Special Reviews

Games Played by Rogue Proteins in Prion Disorders and Alzheimer's Disease

See allHide authors and affiliations

Science  31 Oct 2003:
Vol. 302, Issue 5646, pp. 814-818
DOI: 10.1126/science.1087348


The incidence of Alzheimer's disease (AD) and that of prion disorders (PrD) could not be more different. One-third of octogenarians succumb to AD, whereas Creutzfeldt-Jakob disease typically affects one individual in a million each year. However, these diseases have many common features impinging on the metabolism of neuronal membrane proteins: the amyloid precursor protein APP in the case of AD, and the cellular prion protein PrPC in PrD. APP begets the Aβ peptide, whereas PrPC begets the malignant prion protein PrPSc. Both Aβ and PrPSc are associated with disease, but we do not know what triggers their accumulation and neurotoxicity. A great deal has been learned, however, about protein folding, misfolding, and aggregation; an entirely new class of intramembrane proteases has been identified; and unsuspected roles for the immune system have been uncovered. There is reason to expect that prion research will profit from advances in the understanding of AD, and vice versa.

The key pathogenetic factors of both AD and PrD are neuronal membrane proteins: the amyloid precursor protein APP and the prion protein PrPC. Their adducts, Aβ and PrPSc, are major constituents of the deposits littering the brain of AD and PrD patients (Fig. 1). The role of Aβ and PrPSc in disease was validated by the discovery that mutations of the respective genes result in autosomal-dominant AD and PrD.

Fig. 1.

Both Alzheimer's and prion diseases are characterized by the deposition of pathological proteins in the brain, often in the form of plaques. The brown color is indicative of immunostained cortical deposits of the Aβ peptide and of the PrPSc protein in brains of patients suffering from Alzheimer's disease (A) and Creutzfeldt-Jakob disease (B), respectively. Scale bar: 20 μm.

In addition to APP, positional cloning of the genes causing familial AD led to the identification of the presenilins (1), which encode the presumed proteolytic core of the γ-secretase (2). This landmark discovery immensely accelerated AD research, and its reverberations span from developmental biology (3) to microbiology (46), basic cell biology (7), and protease biochemistry (4).

Two decades after Prusiner postulated that PrPSc is identical with the prion [i.e., the infectious principle (8, 9)], we still have an incomplete idea of the mechanism by which PrPC is converted into a pathogenic moiety. PrPC is necessary for replication of the infectious agent (10) and for pathogenesis (11). However, human genetics was less helpful in PrD than in AD; all cases of familial PrD segregate with PrPC mutations. No relevant genetic or physical interactors were identified. A report that histocompatibility loci would be linked to PrD susceptibility (12) was not confirmed by others (13). Thus, a number of basic questions are still open (Table 1).

Table 1.

Major open questions in prion science.

What is the nature of the infectious agent?
    Is PrPSc identical with the prion?
    How does PrPSc promote PrPC conversion into further PrPSc?
    Which other proteins assist in this process?
    Can we cure prion diseases by interfering with the conversion process?
    What are prion strains, molecularly speaking? How are the strain-specific properties encoded?
How do peripherally administered prions travel to the brain?
    Which molecules and cells are involved in neuroinvasion?
    Which inhibitory strategies have a realistic chance of succeeding?
What are the mechanisms of spongiform neurodegeneration?
    Which pathogenetic cascades are activated?
    What are the biochemical executioners of brain damage?
What is the physiological function of the normal prion protein, PrPC?
    PrPC is highly conserved, so it is probably useful.
    The Prnp gene was identified in 1985, and Prnp knockout mice in 1992, yet the function of PrPC remains unknown.
    None of the phenotypes ascribed to the absence of PrPC has been explained in molecular terms.
    Is abnormal function of PrPC involved in prion disease pathogenesis?

Despite fundamental differences in their biochemistry and genetics, the recent advances in prion and AD research discussed here suggest that AD and PrD converge in many pathogenetic aspects and may even be amenable to similar therapeutic principles.

Prions: A Progress Report

The bewildering microcosm of prion terminology. Prion nomenclature continues to be complex, inconsistent, and intimidating to outsiders. Here we use the term “prion” to denote the transmissible principle that causes transmissible spongiform encephalopathies (TSE), without implicitly assigning a specific structural hypothesis to this term. Instead, we refer to PrPSc as a modified form of the cellular protein PrPC, which accumulates in TSE-affected individuals (8).

PrPSc was originally defined as protease-resistant, aggregated PrP. However, PrPC may undergo disease-associated, diagnostically important modifications that do not lead to protease resistance (protease-sensitive PrPSc, or sPrPSc) (14). It appears advisable, therefore, to define PrPSc on the basis of generic disease-associated structural modifications, rather than protease resistance.

PrPSc (15) may or may not be identical with the prion. For all we know, the infectious principle may consist of (i) a PrPSc subspecies, (ii) an unstable intermediate provisionally termed PrP* (16), or a complex of PrPSc with (iii) some other protein that may be host-derived (17) or (iv) nonprotein compounds (18).

How do prions replicate? Thus, we still do not know whether the prion hypothesis is entirely correct (14). If the prion consists entirely of PrPSc, it may propagate by template-directed refolding or by nucleation (19). Does PrPSc catalyze the conversion of PrPC into more PrPSc by direct action, or does it take advantage of ancillary, hitherto unknown molecules? The second hypothesis is plausible, because innumerable attempts to “grow” prion infectivity in chemically defined cell-free systems have failed. Genetic evidence suggests that at least one further locus distinct from Prnp controls prion replication (17), but no physical evidence for such “protein X” has come forward in the past 8 years. Maybe advances of mass-spectrometric technologies, in concert with tagged PrPSc baits (20), will facilitate the resolution of this question.

Blocking prion replication. Several molecules have been identified as potential leads for inhibiting prion replication. Most therapeutic leads were identified in cell culture assays, in which chronically prion-infected neuroblastoma cells are “cured” of their PrPSc burden. A startling variety of substances appears to have such neuroblastoma-curing properties (2129).

Unfortunately, none of these compounds has, so far, proved particularly effective for therapy of sick animals, let alone of human patients. This raises troubling questions as to the predictive value of infected cell models for antiprion efficacy in vivo. Is it premature to treat patients with drugs on the sole basis of antiprion efficacy in neuroblastoma cells? Such a shortcut was taken in the case of the antimalarial drug quinacrine, which cures cells (30) but not scrapie-infected mice (31) or Creutzfeldt-Jakob disease (CJD) patients (32), and which is actually quite hepatotoxic (33).

In several paradigms, expression of two PrPC moieties subtly different from each other antagonizes prion replication, and humans heterozygous for a common Prnp polymorphism at codon 129 are largely protected from CJD. Similarly, transgenic expression of hamster PrPC renders Prnpo/o mice highly susceptible to hamster prions, whereas coexpression of mouse PrPC diminishes this effect. The molecular basis for these effects is unknown; perhaps the subtly modified PrPC acts as a “decoy” by binding incoming PrPSc and sequestering it in a complex incapable of further replication.

The latter hypothesis was tested by fusing an immunoglobulin Fcγ domain to PrPC (20), which yields a soluble, tagged dimeric PrPC immunoadhesin (34). When expressed in transgenic mice, the PrP-Fc2 fusion protein was found to compete with PrPC for PrPSc (Fig. 2) and to prolong the latency period of prion infection (20). Soluble prion protein mutants may therefore represent lead prionostatic compounds. Besides, PrP-Fc2 engages PrPSc in vivo, which makes it an interesting bait for recovering proteins associated with PrPSc, including the hitherto elusive “protein X,” which was proposed to mediate prion conversion (17), and maybe also “protein Y” and “protein Z,” if these happen to exist.

Fig. 2.

How does soluble dimeric PrP counteract prions? (A) The template refolding model of prion replication postulates a transient dimerization of PrPC and PrPSc. PrPSc may impart its own beta-sheet-rich, protease-resistant conformation onto PrPC. (B) In the absence of PrPC, soluble dimeric PrP does not support replication of the infectious agent or formation of a protease-resistant moiety. Although several lines of evidence indicate that it can associate with PrPSc, this association is nonproductive. (C) In mice coexpressing PrPC and soluble dimeric PrP, scrapie pathology is delayed, because soluble PrP sequesters PrPSc in an inactive complex.

Prion strains. Extracts prepared from the brains of mice inoculated with a particular preparation of prions produce a similar disease in inoculated recipients (3537). These strains differ in their incubation times, the distribution of central nervous system vacuolation that they produce, and whether or not they induce amyloid plaques. That such isolates could be propagated through multiple passages suggested that the prion has a nucleic acid genome that is copied onto nascent prions (36, 38), a hypothesis for which no evidence has come forth.

Currently, understanding the existence of prion strains appears harder than ever. Because strains are stable in inbred populations, whose genome is theoretically homogeneous, strains cannot be encoded by the host genome. Circumstantial evidence suggests that the strain phenotypes are enciphered in distinct tertiary or even quaternary structures of PrPSc—which exhibit, for example, distinct stabilities against chaotropic salts and thermal energy— but formal proof is still absent.

The strain problem is also dramatically important for public health. Whereas the prion strain associated with sheep scrapie is relatively harmless to humans, bovine spongiform encephalopathy (BSE) prions are promiscuous in their species tropism.

Some strains are more stable than expected. Mice inoculated with hamster prions live a long, TSE-free life, and most do not accumulate PrPSc. As shown by Chesebro, PrPSc-negative brain cells injected into other mice do not elicit clinical disease after >650 days (15). When cells from these brains were passaged to hamsters, however, rapid lethality resulted. Therefore, the agent had silently replicated for several years in mice but had maintained its full virulence toward hamsters. Other strains display aberrant behavior and may mutate to unexpected biochemical phenotypes (39). Viewed from the framework of the protein-only hypothesis, this result means that epigenetic strain characteristics of prions generally dominate over the primary prion protein sequence of the infected host and persist even when transmitted across species. The relevance to public health is obvious: BSE prions may maintain their malignant characteristics when passaged to other hosts, such as sheep.

The march of prions. Spongiform encephalopathy is most efficiently induced by intracerebral inoculation. In the field, however, acquisition of prion infectivity typically occurs through peripheral routes, including feeding (4042), intravenous and intraperitoneal injection (43), conjunctival instillation (44), and organ grafts (45). Understanding of how prions reach the brain from peripheral sites has progressed considerably: Lymphoid organs (46, 47), immune cells (48, 49), and peripheral nerves (50, 51) are involved.

Because the process of neuroinvasion takes a long time, there may be ways to achieve postexposure prophylaxis (52, 53). In transgenic systems, antibodies to PrPC can prevent peripheral prion replication (54) and may retard, to some extent, the onset of disease (55).

The executioners of brain damage. Brains of CJD patients look truly frightening. In heavily affected areas, there is hardly any neuron left, and the brain tissue texture is coarsened by the abnormal growth of astrocytes (“gliosis”) and microglial cells. The most telling hallmark is spongiosis, a peculiar microvacuolation affecting residual neural cells.

The molecular steps that emanate from prion replication and lead to such destruction are unknown. A lively discussion is ongoing on the role of abnormal PrPC topologies. Targeting of PrP to the cytosol results in rapidly lethal neurodegeneration (yet without PrPSc), and proteasome inhibition induces a slightly protease-resistant PrP species in cultured cells, which may be self-sustaining, at least for a while (56, 57). Therefore, prion toxicity may start with retrotranslocation of PrPC from the endoplasmic reticulum to the cytosol, in conjunction with impaired proteasomal function. On the other hand, PrPC was found to assume a transmembrane topology (CtmPrP) whose concentration correlates with neurotoxicity (58, 59).

Although PrP is clearly toxic in the cytosol, the details of how it may get there are hotly debated. Cytosolic PrP retains its secretory leader peptide and does not contain a glycosyl phosphatidyl inositol anchor, suggesting that it never enters the endoplasmic reticulum (60).

We still know nothing of the biochemical pathways leading to brain damage, be they triggered by cytoplasmic PrP or by CtmPrP. An understanding of these pathways may lead to the identification of therapeutic targets and may be applicable to other neurodegenerative diseases.

Why do we have prions? A variety of mild phenotypes has been described in Prnp-ablated mice, but the molecular steps leading to such phenotypes are unknown, and there is no unified understanding of PrPC function.

However, there appears to be considerable evolutionary pressure for maintaining the prion gene. Protective polymorphisms in the human prion gene were heavily selected for, possibly because of evolutionary pressure from cannibalism-propagated prions (61). If prion diseases were so frequent as to skew the distribution of Prnp alleles, though, why were Prnp–/– individuals not selected for? This circumstance argues for an important, hitherto unidentified function of the cellular prion protein (62).

Paradoxically, there is rather good knowledge of the domains necessary for the prion protein to exert its unknown function. Ablation of the amino terminal half of PrPC produces a highly neurotoxic moiety, yet toxicity is abolished by coexpression of full-length PrPC (63). This situation suggests competition for a common unidentified receptor. There happens to be a cellular protein structurally reminiscent of amino terminally truncated PrPC, which was aptly called Doppel (doppelganger of prion) or Dpl (downstream of the Prnp locus, because of its chromosomal location) (64, 65). In contrast to PrP, a function has now been assigned to Dpl: It is indispensable for male fertility (6668). Therefore, the obstinacy with which a function for PrPC was sought for in the brain may be out of place. Perhaps very different physiological functions will yield the answers.

Presenilins and the γ-Secretase Quartet in Alzheimer's Disease

AD researchers are arguably in a more comfortable spot than are their prion colleagues. Whereas the latter struggle with the most basic questions (including the actual physical nature of the prion), AD researchers avail themselves of an arsenal of molecules encompassing the APP substrate, three classes of cleaving enzymes termed α-, β-, and γ-secretases, and a sizable list of modifier loci. We focus here on the γ-secretase pathway, the understanding of which is progressing rapidly.

Intramembraneous endoproteolysis was thought to be biochemically impossible (69). γ-Secretase cleaves within the hydrophobic environment of membranes and consists of a quaternary complex. This quartet plays a pivotal role in the generation of the neurotoxic amyloid β-peptide (Aβ) of AD.

Here we discuss this novel aspartyl protease, which requires at least four components to do its job. This complex is required to liberate Aβ from the β-amyloid precursor protein (APP) C-terminal fragment (4), which in turn is generated by the β-secretase cleavage (70). Whereas β-secretase is a conventional aspartyl protease whose typical catalytic center contains the evolutionary conserved D(T/S)G(T/S) motifs (70), γ-secretase turns out to be a quite unusual aspartyl protease (2, 4).

The catalytic core. The presenilin genes, PS1 and PS2, were identified in 1995 (1). Mutations in either gene are associated with familial AD, and all of them affect the precision of γ-secretase cleavage (4): They shift the intramembraneous cleavage by two amino acids toward the C-terminus, favoring generation of the highly amyloidogenic 42 amino acid Aβ (Aβ42). This finding suggests that presenilins may represent the γ-secretase.

All presenilins undergo endoproteolytic cleavage within their large loop (71) (Fig. 3). The resulting N- and C-terminal fragments (NTF and CTF) remain associated (72), and this association seems to be critical for PS function (4). Direct evidence for PS presenting the catalytic core of γ-secretase came from the finding that all presenilins contain two functionally important aspartate residues within transmembrane domains 6 and 7 (TM6 and TM7), and consequently within the NTF and CTF (2) (Fig. 3). If the aspartates are mutagenized, endoproteolysis of PS and γ-secretase activity (and consequently Aβ production) are blocked. Thus, full-length PS may represent an inactive zymogen (Fig. 3), which is proteolytically cleaved to form the active NTF/CTF heterodimer. This idea is supported by the finding that γ-secretase inhibitors (some of which have the characteristics of typical transition analog state aspartyl protease inhibitors) could be cross-linked to the fragments (73, 74). Some γ-secretase inhibitors block PS endoproteolysis, suggesting an autoproteolytic cleavage mechanism (75). All this suggests that presenilins may be novel aspartyl proteases. However, presenilins do not contain the expected D(T/S)G(T/S) aspartyl protease active site. Instead, they contain the GxGD motif around a critical D in TM7 (4, 5). This motif is fully conserved between presenilins and the members of other unusual polytopic proteases, including the type 4 prepilin peptidases (6), the intramembrane cleaving signal peptide peptidases, and their close homologs (4, 7, 69).

Fig. 3.

Presenilins are cleaved (black arrow) to form the physiologically active N- and C-terminal fragments (NTF and CTF). Additional cofactors are required to form the fully active γ-secretase complex. These factors include APH1a/b, Nct, and PEN-2. The quartet acts as a native protease, liberating Aβ into the extracellular space, where it is deposited upon aggregation.

PS substrates are cleaved twice within their membrane domains. A number of substrates of presenilins have so far been identified (76). Besides APP, the Notch 1 to 4 signaling proteins figure among the most prominent substrates (77). Intramembraneous cleavage of Notch is required to liberate the Notch intracellular domain (ICD) for nuclear signaling (3, 77). This cleavage is fully PS dependent; a complete PS knockout, mutagenesis of the aspartate residues in TM6 or TM7, or γ-secretase inhibitors block Notch signaling (3, 4, 77). Other substrates include the APP homologs APLP-1 and -2, ErbB-4, E-cadherin, LRP, Nectin-1-α, the Notch ligands Delta and Jagged, and CD44 (76). Most substrates are type 1 transmembrane proteins, which shed their ectodomain before γ-secretase cleavage. Furthermore, γ-secretase appears to cleave its substrates twice (Fig. 4) (78). Aβ and the corresponding Aβ-like domains of other substrates are generated by a cut in the middle of the transmembrane domain. A second entirely PS-dependent cut, much closer to the cytosolic face of the membrane, appears to liberate ICD (Fig. 4). Thus, γ-secretase generates its products (Aβ, Aβ-like peptides, and ICDs) by a dual cleavage mechanism (4). The order of these cleavage events is currently unknown and, with the exception of Notch ICD, little is known about the target of the many other ICDs.

Fig. 4.

A dual PS-dependent cleavage mechanism liberates Aβ and Aβ-like peptides into the extracellular space, and ICD into the cytoplasm. The fate of the small peptide labeled with “?” is unknown.

Complex composition and coordinated regulation. γ-Secretase activity is associated with a high molecular weight PS-containing complex (4, 72, 76). Biochemical purification of this complex led to the identification of the first γ-secretase complex component, termed nicastrin (Nct) (79). Genetic screens for enhancers of a PS-dependent Notch-deficient phenotype in Caenorhabditis elegans led to the identification of two additional components: APH-1 (anterior pharynx-defective phenotype) and PEN-2 (PS-enhancer) (Fig. 3) (76, 80, 81). Because these genetic screens were probably saturating (81), no additional components of equal functional importance are to be expected. Accordingly, these four components are sufficient for reconstituting γ-secretase–complex formation, PS endoproteolysis, and Aβ and ICD production in Saccharomyces cerevisiae, an organism that lacks endogenous γ-secretase (82). Coexpression of PS, APH-1, PEN-2, and Nct in cultured mammalian and fly cells also enhances γ-secretase activity (8385). However, this finding does not exclude the possibility that additional regulatory subunits may exist, which may, for example, modulate substrate specificity.

PS expression and γ-secretase activity are regulated in a coordinated way. Because overexpression of PS does not lead to the expression of higher levels of PS fragments (71), nor to enhanced γ-secretase activity, a limiting factor regulating γ-secretase activity was postulated. Maybe there is no defined limiting factor; instead, the least abundant γ-secretase complex component, be it PS, PEN-2, APH-1, or Nct, may become limiting. Thus, all four components are equally important for the assembly of the complex (8285). Only when all four components are coexpressed is it possible to achieve higher γ-secretase activity and to stabilize γ-secretase complex components (8385). If there is a disequilibrium of the components, excess amounts may be destabilized or, in the case of Nct, just fail to mature (8694).

A variety of different γ-secretase complexes may be assembled (76). For example, PS1 and PS2 contribute to independent γ-secretase complexes (90, 95). APH-1a, APH-1b, and the splice variants of APH-1a may also contribute to individual complexes (94). However, because the APH-1 isoforms coordinately regulate their own expression, they may exhibit closely related biological functions (94). Furthermore, the molecular weight of the complex is being debated (76). A 1:1:1:1 stoichiometry of the four components would be consistent with the reported molecular weight of ∼200 to 250 kD (83), but others reported a molecular weight of the endogenous complex of ∼500 to 600 kD (86, 96, 97), which suggests either additional components (an unlikely proposition, because of the saturating genetic screens and the reconstitution in yeast) or a dimeric complex (98). Maybe two individual presenilins contained within one dimeric γ-secretase complex are required for executing the two individual cuts characteristic of many if not all γ-secretase substrates. Finally, the fully functional γ-secretase complex may be required to activate a premature complex by inducing PS endoproteolysis. This scenario would require dimerization of two complexes, which may be stabilized upon endoproteolytic activation of the binding partner.

The function of individual γ-secretase complex components. Although the composition and the biological function of the γ-secretase complex seem to be largely solved, little is known about the function of the individual components because of the difficulties in analyzing these components individually in the absence of the other partners. Clearly, every single component is absolutely required and equally important for γ-secretase function.

Nct is inserted into the endoplasmic reticulum and matures by complex glycosylation during its trafficking through the secretory pathway (8688, 99). Elimination of PS, APH-1, or PEN-2 abolished maturation of Nct by complex glycosylation (8694). However, glycosylation is clearly not intrinsically required for Nct function within the γ-secretase complex (89, 100). During its maturation, Nct undergoes a major structural alteration within its large luminal domain (100). The entire luminal domain is required for Nct activity within the complex, and it is tempting to speculate that a globular, tightly folded ectodomain measures the length of the membrane-retained stublike substrates. Such a mechanism must be postulated for one of the γ-secretase complex components, because substrates are only recognized and proteolyzed if their ectodomain was previously truncated (101). Nct would be a natural candidate, because it is the only γ-secretase complex component with a luminal domain large enough to interact with the luminal domain of putative substrates.

The existence and composition of putative precomplexes is less clear. Precomplexes composed of full-length PS, Nct, and APH-1 (84) or Nct, APH-1, and probably PEN-2 (85) have been suggested. In the latter case, PS would displace APH-1 from the mature complex (85). However, these apparently contradictory findings remain to be verified.

In the absence of PEN-2, PS is stabilized as a full-length protein with the possible help of APH-1 (84, 92). Only upon the expression of PEN-2 is PS efficiently endoproteolyzed (84, 85, 92), so that full γ-secretase activity can be obtained. Thus, APH-1 appears to be a stabilizer of full-length PS, whereas PEN-2 is required to initiate its endoproteolysis and consequently its activation. Although PEN-2 is proposed to bind preferentially to full-length PS (92), it remains bound to the activated γ-secretase complex containing the PS fragments (83, 84, 90). Thus, PEN-2 may have a function independent of its proposed activity in stimulation of PS endoproteolysis.

An important question, however, remains. Evolution created the γ-secretase quartet to achieve intramembrane proteolysis of truncated type 1 membrane proteins. Why do presenilins need the additional three cofactors, whereas the other intramembraneous aspartyl proteases apparently perform their functions by themselves?

Diversity and Unity

Both prion and AD research enjoy rapid progress, which is fueled by technologically sophisticated, multidisciplinary approaches. Only 8 years after presenilins were discovered, the composition of the γ-secretase complex was solved, its reconstitution established, and its activity blocked by highly selective small compounds. Likewise, some of the major culprits in prion replication have been identified, and we are witnessing the first successful attempts at interfering with their functions.

Regrettably, the parallels between the AD and prion fields extend to areas in which progress has been slow. For all the granularity of the molecular events that we have described, we still lack a mechanistic understanding of how these diseases induce brain damage.

Failure to understand the downstream effectors of neurotoxicity is not limited to AD and PrD. It applies also to Parkinson's, Huntington's, and many other neurodegenerative diseases. Such predicaments make it tempting to search for general hypotheses of pathogenesis. Because a large share of neurodegenerative diseases go along with inappropriate protein deposits, we (102) and many others have proposed that dangerous liaisons between aggregated proteins are a generic unifying trait of many different syndromes.

Upon closer inspection, such simplified views may look naïve. Or do they? Recent work has identified a conformation-dependent structure common to soluble Aβ oligomers (103). Bewilderingly, antibodies to this structure cross-react to amyloid oligomers with unrelated sequences. The latter finding is particularly important in view of the accruing evidence that immunotherapy may be applicable to both AD (104) and prion diseases (54). Such combined approaches, in our view, hold promise of putting an end to the games played by the rogue proteins in the brain.


View Abstract

Navigate This Article