Prion Diseases

Stephen J. DeArmond and Stanley B. Prusiner

The prion diseases, sometimes referred to as the "transmissible spongiform encephalopathies," include kuru, Creutzfeldt–Jakob disease (CJD), and Gerstmann– Sträussler–Scheinker disease (GSS) of humans as well as scrapie and bovine spongiform encephalopathy (BSE) of animals (Table 1 and Table 2). For many years, prion diseases were thought to be caused by viruses despite intriguing evidence to the contrary (1). The unique characteristic common to all of these disorders whether sporadic, dominantly inherited, or acquired by infection is that they involve the aberrant metabolism of the prion protein (PrP) (59). In many cases, the normal cellular prion protein, PrPC, is converted into an abnormal, protease-resistant isoform, PrPCJD in humans and PrPSc in animals, by a post-translational process which involves a conformational change. The particle (prion) which transmits scrapie appears to be composed largely, if not exclusively, of PrPSc (58). In the case of human prion diseases, many sporadic and familial cases have been transmitted to experimental animals (30, 34, 50).

Because of the complexity of prion diseases, the nomenclature to designate the wild-type and mutant prion protein genes, the pathogenic prion protein synthesized by humans and animals, the prion protein associated with the infectious prion and different prion isolates, and the normal and artificial prion proteins expressed in transgenic mice are currently evolving. A provisional, simplified nomenclature is used here. PrPC designates the normal cellular isoform synthesized constitutively by humans and animals. PrPSc designates pathogenic PrP synthesized in any animal whether its synthesis is initiated by inoculation with human or animal prions or by expression of a mutated PrP gene. PrPCJD designates pathogenic PrP synthesized in humans irrespective of whether the human prion disorder is sporadic, familial, or acquired by prion infection. The human PrP gene maps to the short arm of chromosome 20 and is designated PRNP. The mouse PrP gene maps to the homologous region of chromosome 2 and is designated Prn-p (67). Transgenic (Tg) mice expressing foreign or mutant PrP genes are labeled with the transgene in parentheses. For example, transgenic mice harboring a mouse (Mo) PrP gene which mimics the codon 102 mutation of human GSS, which causes a leucine for proline substitution, is designated Tg(MoPrP-P102L) (see Basic Concept and Techniques in Molecular Genetics, for background).

Until recently, only two familial forms of human prion disease were recognized based on dominant inheritance, transmissibility to animals, and the type of amyloid plaques: familial CJD (50, 52) and Gerstmann–Sträussler–Scheinker syndrome (GSS) (50). By definition, the diagnosis of GSS requires multicentric PrP amyloid plaques. GSS cases in which ataxia predominates have been linked to a mutation in PRNP gene codon 102 which leads to a proline-to-leucine substitution (41). The cause–effect relationship between the codon 102 mutation and prion disease was verified in Tg mice expressing a mouse (Mo) PrP transgene mimicking the human codon 102 mutation, Tg(MoPrP-P102L) mice (45). The founder and its offspring expressing the transgene developed a spontaneous neurodegenerative disorder with spongiform degeneration and multicentric amyloid plaques similar to the neuropathology in human GSS. The primarily dementing form of GSS has been linked to a PRNP gene codon 117 mutation (22, 44). A PRNP gene codon 200 mutation has been identified in some familial CJD pedigrees (28, 35, 39, 43). This mutation among Libyan Jews represents the largest focus of CJD in the world, with an incidence about 100 times greater than in the worldwide population (52). To date, this mutation has been found in 51 Libyan Jews with CJD and except for one homozygous patient, all were heterozygous for the mutation (29). In addition to point mutations in the PRNP gene, octapeptide-repeat insertions at codon 53 have been identified in other families with CJD (38, 54). Molecular genetics has led to the discovery of new prion disorders. PRNP gene codon 198 and 217 mutations (21, 42) have been found in a unique form of GSS in which Alzheimer's disease-like neuropathological changes, including neuritic plaques and neurofibrillary tangles, are associated with deposition of PrP amyloid plaques and not the bA4 peptide (31, 32, 68). These unique pedigrees raise questions about the relationship of the bA4 peptide deposition and senile plaques in Alzheimer's disease and suggest that there is an overlap of pathogenic mechanisms in Alzheimer's disease and prion diseases. A codon 178 mutation (51) has been found in families with fatal familial insomnia; however, the same codon 178 mutation has also been described in some familial CJD pedigrees without features of insomnia (37). Although the mutation at codon 178 causes an Asn-for-Asp substitution in both FFI and 178-CJD, the two diseases differ in clinical presentation and distribution of neuropathology. In FFI, neuropathological changes were confined largely to the mediodorsal and anterior ventral nuclei of the thalamus whereas they were widespread in 178-CJD. Subsequently, in a collaborative molecular genetic effort, it was discovered that the 178 mutation in FFI is also associated with a methionine (Met) polymorphism at codon 129 (37). Furthermore, it was reported that methionine homozygosity at codon 129 in FFI results in a more severe form of FFI, and valine (Val) homozygosity results in a more severe form of 178-CJD.

Therefore, molecular genetic studies indicate that both CJD and GSS are not single, discrete disorders, but instead are both syndromes with multiple molecular, clinical, and neuropathological characteristics. While we only know of familial forms of GSS, those disorders we classify as CJD because of common clinical and neuropathological features are clearly more complicated because they can have familial, infectious, or sporadic etiologies.

Much has been assumed about the pathogenesis of infectious, genetic, and sporadic human prion diseases from studies of infectious forms of scrapie in experimental animals and, more recently, by creating genetic forms of prion disease in Tg mice. As a result of these studies, it appears that abnormal forms of the prion protein play the predominant role in both the etiology and pathogenesis of all forms of prion disorders. The data argue that a protease-resistant form of the prion protein, PrPSc, is the sole functional component of the infectious particle, termed a prion, which transmits scrapie (58). Multiple studies have verified that PrPSc purifies with scrapie infectivity and have shown that procedures which hydrolyze or modify proteins inactivate prions (26, 27, 58). In contrast, those procedures that alter nucleic acids do not affect infectivity (1, 3, 58). Physicochemical analysis of nucleic acids in highly purified prion preparations have failed to reveal viral-like polynucleotides; however, small nucleic acid fragments were found, the vast majority of which were between 40 and 80 nucleotides in length (48). Whether these nucleic acids are contaminants or a functional component of the prion is unknown. It has been proposed that an accessory cellular RNA, designated a "co-prion," can modify the properties of PrPSc to account for strain-like differences among different prion isolates; however, the failure of ultraviolet (UV) radiation to modify the properties of prions argues against the nucleic acid co-prion hypothesis.

PrPC is synthesized by a single copy cellular gene in which the entire open reading frame is in a single exon (2, 53). PrPC is distinguished from PrPSc in several ways. PrPC is completely digested by limited hydrolysis with proteinase K, whereas it only removes the N-terminal 67 amino acids from PrPSc to yield PrP 27-30 without loss of infectivity. PrPSc accumulates in the brain during scrapie infection and attains concentrations locally 10–100 times greater than that of PrPC (17, 46). PrP 27-30 forms into rod-like particles in vitro, whereas PrPC does not. These rods resemble the structures purified from amyloids and, like them, bind Congo red dye, which displays green birefringence in polarized light. Immunogold studies for electron microscopy verify that the rods are composed of PrP (16). PrP-specific antibodies indicate that the amyloid plaques which form in scrapie-infected brains (16), as well as those found in cases of CJD, GSS, and kuru (60, 64, 66), are composed of protease-resistant PrP.

Whether mutated PrP in inherited forms of human prion disease is protease-resistant or must act on PrPC to transform it into protease-resistant PrPCJD is not known. Transgenic mice which express PrP containing the P102L mutation mimicking the PRNP mutation of ataxic human GSS spontaneously develop a neurodegenerative disorder with neuropathological characteristics of scrapie (45). Subsequently we found that about 30% of these mice form PrP-immunopositive amyloid plaques spontaneously (Fig. 1).

A single host animal species can synthesize multiple types of scrapie prions each capable of faithfully transmitting a different clinical–pathological syndrome (18, 40). We have chosen to avoid using the term "strain" to describe distinct prion inocula which differ in their individual passage history among animals and which produce distinct clinical–neuropathological syndromes because of the implication from strains of bacteria, viruses, and viroids that diversity resides in the sequence of their nucleic acid genomes. We currently prefer to use the term "isolate" for prions to avoid prejudging the molecular mechanisms which determine reproducible clinical and neuropathological features.

The existence of distinct prion isolates was first discovered during laboratory transmission studies of sheep scrapie to sheep and goats (56). Subsequently, more than 15 scrapie prion isolates were identified in rodents (8) which could be distinguished by scrapie incubation time, the distribution and intensity of spongiform degeneration, and whether or not cerebral amyloid plaques formed. These characteristics were preserved during multiple sequential passages of a given prion isolate within a single mouse or hamster strain but varied markedly or even failed to appear when transferred to a different animal species (host species barrier). Classic genetic studies in different mouse strains which exhibited short (100–180 days) or long (greater than 280 days) incubation times with different prion isolates indicated that scrapie incubation time was determined largely by a single autosomal gene (20). Following the discovery of PrP, molecular genetic studies showed that the scrapie incubation time gene is tightly linked to, or identical to, the Prn-p gene (11).

There is considerable evidence from studies of PrP turnover in scrapie-infected cell lines that PrPSc is derived from preexisting PrPC (5, 12, 69). It is believed that exogenous PrPSc in the form of inoculated prions, as well as nascent PrPSc produced endogenously, form a transient complex (a heterodimer) with PrPC molecules and that this interaction leads to the conversion of PrPC into nascent PrPSc. The fact that each animal species can synthesize multiple prion isolates with strain-like characteristics implies that specific information is transferred from PrPSc to PrPC during its conversion to nascent PrPSc. It is further assumed that the information must be "coded" by specific stable structural configurations of PrPSc.

There are three possible mechanisms to explain how PrPSc can attain multiple stable structural states. First, it has been proposed that an accessory cellular RNA called a "co-prion" can modify the properties of PrPSc; however, there is no physical or chemical evidence to support the existence of a co-prion. The alternative hypothesis is that transferable information is stored exclusively in prion protein molecules. Thus, a second mechanism may be that each cell in an animal synthesizes a single isoform of PrPC with the same amino acid sequence and same secondary and tertiary structures. In this case, one would have to postulate that specific structural information coding for individual prion isolate behavior in PrPSc of the infecting prion is transferred stably to PrPC. A third mechanism has been suggested by studies with transgenic mice which express both Syrian hamster (SHa) PrPC and mouse (Mo) PrPC (62, 65), as well as by our recent studies of the patterns of PrPSc deposition in the brain as a function of prion isolate (18, 19, 40). The former study indicates that prion isolates selectively bind to homologous PrPC molecules and convert them into specific kinds of PrPSc molecules. Syrian-hamster-adapted Sc237 scrapie prions which contain SHaPrPSc selectively interacted with SHaPrPC in these Tg(SHaPrP) mice to form nascent Sc237 prions, and mouse-adapted RML scrapie prions which contain MoPrPSc selectively interacted with MoPrPC to form nascent RML prions. In the second set of studies, the histoblot technique revealed that the neuroanatomical location of PrPSc deposition was different and characteristic for each prion isolate (Fig. 2). When combined, these findings argue for the possibility that each subset of neurons synthesizes a different PrPC isoform with the same amino acid sequence but with perhaps different secondary or tertiary structures. Prion-stimulated conversion of PrPC into nascent PrPSc would occur only in those neurons which synthesize an isoform of PrPC which is compatible with PrPSc of the infecting prion.

There is ample evidence for variation in the glycosylation patterns of PrPSc isolated from SHa brain. Multiple charge isomers due to variations in sialylation have been demonstrated by 2-D electrophoresis (4). The diversity of PrPSc glycoforms is sufficient to account for a large number of prion isolates. Furthermore, cells potentially possess distinct repertoires of glycosyltransferases which could synthesize PrPC molecules with diverse carbohydrate structures. Presumably, the interaction between PrPSc and PrPC would be fostered by receptors on the surface of cells which recognize PrPSc carbohydrates homologous to those which are attached to PrPC synthesized within the cells. This hypothesis not only is supported by a number of experimental observations, but also might explain how each prion isolate exhibits a specific scrapie incubation time, neuropathologic lesion profile, and pattern of PrPSc accumulation. The postulate set forth here does not preclude PrPSc interacting with receptor proteins on the surface of cells which might restrict its entry into the cell or its binding to PrPC.

Although the diversity of structures found among Asn-linked sugar chains is attractive as a possible source of the biological diversity manifest by prion isolates, the large array of complex-type oligosaccharides found linked to PrP 27-30 isolated from Syrian hamster brains inoculated with the Sc237 isolate is also at odds with this hypothesis. Bi-, tri-, and tetraternary structures were found, some of which contained branched fucoses (24). Some of the sugar chains were also sialylated. The extreme diversity of Asn-linked oligosaccharides found in one isolate makes it difficult to envision entirely distinct populations of Asn-linked sugar chains for each of five additional isolates or "strains." However, these PrP27-30 preparations were made from whole hamster brains and not from specific brain regions, and therefore the hypothesis has not yet been rigorously tested. It is also possible that many different PrPSc molecules are formed in the brain, but that only a subset will determine clinical behavior.

Although the differences in location of PrPSc accumulation raise the possibility of the neuronal origin of prion isolates, the fact that there are also regions common to each of the prion isolates in which PrPSc accumulated (19) requires comment. One possibility is that these regions contain subpopulations of neurons, each with differential susceptibility to prion infection. Another possibility is that they represent neurons which nonspecifically accumulate PrPSc of any form perhaps through a retrograde transport mechanism. Similarly, it is possible that PrPSc is transported by anterograde axonal transport from multiple neurons, each with their own selective vulnerability to prion infection.

Our recent finding that the pattern of PrPSc deposition is different for each prion isolate argues that it is another fundamental characteristic (18, 19, 40). The neuropathological feature common to all forms of human and animal prion disorders is spongiform (vacuolar) degeneration of synaptic regions of the grey matter. This is accompanied by varying degrees of nerve cell loss and reactive astrocytic gliosis. Immunohistochemical and neurochemical localization of PrPSc during scrapie have revealed a precise topographical correlation between PrPSc, spongiform degeneration, and reactive astrocytic gliosis (17). This correlation has been found with both Sc237 and 139H isolates of the scrapie agent (40) and more recently with the Me7H isolate (19). There is also a temporal correlation between the accumulation of PrPSc and the development of neuropathology (46). However, the most convincing argument that abnormal PrP causes pathology comes from the molecular genetic studies which link dominantly inherited human prion diseases with mutations in the PRNP gene described above.

Thus a wealth of the experimental data argues that PrPSc is a necessary component of the scrapie infectious agent and that its accumulation in the brain causes the clinically relevant neuropathology. Furthermore, the rate and pattern of PrPSc accumulation—and, therefore, the rate of formation of neuropathology—determine scrapie incubation time (15, 40). For these reasons, it appears that the rate and pattern of PrPSc deposition may be the most relevant characteristics of each prion isolate.

A major unresolved issue concerns the etiology and pathogenesis of sporadic forms of CJD which account for the great majority, about 85%, of human prion diseases. Because neither an infectious nor familial etiology has been found for these cases and because molecular genetic studies of familial prion disorders indicate that multiple mutations of the PRNP gene are pathogenic, the possibility of an age-related acquired mutation of the PRNP gene must be considered. In this regard, sporadic CJD is an age-related disorder with a peak onset at about 60 years of age. It is likely that the PRNP genes in neurons or other cell types occasionally experience mutations. Some of these may be repaired and some may not be pathogenic; however, some may be pathogenic mutations and not repaired. If a mutation leads to the formation of PrPCJD even in a single neuron, the experience from infectious scrapie argues that a chain reaction could result which leads to the spread of disease to other susceptible neurons. During the process of spreading, the neuron in which the mutated PrP was synthesized might degenerate, leaving no objective evidence of a somatic mutation. Because prions do not stimulate neoplastic growth of cells to produce a tumor in which the putative mutation leads to the production of PrPCJD, this hypothesis will be difficult to verify. The incidence of CJD in populations throughout the world is about 1 case per 106. This may represent the combined probabilities that a mutation occurs in the PRNP gene, the probability that the mutation leads to the synthesis of PrPCJD, and the probability that the resultant PrPCJD targets other neurons for the synthesis of more PrPCJD at a rate fast enough to cause clinical disease in the patient's lifetime.

The somatic mutation hypothesis predicts that any neuron could be the source of PrPCJD. In the context of the hypothesis that neurons determine prion isolates through an inherent control of PrP structure, one would predict that sporadic CJD consists of multiple distinct clinical–neuropathological syndromes. Consistent with this postulate, multiple clinical–neuropathological subtypes of CJD have been described, including primarily cortical, corticostriatal with visual impairment, corticostriatal without visual impairment, corticostriatocerebellar, corticospinal, and corticonigral types. That there are "strains" of CJD prions is also suggested by transmission experiments. Transmission of CJD and kuru from many patients into a variety of subhuman primates, cats, and rodents revealed that the susceptible host range, incubation times, duration of illness, and type of clinical disease varied significantly among human prion isolates (33). Once a CJD isolate was transmitted to one rodent species or strain, other rodent strains which had previously been resistant to primary infection with that particular prion isolate could be infected with it. In terms of the prion protein hypothesis, the explanation for this form of species barrier breach is that human CJD prions, composed of PrPCJD, when passaged in a mouse, become composed of rodent PrPSc.

While the somatic mutation hypothesis may account for some cases of CJD, equally plausible is the notion that sporadic CJD results from the spontaneous conversion of PrPC into PrPSc in one or a limited subpopulation of cells. Most of the features of sporadic CJD described above are equally plausible with this model. A third possible mechanism of sporadic CJD is suggested by its association with homozygousity at PRNP codon 129 (see below).

Molecular studies of iatrogenic CJD, particularly those related to human growth hormone treatment, suggest that normal polymorphisms of the PRNP gene may cause a predisposition to prion infection.

It is believed that kuru, which is virtually indistinguishable from CJD, was transmitted among the Fore people of New Guinea by ritualistic cannibalism. It is conjectured that this infectious form of human prion disease was initiated in the Fore as the result of spontaneous development of a case of sporadic CJD in one of its members. The only proven cases of infectious CJD in the "civilized" world have resulted from medical procedures (7). Cases of CJD thought to be transmitted by medical procedures include the following: (a) Depth-recording electroencephalographic electrodes sterilized with 70% ethanol and formaldehyde vapor (two cases); electrodes were used previously in a patient with known CJD. While this sterilization procedure was effective for viruses, it was not effective for prions which requires 1 N NaOH denaturation of PrPCJD. (b) Corneal transplant; brain extracts from the donor transmitted CJD to chimpanzee. (c) Cadaveric dura homografts have been implicated in seven cases of iatrogenic CJD. Five of the seven dura specimens came from a single manufacturer where they had been treated with 10% hydrogen peroxide and ionizing radiation. They are now treated for 1 hr with 1 N NaOH. (d) A pericardial homograft replacement for a tympanic membrane has been implicated in one case. (e) The most devastating in terms of numbers has been the transmission of CJD via human growth hormone (hGH) therapy (47). In the past, hGH was prepared from pools containing as many as 20,000 pituitaries obtained at necropsy. The first case of growth-hormone-associated CJD was described almost 10 years ago (49). Today, more than 43 cases are known. Most of these have occurred before the age of 40. The risk among those who have been treated with hGH is estimated to be about 1 per 200 currently, whereas the risk of CJD in the general population under 40 years of age is about 1 per 20 {ewc MVIMG, MVIMAGE,!times.bmp} 106 (25). (f) Human gonadotropin therapy has been associated with CJD in two Australian women (13, 23).

Seventeen patients who developed CJD following hGH inoculations did not have any known pathogenic mutations of the PRNP gene (14). However, Collinge et al. (14) found that four of seven patients in the United Kingdom with iatrogenic CJD were homozygous for Val at PRNP codon 129. Brown et al. (7) found that four of nine patients from the United States and France were homozygous for Val, while four others were homozygous for Met. One of the nine was heterozygous for this polymorphism. The normal distribution of the codon 129 polymorphism in the general Caucasian population of the United States, France, and United Kingdom was found to be virtually identical: Val/Val = 11–12%, Met/Met = 37–38%, and Met/Val = 51% (7, 14). These findings indicate that homozygosity at codon 129 is not necessary for the development of infectious CJD, but they do suggest that it is associated with increased susceptibility. Interestingly, two of three kuru patients were reported to be homozygous for Val at codon 129 (36).

Because of this association, Palmer et al. (55) tested the possibility that there is a significant homozygosity at codon 129 among cases of sporadic CJD. Twenty-two sporadic CJD cases and 23 suspected sporadic CJD cases in the United Kingdom were examined. Ninety-five percent of the sporadic CJD cases were homozygous (16 Met/Met, 5 Val/Val, and 1 Met/Val). Whether the one heterozygous patient was truly a sporadic case came into question when it was learned that his father died of dementia, raising the possibility that it was familial CJD. Eighty-two percent of the suspected sporadic CJD cases were also homozygous (11 Met/Met, 6 Val/Val, and 4 Met/Val).

Other examples of the apparent importance of homozygosity at codon 129 was its association with more severe forms of fatal familial insomnia and 178-CJD mentioned above. It has also been reported that individuals with CJD associated with a heterozygous 144-base-pair insert between codons 51 and 91 in the PRNP gene and homozygous for Met at codon 129 died at a significantly earlier age than those heterozygous at codon 129 (mean age 42.7 years versus 56.9 years) (57). The PRNP allele containing the insert always coded for methionine at codon 129.

One implication of these findings concerns the etiology of sporadic CJD. They present an alternative to the acquired mutation hypothesis by raising the possibility that some cases of "sporadic" CJD are due to a codon-129-determined susceptibility to environmental prions. That it is possible to acquire a prion disease from the environment is attested to by the experience with kuru and its association with ritualistic cannibalism and by the recent bovine spongiform encephalopathy (BSE) epidemic (70). With regard to the latter, cattle developed BSE, a spongiform encephalopathy with accumulation of proteaseresistant PrPSc primarily in the brainstem (61), after they were given sheep-scrapie-contaminated feed.

The influence of codon 129 homozygosity on the severity and rapidity of prion diseases may shed some light on the mechanism of PrPC transformation to PrPSc in which a PrPC–PrPSc complex is believed to be an essential step. It raises the possibility that the interaction of PrPSc with PrPC requires the formation of a heterodimer composed of a dimer of PrPSc and a dimer of PrPC. Putative PrPC dimer formation would be favored when the amino acids at codon 129 are the same.

The Met and Val polymorphism at codon 129 appears to be the most common and generally nonpathogenic polymorphism among human PrP alleles in the general population. An uncommon polymorphism has been identified in some kindreds with the D178N mutation which consists of 24-base-pair deletion of the PRNP gene in the octapeptide repeat portion of the prion protein (6).

PrP gene knockout mice appear to live a normal lifespan with no morphological or behavioral abnormalities (10). These animals do not develop scrapie following inoculation with a spectrum of prion isolates (9, 63). The negative results with the PrP null mice are consistent with the preeminence of the prion protein in prion disorders. They also argue that scrapie and CJD are not due to the loss of the normal isoform, PrPC. PrP null mice are now being used to express a broad variety of PrP genes in the absence of normal mouse PrP gene expression.

The apparent normal mouse phenotype in PrP null mice suggests possible methods for treatment and/or prevention of prion diseases. In the case of the economically devastating animal diseases related to the prion protein, scrapie in sheep and BSE in beef and dairy cattle, it may be possible to breed animals without PrP genes to produce prion-disease-resistant animal strains which are safe for humans. It is interesting to conjecture whether elimination of prion diseases in domestic animals would have an impact on the incidence of "sporadic" CJD. The results with null mice also raise the possibility of gene therapy for human prion diseases. Human prion diseases are universally fatal with a relatively rapid, unrelenting course to death once signs and symptoms present. One can hope that someday with antisense PrP gene therapy, it will be possible to diminish synthesis of PrPC and therefore prevent or slow the formation of PrPCJD as well as prevent the synthesis of mutated forms of PrP. The experience with PrP null mice suggests that attenuating PrPC synthesis may not be detrimental; however, we do not yet know the normal function of PrPC or whether critical cell functions become dependent on it when it is expressed normally throughout life. These as well as multiple other questions remain to be answered about the etiology, pathogenesis, and treatment of this set of neurodegenerative diseases which are uniquely manifest as genetic and infectious disorders.

The authors wish to thank Dr. Shu-Lian Yang and Audrey Lee for preparing histoblots, Mrs. Juliana Cayetano-Canlas for neurohistology, and Mrs. Jo Nelson for help in manuscript preparation. This work was supported by research grants AG02132, NS14069, AG08967, and NS22786 from the National Institutes of Health as well as by gifts from the Sherman Fairchild Foundation and National Medical Enterprises.

published 2000