and PrP^Sc to specific membrane micro-domains known as rafts,¹⁴ patches of membranes that are highly enriched in cholesterol and sphingolipids. The presence of GPI-anchored proteins is thought to help the formation of rafts as the saturated fatty acids within GPIs facilitate the solubilization of cholesterol in the membrane.¹⁵ The addition of PrP^Sc, or cross-linked PrP^C, increased the amount of cholesterol in synapse membranes in a GPI-dependent manner, as did preparations of isolated GPI anchors.¹⁶ However, cross-linked monoacylated PrP^C did not affect the cholesterol content of cell membranes, an observation that highlights the critical role of the two saturated fatty acids contained within the GPI anchor to sequester cholesterol and precipitate the formation of rafts.¹⁵ The change in cholesterol concentrations/formation of rafts alone was not sufficient to cause synapse damage, as cross-linked desialylated-PrP^C also precipitated an increase in synaptic cholesterol without affecting synaptophysin concentrations.

Rafts are heterogeneous and appear to have multiple functions. Some rafts are enriched with signaling molecules and can act as platforms in which the GPI anchors attached to PrP^C interact with cell signaling pathways.¹⁷ We studied the effects of PrP^Sc and cross-linked PrP^C on the activation of cytoplasmic PLA₂ (cPLA₂) for two reasons. First, because upon activation cPLA₂ migrates from the cytoplasm into PrP-containing rafts, and second because inhibition of cPLA₂ reduced synapse damage.¹⁸ Our study showed that both PrP^Sc and cross-linked PrP^C, but not native PrP^C, activated cPLA₂ within synapses. The activation of cPLA₂ was dependent upon the composition of the GPI and required both diacylation of the GPI anchor and the presence of a terminal sialic acid. The role of activated cPLA₂ in synapse damage was confirmed by our observation that pre-treatment of neurons with selective cPLA₂ inhibitors reduced PrP^Sc-induced synapse damage. Collectively, these observations underpin the hypothesis that the clustering of GPI anchors attached to PrP proteins caused activation of cPLA₂ leading to synapse damage. This hypothesis was tested further by loading neurons from Prnp knockout mice with PrP^C containing different GPIs, and cross linking PrP^C with mAbs. The cross-linkage of PrP^C by mAb 4F2 caused the activation of cPLA₂ and synapse damage (a finding we have also obtained with several other PrP^C-reactive antibodies) whereas the addition of Fab fragments (that do not cross-link PrP^C) had no significant effect. Second, we found that mAb-induced cross linkage of monoacylated PrP^C or desialylated PrP^C did not activate cPLA₂, nor did it cause synapse damage, confirming that it is the cross-linkage of PrP^C’s with specific GPIs that is necessary to trigger synapse damage.

So how does exogenously applied PrP^Sc or cross-linkage of PrP^C cause cell signaling? It is difficult to conceptualize how a protein attached to the outer cell membrane bilayer by phosphatidylinositol alone can influence cell signaling molecules associated with the cytoplasmic leaflet of the cell membrane. Under these circumstances most authors invoke the presence of a hypothetical transmembrane signaling protein that binds to PrP proteins and mediates cell activation. While the concept that PrP acts as a scaffold protein that helps assemble a signaling complex should not be ignored, it is not the only explanation. The selective associative properties of cholesterol, sphingolipids and GPI-anchored proteins are capable of altering raft composition.¹⁹ GPI anchored proteins are surrounded by a shell of membrane lipids, the composition of which is dependent upon the glycan and lipid components of the GPI anchor.²⁰ Since the composition and function of membrane rafts is controlled by an induced fit model,²¹ the aggregation of GPIs attached to PrP proteins could alter the composition of the underlying cell membrane. Such a process has been reported in T cell signaling studies where the cross-linkage of GPI anchored receptors at the cell surface influences the lipid and protein composition of the cytoplasmic leaflet of the cell membrane.^11,22 Although the precise role of sialic acid in the GPI is unknown, we recently discovered that rafts containing desialylated PrP^C contained more gangliosides than did rafts containing PrP^C (unpublished data). Such results suggest that the sialic acid on the GPI attached to PrP^C may be competing with sialylated gangliosides for specific partner proteins.

Two observations, that PrP^C acts as a receptor for toxic Aβ assemblies²³ and that Aβ oligomers cross-link PrP^C at synapses,²⁴ link the pathogenesis of prion and Alzheimer diseases. Thus, Aβ-induced cross-linkage of PrP^C mimics the effects of cross-linked PrP^C, self-aggregated PrP^Sc or PrP^C cross-linked by specific mAbs; all of which cause activation of cPLA₂ and trigger synapse damage. Such a model would predict that Aβ monomers would not cross-link PrP^C, activate cPLA₂ or cause synapse damage and would be consistent with reports that Aβ monomers are not toxic.²⁵ Such observations also raise the possibility that the pharmacological modification of GPI anchors might constitute a novel therapeutic approach to both prion and Alzheimer diseases. However, this approach may also change the normal function of PrP^C and should be regarded with caution at this stage.

In conclusion our study showed that the addition of PrP^Sc, cross-linked PrP^C or mAbs that cross-link PrP^C altered cell membranes increasing the activation of cPLA₂ and triggering synapse damage. Both the density and the composition of the GPI anchors were crucial to these processes that were dependent upon acylation and a terminal sialic acid moiety. These results raise the possibility that targeting the GPI anchor attached to PrP^C may reveal novel treatments for prion diseases and possibly Alzheimer disease