Cellular prion protein: A co-receptor mediating neuronal cofilin-actin rod formation induced by β-amyloid and proinflammatory cytokines

Abstract

Increasing evidence suggests that proteins exhibiting “prion-like” behavior cause distinct neurodegenerative diseases, including inherited, sporadic and acquired types. The conversion of cellular prion protein (PrP^C) to its infectious protease resistant counterpart (PrP^Res) is the essential feature of prion diseases. However, PrP^C also performs important functions in transmembrane signaling, especially in neurodegenerative processes. Beta-amyloid (Aβ) synaptotoxicity and cognitive dysfunction in mouse models of Alzheimer disease are mediated by a PrP^C-dependent pathway. Here we review how this pathway converges with proinflammatory cytokine signaling to activate membrane NADPH oxidase (NOX) and generate reactive oxygen species (ROS) leading to dynamic remodeling of the actin cytoskeleton. The NOX signaling pathway may also be integrated with those of other transmembrane receptors clustered in PrP^C-enriched membrane domains. Such a signal convergence along the PrP^C-NOX axis could explain the relevance of PrP^C in a broad spectrum of neurodegenerative disorders, including neuroinflammatory-mediated alterations in synaptic function following traumatic brain injury. PrP^C overexpression alone activates NOX and generates a local increase in ROS that initiates cofilin activation and formation of cofilin-saturated actin bundles (rods). Rods sequester cofilin from synaptic regions where it is required for plasticity associated with learning and memory. Rods can also interrupt vesicular transport by occluding the neurite within which they form. Through either or both mechanisms, rods may directly mediate the synaptic dysfunction that accompanies various neurodegenerative disorders.

Keywords:

• actin
• Alzheimer disease
• cofilin
• caveolin-1
• cellular prion protein
• lipid rafts
• oxidative stress
• NADPH oxidase
• proinflammatory cytokines

Introduction

The cellular prion protein (PrP^C) is expressed in many cell types but is particularly enriched in neurons. Here we review evidence that PrP^C has a role in transmembrane signaling in neurons. We will discuss a possible mechanism by which diverse extracellular ligands, such as amyloid-β and proinflammatory cytokines, utilize PrP^C to induce an identical alteration in the actin cytoskeleton that affects synaptic function.

The encoded PrP^C contains an N-terminal signal peptide removed after transport across the endoplasmic reticulum and a C-terminal peptide that is removed when the protein is transferred to its glycosylphosphatidylinositiol (GPI) lipid that keeps it anchored in the outer leaflet of the plasma membrane (Fig. 1). Near the N-terminus of the processed protein are 4 to 5 octapeptide repeats that contain up to 5 copper binding sites.¹ Copper binding induces significant conformational changes to PrP^C, which may have important implications in prion formation and in modulating PrP^C interactions with other membrane proteins.² Neuronal, GPI-anchored PrP^C serves as a relay for transmissible spongiform encephalopathy (prion disease)³ induced by the β-sheet-containing form of the prion protein, referred to as PrP^Sc (scrapie) or PrP^Res (protease resistant). The C-terminal region of PrP^C contains a sphingolipid-binding domain and 3 α helices that undergo a conformational transformation to β-sheet when the PrP^C interacts with an appropriate PrP^Res. The converted PrP^Res self-assembles into fibrils and insoluble aggregates that accumulate in brain. Although PrP^C is highly conserved across mammalian species, only a few amino acid differences, of which many may be conservative, are sufficient to provide a species barrier for prion disease transmission.

Figure 1. Schematic of structural domains and interactions of PrP^C. PrP^C contains an N-terminal signal sequence (1–22) and 2 charged clusters (CC1/2) on either side of 5 copper-binding octapeptide repeats (OR). Two principal binding sites have been identified for Aβ (23–50 and 92–110) whereas caveolin-1 has been shown to interact with ORs, where it may serve to link cellular prion to the activation of fyn tyrosine-kinase. The C-terminal domain contains 3 α helices, 2 of which are stabilized by a disulfide bond, while a third (α1) unfolds and is converted to the PrP^Res conformation. A sphingolipid binding domain is located between the 2 cysteine residues (179 and 214). GPI attachment at amino acid 231 anchors PrP^C to the extracellular leaflet of the plasma membrane upon elimination of the C-terminal peptide (232–254).

PrP^C contains 2 potentially glycosylated asparagine residues,² undergoes 3 different proteolytic cleavages, and thus, exists as multiple post-translationally modified isoforms whose distribution and relevance may vary based on distinct brain region or cell type. This concept of prion speciation might account for its ability to affect specific neuronal populations and thus play a role in a spectrum of neurodegenerative disorders beyond those induced by infectious prions. These include Alzheimer disease (AD), Parkinson disease, Huntington disease, other frontotemporal dementias such as chronic traumatic encephalopathy, as well as traumatic brain injury (TBI).^4,5 Although PrP^C is widely expressed in neurons, there are certain neuronal populations and neurite populations in which its expression is below detection.⁶ In addition and of relevance to implications on synaptic function, PrP^C was found highly enriched in post-synaptic densities.⁷

Knockout of the PrP^C-encoding gene Prnp in mice results in viable animals that develop normally but which are resistant to prion infection.² Although the animals show increased vulnerability to various types of stress, specific roles for the PrP^C in normal function have been difficult to identify. However, PrP^C was suggested to be a protein involved in signal transduction many years ago⁸ and recent work supports a role for PrP^C as a receptor or co-receptor to mediate signal transduction processes, some of which may be especially important in the neurodegenerative cascade leading to loss of synaptic function.

PrP^C is a component of distinct membrane microdomains

Transmembrane and membrane associated proteins have lipid interacting domains of various types that are important for their conformation and function. Many membrane proteins have transmembrane segments that are accommodated better within slightly thicker membrane microdomains, often referred to as lipid rafts, containing phosphatidylinositols and cholesterol and in which sphingolipids are concentrated in the exoplasmic face of the plasma membrane. Although their exact character is debated, lipid rafts are spatially and temporally dynamic and play a vital role in many cellular signaling pathways, including those of cytokines, growth factors, and hormones.⁹ GPI-linked proteins, such as PrP^C, are also concentrated in these domains. Specific hydrogen bonding between cholesterol and sphingomyelin appears critical for the formation and stability of lipid rafts. The biophysical properties of the lipid raft environment affect the recruitment, conformation, interaction, and physiological function of raft-associated proteins. Several ligand-gated ion channels and G-protein coupled receptors contain specific binding sites, either in their transmembrane helices or extracellular loop structures, for cholesterol and sphingolipids, the binding of which allows conformational changes necessary for function.¹⁰ Subtle changes in lipid species alter membrane architecture and impose constraints on lipid-protein and/or protein-protein interactions influencing the function of membrane proteins.

Levels of raft-associated lipids such as sphingomyelin, ceramides, and cholesterol significantly increase with age,¹¹ suggesting that lipid raft densities and associated signaling process are likely subjected to age-dependent alterations that could contribute to the initiation and/or progression of AD,⁹ as could dietary factors that affect lipid accumulation and metabolism and alterations in membrane lipids resulting from oxidative stress (e.g. lipid peroxidation).¹² Although the lipid raft microenvironment is essential for the conversion of PrP^C to the neurotoxic PrP^Res,¹³ perhaps through the GM1 ganglioside binding epitope in the C-terminal portion implicated in conferring structural changes to the prion protein,¹⁴ the interactions of PrP^C with both lipid and protein components within the membrane microdomains also are likely to influence the signaling output of the system.

PrP^C links Aβ and proinflammatory cytokines to cognitive impairment

GPI-anchored PrP^C serves as a receptor for Aβ oligomers^2,7 (Fig. 1) and is required for the synaptotoxicity and cognitive deficits of Aβ oligomers in mice.^2,7,15 Cognitive decline in human subjects appears to depend more on the length (Aβ_1–42 vs shorter forms) and post-production alterations of Aβ (formation of dimers/oligomers) than simply the total amount produced.¹⁶ SDS-stable dimers are a major form of soluble Aβ extracted from most AD brains and may be the essential building block of the most synaptotoxic Aβ species that can induce dendritic spine loss and AD progression.^16,17

Synapse loss can be detected in amyloid precursor protein (APP)_swe/presenilin 1 (PS1)ΔE9 mice by staining for synaptic marker proteins, and the loss of synaptic markers is dependent on the presence of PrP^C.⁷ Treatment of aged APP_swe/PS1ΔE9 mice with anti-PrP^C antibodies allows a recovery of synaptic markers in the dentate gyrus of the hippocampus, a brain region associated with learning and memory.¹⁸

Neuroinflammation occurs in TBI and many other neurodegenerative disorders. Proinflammatory cytokines, including tumor necrosis factor α (TNFα) and interleukin (IL)-1β, are increased in mild TBI and modulate deficits in motor coordination.¹⁹ Although the molecular mechanism by which the proinflammatory cytokines mediate the decline in cognitive function associated with the neuroinflammatory responses in mild TBI is incompletely understood, neuronal receptors for these cytokines are concentrated in lipid raft domains, where they enhance redox-signaling events.^20,21

PrP^C dependency for cofilin-actin rod formation induced by Aβ and proinflammatory cytokines

Dynamic remodeling of the actin cytoskeleton into rod-like neuritic inclusions may be a final common mechanism for synaptic dysfunction in many neurodegenerative disorders. Activation (dephosphorylation) of actin-depolymerizing factor (ADF)/cofilin, a family of proteins that enhances the severing and turnover of actin filaments, initiates the formation of short, stable ADF/cofilin-saturated actin filaments, which under oxidative conditions can bundle to form rod-like inclusions (rods) within neurites (Fig. 2).²² Cofilin is 5–10 fold more abundant in mammalian neurons than ADF, thus the rods are called cofilin-actin rods. Rod-like neuritic inclusions have been observed in human AD brain and in mouse models of AD.^23,24 Rods can grow to occlude neurites, blocking vesicular transport necessary for synaptic function.^23,25,26 Specifically, formation of rods results in a global loss of trafficking of amyloid precursor protein (APP)-containing vesicles²⁷ that normally traffic both anterogradely to the plasma membrane and retrogradely for lysosomal degradation.²⁸ Rods may also mediate synaptic dysfunction by sequestering ADF/cofilin. In dendritic spines, cofilin functions in many aspects of receptor insertion and spine dynamics associated with memory and learning, whereas ADF/cofilin together appear to function in presynaptiC-terminals.²⁹

Figure 2:. Schematic model of PrP^C-mediated membrane signaling complex for Aβ and proinflammatory cytokine-induced cofilin-actin rod formation. This proposed hypothetical model links Aβd/t and the proinflammatory cytokine TNFα to cofilin-actin rod formation in membrane raft domains in which PrP^C is highly concentrated. Overexpression of PrP^C alone is sufficient to induce rods in a NOX-dependent manner, suggesting that enlargement of membrane signaling domains from which resultant ROS generation exceeds a rod-inducing threshold is regulated by PrP^C levels. This inducible oxidizing environment both maintains dephosphorylation-dependent activation of cofilin and initiates intermolecular disulfide bond formation in dephosphorylated cofilin to permit bundling of cofilin-saturated actin filaments. The dynamics of rod dissolution and reversal upon addition of several different NOX inhibitors further suggests that rods may be transient in nature until such time as the stimulus for their formation maintains ROS above the threshold required for rod maintenance.

Rod formation is triggered in cultured neurons by Aβ oligomers, but especially by those forms most associated with AD progression, soluble Aβ dimer/trimer (Aβd/t).^30,31 Rods are also induced by several proinflammatory cytokines in the same neuronal subpopulation that responds to Aβd/t, and both inducers require the expression of PrP^C for rod signaling.³¹ Formation of rods downstream from PrP^C is dependent on the activation of NADPH oxidase (NOX) for production of reactive oxygen species (ROS), which is required for oxidation of cofilin sulfhydryls to form an intermolecular disulfide bond.^22,31 Rods induced by Aβ or proinflammatory cytokines are reversible and their stability appears to depend upon the continuous production of ROS above a threshold level to maintain them. Addition of NOX inhibitors or washout of the inducer reverses rod formation with a similar half-life of about 35 min.³¹

Previous work has demonstrated that several different Aβ-binding proteins are able to mediate cognitive decline,² including at least one (mGluR5) for which an association with PrP^C has been established.⁷ Each of the cytokines shown to induce rods in the same population of neurons as Aβ has its own unique receptor and all of them use a PrP^C-dependent pathway for rod formation that involves NOX activation and ROS production. NOX activity in human brain is inversely correlated with cognition³⁰ as its inhibition had beneficial effects on various central nervous system pathologies as well as inhibition and reversal of PrP^C-induced rods in cultured neurons.³¹ Thus, rod formation represents a common mechanism by which a multitude of membrane proteins that bind the promiscuous Aβ could work through a PrP^C-dependent signaling pathway to affect synapse plasticity and ultimately cognitive function.

PrP^C signaling in synaptic loss

Surprisingly, the overexpression of PrP^C alone is sufficient to induce cofilin-actin rods.³¹ At low levels of PrP^C-overexpression, rod formation is enhanced by the presence of a stressor such as Aβd/t or TNFα, but at higher levels of expression, no additional rod formation occurs with stressor addition. Rods induced by PrP^C-overexpression are reversed by inhibiting NOX activity, suggesting the clustering of GPI-anchored PrP^C into membrane domains independently recruits and activates NOX for ROS production. This idea is reinforced by studies that show the clustering of GPI-linked PrP^C alone alters the membrane lipid profile and results in density-dependent synaptic damage.³² Both Aβ and TNFα activate neutral sphingomyelinase generating ceramide and significantly changing the lipid raft environment.^21,33 Such changes could contribute to the proposed signaling pathway by altering interactions between raft components.

NOX-mediated ROS production occurs downstream of fyn, a non-receptor tyrosine kinase that is activated in a PrP^C-dependent manner by Aβ⁷ or by antibody cross-linking.² Fyn activation occurs mainly in neurites where a conformational change in PrP^C results in its octapeptide repeat region interacting directly with the hairpin-like membrane-embedded protein caveolin-1.^2,34 Both of these interactions result in caveolin-1 and fyn phosphorylation. Thus, it is easy to visualize how alterations in metal ion binding to the octapeptide repeats or in the lipid composition of raft domains could influence PrP^C signaling. Fyn activation through PrP^C may mediate both neural cell adhesion molecule-dependent neurite outgrowth and synaptic dysfunction.² The downstream effects of PrP^C-fyn signaling, which includes NOX activation by phosphorylation and enhanced recruitment of cytoplasmic NOX subunits, are enhanced by Aβ, which clusters and crosslinks PrP^C at the neuronal cell membrane.^2,32 Aβ-dependent PrP^C-fyn signaling has been tied to the overactivation of metabotropic mGluR5 receptors as well as the NMDA receptor^2,7 and may utilize a Ca²⁺-dependent mechanism for synaptic dysregulation.

Although the specific pathways utilized for Aβ or proinflammatory cytokine activation (dephosphorylation) of cofilin have not been conclusively established, several have been characterized in neurons and may be utilized. The cofilin phosphatase slingshot (SSH) is activated by ROS via the oxidation-dependent removal of 14-3-3 isoform which is inhibitory to SSH activation.²⁵ SSH can also be directly activated by dephosphorylation via the Ca²⁺-dependent phosphatase calcineurin. Alternatively, the cofilin phosphatase chronophin is activated by release from Hsp90 which occurs upon localized ATP depletion.³⁵ In addition, there are potential cofilin-activating pathways downstream of fyn.

About 20% of cultured hippocampal neurons globally treated with Aβd/t or proinflammatory cytokines form rods and these rods form in a subset of neurites in which cofilin activation occurs. More than 40% of neurons will form rods in response to the overexpression of PrP^C in the absence of any additional treatment. Although there is a small but significant (P < 0.001) increase in the number of rods per cell body (from 3.5 in Aβ-treated neurons to 3.9 in neurons overexpressing PrP^C), the number of neurites in which rods form does not significantly increase (average of 1.54 neurites per neuron with Aβ treatment versus 1.59 in PrP^C overexpressing neurons). However, in >80% of neurons exposed to excitotoxic levels of glutamate or that are energy-depleted (i.e. mitochondrial inhibition), rods form within minutes and are present in most neurites. This rapid pathway for rod formation does not require the presence of PrP^C or NOX activity.³¹ The above findings suggest that the components of rods are not limiting but that for PrP^C-dependent rod formation some other factor downstream of PrP^C restricts the signaling pathway to specific neurites in a subset of neurons. The nature of this factor is unknown but it could be caveolin-1, whose presence in neurons was debated for many years because it went undetected by immunostaining.

PrP^C-fyn signaling also promotes tau hyperphosphorylation in response to oligomeric Aβ.² The hyperphosphorylation of tau follows its release from microtubules and enhances its assembly into paired helical filaments (PHFs). The PHFs aggregate to form neuropil threads and neurofibrillary tangles, the latter of which, along with amyloid plaques, are considered the pathological hallmarks of AD, but do not always correlate well with cognitive status.^36,37 However, cofilin-actin rods can disrupt microtubules, causing the release of tau.²³ Because rods are reversed upon the removal of the inducing agent or a decline in ROS below the threshold required for rod maintenance, rods may initiate many changes in neurons but not be present throughout the degenerative cascade.^27,31

Concluding Remarks

Cofilin-actin rod formation provides a common end point for signaling from multiple receptors that interact with PrP^C to stimulate ROS production from NOX. That overexpression of PrP^C alone initiates this pathway suggests that the role of Aβ and other ligands (e.g., proinflammatory cytokines) is to enhance the interactions of membrane raft components for NOX activation (Fig. 2). Since this proposed signaling pathway occurs in the lipid raft environment, distinct lipid species could influence either positively or negatively the conformational changes of proteins involved.

AD is a disease of neuronal connectivity with synapse loss its earliest feature. Given that sporadic AD is the most common form and is likely initiated by a diversity of factors, it would be naïve to think that rods explain all of the synaptic dysfunction that arises. Nevertheless, the rapidity with which rods form, their ability to reverse upon reduction in ROS, and their ability to become persistent with continued elevation of ROS suggest that rods could be an important effector of neurodegenerative signaling and synapse dysfunction. The requirement for a ROS threshold to maintain rods might explain why a decrease in a number of different raft-associated Aβ binding partners has beneficial effects on cognitive function in mouse models.

Acknowledgment

We thank Laurie Minamide, Alisa Shaw and Dr. Mark Zabel for comments and critical reading of the manuscript.

Funding

Supported in part by a grant from National Institutes of Health (AG044812 to JRB).