Copper and the prion protein
Copper: a role for prion protein?
Clusterin: an extracellular version of heat-shock protein
Are strain types a new idea?
Florid plaques unique to nvCJD?
Commentary -- webmaster 11.23.97In the modern era, people don't sequence proteins any more than absolutely necessary, they sequence in a short ways if at all, deduce the DNA sequence, pull the gene, sequence that, and infer the final amino acid sequence. Unless there is absolutely first-rate site recognition software, important post-translational modifications of the amino acids are then missed.
Ironically, the prion protein has covalent modifications that occur early on, which luckily were caught during Edman degradation. But then the ball was fumbled:
Hood and Prusiner noted early on [EJB 176 21 1988] that R25 and R37 but not R48 'appear to be modified...due to a post-translation modification' but dismissed this saying R25 in rogue conformer 'appears to be similarly modified' and that 'other investigators have reported [unmodified] arginine in these cycles for hamster' and 'studies of mouse demonstrate [normal] arginine residues at R25 and R37 as well as an unidentified molecule eluting with threonine.' In a later review article [Science 252 1516 1991], Prusiner noted the unknown arginine modifications again but that 'these are inconsistently reported.'
The supporting references actually say:
J Hope [EMBO 5:2591-2597 1986]: "the PTH-amino acids tyrosine and arginine inverted and arginine eluted before tyrosine' later observing, in the abstract even, J Hope [EJB 172(2):271-277 1988], "we found a novel, as yet unidentified, amino-acid derivative of the arginine residue at position 3 in mouse PrP, which may predispose PrP to form SAF."
Hope et al. then note, "in all sequencing traces ... we see an unidentified derviative or arginine at positions R25 and R37 [that] elutes with the retention time of PTH-thr. We do not think this compound is PTH-thr as this [arginine] derivative is unstable, raidly deprading to dehydro-products which elute just before PTH-pro.
Gibbs' lab [PNAS 87 6373-6377 1990] didn't exactly report smooth sailing either, writing "position 1 was identified as serine (machine call) and we identified position 3 [R25] as arginine."
Not cited was a possibly redundant account by J Hope, N Hunter [Ciba Found Symp 1988;135:146-163] "we also found a novel, as yet unidentified, amino acid derivative of the arginine residue at position 3 in both hamster and mouse PrP 33-35, which may predispose PrP to form SAF.
In summary, 3 labs found that prion protein has two covalently modified arginines in the pre-repeat region just upstream of the putative copper binding site but,12 years later, they remain uncharacterized. The precise nature of the modification is a good clue to prion structure and function: arginine modifications are informative precisely because they are unusual -- side chain modifications happen for excellent reasons. Partial modification may imply a cycle of modification and demodification, rather than inessentiality.
This raises the issue of prion nmr structures are being determined on the wrong molecule: no copper, no arginine modification. This weakens the case for a random coil for the amino terminus. Model peptides used to study copper also need the correct structure of the pre-repeat region. Recombinant E.coli does not modify arginine modifications [Oesch prion mass spec data; also seen in endothelial nitric-oxide synthase expressed in E.coli]. Modified arginine also raises the issue of specific auxillary genes that recognize this site in the prion repeat region at some point during its life cycle.
What else might be found if the whole molecule were studied and what is the situation cross-species? The RP and RYP themes occur again in several places, most notably R 48 and R148, the latter never investigated. Arginine overall occurs at:
MANLGCWMLVLFVATWSDLGLCKKRPKPGGWNTGGSRYPGQGSPGGNRYP PQGGGGWGQPHGGGWGQPHGGGWGQPHGGGWGQPHGGGWGQ GGGTHSQWNKPSKPKTNMKHMAGAAAAGAVVGGLGGYMLGSAMSRPIIHFGSDYEDRYYRENMHRYPNQ VYYRPMDEYSNQNNFVHDCVNITIKQHTVTTTTKGENFTETDVKMMERVVEQMCITQYERESQAYYQRGSSMVLFS SPPVILLISFLIFLIVGAccording to Medline:
The single most common modification of protein arginine in eukaryotes is ADP-ribosylation, invariably done for regulatory purposes, sometimes to cell surface proteins [reverse reaction catalyzed by ADP-ribosylarginine hydrolase]. The modified arginine (ADP-ribosylated) is generally the 'switched-off' condition. Mammalian mono(ADP-ribosyl)transferase is a secretory or GPI-anchored membrane protein, placing it appropriately to interact with prion protein, partially modifying it.. Bacteria toxins specifically ADP-ribosylate the rho family of GTP-binding proteins at a specific asparagine residue in their putative effector domain, interfering with their interaction with a putative effector molecule downstream in signal transduction.
Another common arginine modification in a neural setting is to citrulline [loosely, carbonyl-arginine]. Human myelin basic protein has a charge isomer with six citrullinyl residues which replace arginyl residues at selected sites. Aggregation and mediated adhesion are affected. Citrulline-containing myelin could be studied with antisera specific to the citrullinyl group. Myelin basic protein regulates stack charge, adhesion, and aggregation. Peptidyl-arginine deiminase in the presence of calcium ions makes the citrulline. A keratin intermediate filament matrix protein shows this modification in a high proline-glycine structural settingy.
Ornithine [loosley, short lysine] is a less common related modification.
Phospho-arginine: also occurs in myelin basic protein where modifications are known to decrease the ability of MBP to aggregate acidic lipid vesicles possibly regulating the ability of myelin basic protein to mediate adhesion between the intracellular surfaces of myelin. Phosphoarginine is an interesting candidate as often multiple arginines are affected. Note phosphoarginine is acid-labile and might elute earlier, as needed, unlike methylated arginine. If these arginines bind to negatively charged phospholipid per Warwicker's suggestion, phosphorylation would eliminate this interaction.
Various N-methylation scenarios also occur in lipophosphoglycan-associated protein, basic fibroblast growth factor, and hnRNA [requires RGG site]. Heterogenous nuclear ribonucleoproteins (hnRNPs) bind pre-mRNAs and facilitate their processing into mRNAs. Many of the hnRNPs undergo extensive posttranslational modifications including methylation on arginine residues. hnRNPs contain about 65% of the total NG,NG-dimethylarginine found in the cell nucleus, by protein-arginine N-methyltransferase, evolutionarily conserved from lower eukaryotes to mammals, suggesting that argilnine methylation has a significant role in the function of RNA-binding proteins. Basic fibroblast growth factor (bFGF) is a heparin-binding angiogenic polypeptide mitogen. bFGF purified from guinea pig brain tissue has methylated arginines.
The modifying enzymes (say, peptidylarginine deiminase or ADP-ribosylarginine hydrolase) may specifically recognize sites on the prion molecule. They have genetics of their own, polymorphisms that might cause disease states or affect cross-species transmission.
There doesn't seem to exist online software to identify arginine modification sites in prion protein, evidently an inadequate training set. PSORT doesn't have this feature. SwissProt won't accept arginine AND modif*; Workbench found 22 entries but only PIR:JW0039 and Q08642 were correctly annotated; it is hard to pull the set of sequences with modified arginines and develop recognition software.
Pharmacol Ther 1995;67(3):323-350; Eur J Biochem 1992 Sep 15;208(3):747-752 Phosphohistidine goes undetected in conventional studies of protein phosphorylation, although it may account for 6% of total protein phosphorylation in eukaryotes. Procedures for studying protein N- kinases are described. There is a protein arginine kinase in mouse. N omega-Phosphoarginine hydrolase from rat liver formed arginine and inorganic phosphate, whereas it did not release inorganic phosphate from 19 other phosphorylated compounds, ie it is distinct from both phosphoamidase and arginine kinase. The effect of cyclic AMP-dependent phosphorylation and ADP-ribosylation on the activities of the rat liver bifunctional enzyme, phosphofructo-kinase/fructose-bisphosphatase: Val-Leu-Gln-Arg-Arg-Arg-Gly-Ser-Ser-Ile-Pro-Gln was ADP-ribosylated on all three arginine residues. Sequencing of the ADP-ribosylated native enzyme also demonstrated that the preferred sites were at Arg-29 and Arg-30, which are just N-terminal to Ser-32, whose phosphorylation is catalysed by cAMP-dependent protein kinase. ADP-ribosylation was independent of the phosphorylation state of the enzyme. Stimulation of Ca-calmodulin-activated histone 3 arginine kinase A calmodulin-activated histone 3 kinase was partially purified from nuclear extracts; histone 3 was phosphorylated on arginine. The histone 3 arginine kinase is a component of a complex containing a Ca(2+)-dependent calmodulin-binding protein and is involved with cell cycle exit in eukaryotes. In vitro, the kinase activity forms an acid-labile and alkaline-stable linkage. The putative phosphorylation sites are arginines 2, 128, 129, and 131. N(omega)-phosphoarginine phosphatase is useful for selectively detecting N(omega)-phosphoarginine residue in peptides containing various kinds of phosphorylated amino acids. Protein phosphorylation is one of the major signal transduction mechanisms for controlling and regulating intracellular processes. Model compounds for the phosphorylated and basic amino acids show the Arg guanidinium group can form a bidentate hydrogen bonded structure. Arginine methylation is a prevalent modification found in many RNA binding proteins, yet little is known about its functional consequences. The recent discovery of phosphorylated serine residues suggests a hypothesis in which a molecular switch governed by methylation and phosphorylation regulates biochemical properties N omega-phosphoarginine and phosphocreatine are substrates for phosphoamidase. The main characteristic of this protein kinase is that it is arginine-specific. Isolation of phosphoarginine required the use of proteolytic enzymes at alkaline pH since the phosphate bond is highly acid-labile. This protein kinase is able to autophosphorylate and to phosphorylate a single chromosomal protein tightly bound to DNA. It uses ATP and dATP as phosphate donors and is cAMP-independent. Arginine kinase catalyzes the production of L-phosphoarginine, which is the principal reserve of high energy phosphate compounds in insect muscle. Rapid separation of phosphoamino acids by isocratic hplc of orthophthalaldehyde derivatives: Anal Biochem 1985 Sep;149(2):344-348 High-performance liquid chromatography of acid-stable and acid-labile phosphoamino acids: J Chromatogr 1980 Dec 19;202(2):263-269 A simple preparation of N-phosphorylated lysine and arginine: Prep Biochem 1980;10(2):205-213Phosphorylation under basic conditions of the copper chelate of L-arginine yields N-omega-phosphoarginine as the sole product which is identical to that produced enzymatically.
Biochem J 1997 Nov 15;328(1):45-50 Michel D, Chatelain G, North S, Brun GClusterin/apoJ is an intriguing gene frequently isolated by differential screening in laboratories from different areas of molecular biology, since it is overexpressed in numerous cases of degenerative diseases such as Alzheimer's disease and scrapie. While the dramatic increase of clusterin expression in injured tissues is well established, the molecular basis of the gene induction remains unclear. In this study, we have focused our attention on the only DNA region strictly conserved between clusterin gene proximal promoters from different vertebrate classes. We show that this 14-bp DNA element is specifically recognized by the HSF1 transcription factor and can mediate heat-shock-induced transcription in transient expression assays. Conversely, the avian clusterin proximal promoter, point-mutated at the level of this element, no longer transmits heat-shock activation. These findings provide a possible explanation for the high sensitivity of clusterin expression to environmental changes and allow the classification of clusterin as an extracellular version of heat-shock protein.
Neurosci Lett 1997 Sep 26;234(1):11-14 Chishti MA, Strome R, Carlson GA, Westaway DThe cellular isoform of the prion protein (PrP) serves as a precursor to abnormal PrP isoforms which accumulate in diseases such as scrapie in sheep, and Creutzfeldt-Jakob disease in humans. Since prions can replicate in photoreceptors we surmised that PrP(C) must be expressed in these cells. Accordingly, monoclonal antisera directed against two epitopes of hamster PrP(C) produced retinal immunostaining in hamsters, and in mice bearing a hamster PrP transgene. Immunostaining was most prominent in the inner and outer segments of rod photoreceptors, coinciding with the earliest site of pathologic changes in scrapie-infected hamsters. These data define PrP(C) expression in an experimentally-accessible population of neurons and suggest that the retina may comprise a useful system for studying the biology of wild-type and mutant prion proteins.
Vet Rec 1997 Oct 4;141(14):352-357 Austin AR, Pawson L, Meek S, Webster SHeart rates of healthy cows and cows suspected of having bovine spongiform encephalopathy were measured by auscultation and by a portable cardiac monitor. Bradycardia was demonstrated in suspecte cases which were confirmed histopathologically. Disturbances in cardiac rhythm were also evident in some cases. Healthy cows deprived of food exhibited bradycardia. The administration of pharmacological doses of atropine indicated that bradycardia in BSE was mediated by increased vagal influence, suggesting that the cardioinhibitory reflexes in the caudal brainstem were functionally altered by the disease.
Proc. Natl. Acad. Sci. USA, Vol. 94, pp. 12503-12508, November 1997 Daniel C. Masison, Marie-Lise Maddelein, and Reed B. WicknerThe genetic properties of the non-Mendelian element, [URE3], suggest that it is a prion (infectious protein) form of Ure2p, a mediator of nitrogen regulation in Saccharomyces cerevisiae. Into a ure2 strain (necessarily lacking [URE3]), we introduced a plasmid overproducing Ure2p. This induced the frequent "spontaneous generation" of [URE3], with properties identical to the original [URE3]. Altering the translational frame only in the prion-inducing domain of URE2 shows that it is Ure2 protein (and not URE2 RNA) that induces appearance of [URE3]. The proteinase K-resistance of Ure2p is unique to [URE3] strains and is not seen in nitrogen regulation of normal strains.
The prion-inducing domain of Ure2p (residues 1-65) can propagate [URE3] in the absence of the C-terminal part of the molecule. In contrast, the C-terminal part of Ure2p cannot be converted to the prion (inactive) form without the prion-inducing domain covalently attached. These experiments support the prion model for [URE3] and extend our understanding of its propagation. That overexpression of the chaperone Hsp-104 cures [PSI] also suggests it is a prion.
J Cell Biol 1995 Jun;129(5):1301-1310"In Tetrahymena, at least 17 distinct microtubule structures are assembled from a single primary sequence type of alpha- and beta-tubulin heterodimer, precluding distinctions among microtubular systems based on tubulin primary sequence isotypes. Tetrahymena tubulins also are modified by several types of posttranslational reactions including acetylation of alpha-tubulin at lysine 40, a modification found in most eukaryotes."
Science Nov 21 MEETING BRIEFS: Marcia Barinaga 27th annual meeting of the Society for Neuroscience 25 to 30 October 1997New Orleans was host to 24,000 neuroscientists at the . With more than 14,000 presentations, the offerings were as rich as a New Orleans gumbo, spiced with new tidbits about topics ranging from prion proteins to Gulf War syndrome.
With the award of the Nobel Prize last month to University of California, San Francisco, neuroscientist Stanley Prusiner, prions have been in the news a lot lately. But in spite of the work by Prusiner and others implicating these rogue proteins in a variety of fatal brain diseases, including the United Kingdom's bovine spongiform encephalopathy (BSE or "mad cow disease"), the story has some gaping holes. One of the largest: No one knows the function of the normal prion protein, known as PrPc. Data presented at last month's neuroscience meeting by neurobiologist David Brown of the University of Cambridge in the United Kingdom may help remedy that.
His new findings suggest, Brown says, that PrPc protects neurons by binding copper ions. When they run free, these reactive ions are highly toxic to cells. If the normal prion protein does help sequester the ions, the abnormal prion folding and clumping thought to cause BSE and other prion diseases could cause nerve cell death by interfering with this protective function.
Researchers who study copper binding are all abuzz about Brown's results. "All my friends in neuroscience e-mailed me [after Brown's talk] to ask what is going on with this," says Jonathan Gitlin, an expert on copper-binding proteins at Washington University in St. Louis. Gitlin himself says that PrPc could well play a role in protecting nerve cells from copper ions.
But Brown's theory that a loss of the copper binding contributes to prion disease is more controversial. Many prion researchers believe that the destructive effects of the malformed prion protein, rather than loss of normal PrPc function, cause the diseases. As prion researcher Adriano Aguzzi of the University of Zurich in Switzerland points out, mice that make no PrPc because the gene has been knocked out do not get them.
Because those mice don't show obvious ill effects from the loss of PrPc, Brown took a different route to identifying the role of the normal protein. Previous studies by other researchers had shown that a fragment of PrPc binds copper in the test tube. Working in collaboration with Hans Kretzschmar of the University of G‚ttingen in Germany, Brown went on to see what effect this copper binding might have on brain cells.
In his first experiments, he cultured brain neurons from normal mice and >from mice lacking a functional PrPc gene. He found that cells without PrPc are much more susceptible than the cells from the normal mice to poisoning by copper sulfate. Brown then went on to show that a peptide containing the copper-binding site of PrPc could protect the mutant cells from copper's toxicity.
Further evidence that PrPc may protect cells from copper came in experiments indicating that the protein binds the metal in living brains. PrPc is normally found in the membranes of brain cells, and Brown reported that membranes prepared from the brains of normal mice "have a lot of copper"--almost 20 times more than those >from the knockout mice. Because PrPc is concentrated in the membranes of synapses, the specialized gaps where neurons are chemically coupled to one another, Brown postulated that it may serve to sop up copper ions released into synapses when neurons fire. In support of that idea, he showed that copper diminishes electrical activity in cerebellar neurons from the knockout mice that lack PrPc, but not in cerebellar neurons from normal mice.
Brown proposed that PrPc could have a more indirect role as well: It might pass the copper to other proteins, which could ferry it to enzymes inside the cell that need it for their activity. These include superoxide dismutase, which protects cells from damage by oxidizing chemical species that can form in cells from normal activities, such as energy metabolism. Indeed, he and Kretzschmar found less resistance to oxidative damage in neurons from PrPc knockout mice compared to those of normal mice.
Even though prion researchers generally blame the abnormal protein for the cell death seen in prion diseases, Aguzzi says Brown's findings raise the possibility that the loss of a normal function could also contribute in ways that have been missed or somehow compensated for in the knockout mice. If so, the normal prion protein may end up stealing some of the attention its malignant counterpart is now getting.
Thu, 20 Nov 1997 webmasterHmmm. I'm not sure what is so new in this Science meeting report that seems to have people on the edge of their seats, Volume 278, Number 5342 Issue of 21 November 1997, pp. 1404 - 1405. (No actual article has appeared.)
I posted exhaustive coverage of the copper angle 17 months ago, reviewing all the known copper neuroenzymes (these typically synthesize or degrade neurotransmitters with molecular oxygen), anti-oxidizing agents, and copper transmitters etc. To repeat, the 4 periodic histidines in the repeat region can and do bind a single copper (or zinc) ion in vitro at micro-molar concentrations. These 4 histidines are found back to birds even though the hexamer spacing is quite different. The fifth histidine could come in from the bottom. The question has always been, what would prevent copper or zinc binding from happening in vivo?
Prion copper review Copper enzyme structues More copper coverage Search for copper articlesThis mode of binding is not how copper is bound in any other known copper enzyme, not a problem as the repeat region seems an independent reinvention of something. Also not a serious problem if copper has not been found in purified protein, it would be lost during denaturing purification and use of EDTA in nmr studies. A copper ligand would prevent the unnatural flopping around of the N terminal as I have pointed out numerous times. Copper is either transport or active site.
However, there is no structural mechanism to soak up additional copper, this is no ferritin. There are quite a few metal transporters already, though this could be high affinity. I could not find any GPI transporters or copper enzymes. Neuro copper is more likely to be in short supply than a toxin -- there is already strong homeostatis of copper throughout the body. And the prion protein is expressed in many non-neural cell types (no synapses) where copper is not known to be such a big deal. Copper can knock out cysteine thiols and also generate indiscriminant oxidants.
"Brown then went on to show that a peptide containing the copper-binding site of PrPc could protect the mutant cells from copper's toxicity."
I have no idea what the reporter means here -- if the bare peptide protected, or already loaded peptide protected. The latter suggests an anti-oxidant function (hydroxyl radical, superoxide anion, peroxide). Superoxide dismutase is a copper-zinc enzyme (or Mn), neither family exhibits the slightest homology to prion protein. Neurotransmitters are so well studied they would have shown up years ago on Blast or Scop searches. There is an obscure neuro-sterol hydroxylase needed in sleep regulation that is worth checking out, might have fallen throught the cracks, would fit prion in buffy coat better than neurotransmitter.
"Brown postulated that it may serve to sop up copper ions released into synapses when neurons fire."
I am not aware of significant levels of copper being released from synaptic packets, seems like it would be easier to prevent waste at the source, had 600 million years to figure it out.
"Brown reported that membranes prepared from the brains of normal mice "have a lot of copper"--almost 20 times more than those from the knockout mice."
Interesting, a dramatic effect. Wonder if this refers to some specific membrane fraction, rather than total brain membrane including mitochondria and ER. Doesn't seem that there are enough prion molecules around to give much of an effect with 1-1 stoichiometric binding. Have to wonder, as always, how far downstream is the effect from the ultimate cause.
"Even though prion researchers generally blame the abnormal protein for the cell death seen in prion diseases, Aguzzi says Brown's findings raise the possibility that the loss of a normal function could also contribute in ways that have been missed or somehow compensated for in the knockout mice."
Aguzzi-Weissmann rightly pointed out that knockouts have from fertilization on to develop compensatory mechanisms, knockouts might be more lethal if suddenly turned on in an adult. There has always been the question of whether CJD by definition screens for accumulators of bad conformer or if conventional loss-of-function mutations are some other unrecognized disorder, possibly with quite dissimilar phenotype, possibly along the lines of the new Brazilian psychiatric disorder, N171S. The earliest CJD mutant is at codon 102, well after any copper site, which would be easy to knock out.
Lancet 15 Nov 97 James W Ironside, Jeanne E BellSir--Shutaro Takashima and co-workers (Sept 20, p 865)1 describe a case of iatrogenic Creutzfeldt-Jakob disease (CJD) in a dura-mater-graft recipient, in whom the neuropathological features were characterised by the presence of prion protein (PrP) plaques surrounded by areas of spongiform change, the so-called florid plaque.2 On the basis of this and two other similar cases, the investigators conclude that florid plaques are not specific to new-variant CJD (nvCJD), thereby raising questions about the neuropathological diagnostic criteria for this disorder.
Neuropathology is essential for the diagnosis of nvCJD, because the clinical features of this disorder are not specific and overlap with both sporadic CJD and other diseases during the course of the illness.3 In our original report,4 we did not claim that florid plaques were specific for nvCJD, since they were first described in a murine scrapie model,2 and similar plaques are a prominent feature of the pathology of chronic wasting disease in mule deer and elk.5 However, large fibrillary PrP amyloid plaques surrounded by a halo of spongiform change are characteristic for nvCJD and are easily recognised in routinely stained sections.
The illustration by Takashima and co-workers shows PrP plaques stained by immunocytochemistry in an area of confluent spongiform change, which is not characteristic for nvCJD florid plaques in which the halo of peripheral spongiform change is typically discontinuous, probably representing dilated neuronal processes within otherwise intact neuropil.2,4,5 Irrespective of the presence of florid plaques in their case, the other characteristic neuropathological feat-ures of nvCJD--spongiform change most pronounced in the basal ganglia, abundant PrP deposition in the occipital cortex and cerebellar molecular layer with perineuronal and perivascular deposits, and marked thalamic gliosis--were absent.
These features allow the clear distinction of nvCJD from other varieties of human prion disease. Obviously, not all these features can be detected in small cerebral-cortical samples and diagnostic difficulties may be encountered in the interpretation of cortical biopsies from patients with suspected nvCJD. For this reason, we would not recommend brain biopsy as a reliable diagnostic investigation for nvCJD. Analysis of PrP glycoform provides a potential molecular marker for nvCJD, and frozen brain tissue should be retained from all suspected cases for this purpose. Although there is much current interest in diagnostic clinical studies3 and tonsillar biopsy in the investigation of patients with suspected nvCJD, necropsy neuropathology is essential for diagnosis.
1 Takashima S, Tateishi J, Taguchi Y, Inoue H. Creutzfeldt-Jakob disease with florid plaques after cadaveric dural graft in a Japanese woman. Lancet 1997; 350: 865?66. 2 Fraser H. The pathogenesis and pathology of scrapie. In: Tyrell DAJ, ed. Aspects of slow and persistent virus infections. The Hague: Martinus Nijhoff, 1979: 30?58. 3 Zeidler M, Stewart GE, Barraclough CR, et al. New variant Creutzfeldt-Jakob disease: neurological features and diagnostic tests. Lancet 1997; 350: 903?07. 4 Will RG, Ironside JW, Zeidler M, et al. A new variant Creutzfeldt-Jakob disease in the UK. Lancet 1996; 347: 921?25. 5 Guiroy DC, Williams ES, Yanagihara R, Gajdusek DC. Topographic distribution of scrapie amyloid-immunoreactive plaques in chronic wasting disease in captive mule deer. Acta Neuropathologica 1991; 81: 475?78.