Alteration of theSerotonergic Nervous Systemin Fatal Familial InsomniaJulia Wanschitz,* Stefan Kloppel,* Christa Jarius,*Peter Birner,† Helga Flicker,* Johannes A. Hainfellner,*Pierluigi Gambetti,‡ Marin Guentchev,*and Herbert Budka*

Fatal familial insomnia (FFI) is a unique hereditary priondisease with characteristic disturbances of sleep. We stud-ied the serotonergic system in 8 FFI-affected subjects byimmunohistochemistry for the serotonin-synthesizing en-zyme, tryptophan hydroxylase (TH). Quantification ofneurons in median raphe nuclei showed no total neuro-nal loss in FFI but a substantial increase of TH1 neurons(;62%) in FFI subjects compared with controls. Ourdata indicate an alteration of the serotonergic system thatmight represent the functional substrate of some typicalsymptoms of FFI.

Wanschitz J, Kloppel S, Jarius C, Birner P,Flicker H, Hainfellner JA, Gambetti P,

Guentchev M, Budka H. Alteration of theserotonergic nervous system in fatal familial

insomnia. Ann Neurol 2000;48:788–791

Fatal familial insomnia (FFI) is a rare form of humantransmissible spongiform encephalopathies (TSEs), orprion diseases. FFI is inherited by an autosomal dom-inant trait and is genetically linked to a point mutationat codon 178 of the prion protein (PRNP) gene, cou-pled with the methionine polymorphism at codon 129of the mutant allele.1,2 Neuropathology differs fromother human TSEs, with predominance of lesions inthe thalamus, whereas involvement of the cortex ismild or absent.3,4

Patients with FFI display unique clinical features, in-cluding complex alterations of the sleep-wake cycle, at-tention deficit, sympathetic hyperactivity, attenuationof autonomic circadian and endocrine oscillations, andmotor signs.5 On polysomnography, non–rapid eyemovement (NREM) sleep is abolished, sleep spindlesand K-complexes disappear early, and brief, often clus-

tered, REM sleep episodes are associated with oneiricbehavior because of lack of physiological muscleatonia.6 Impairment of sleep and autonomic and endo-crine functions in FFI has been related to selectivedamage of anterioventral and mediodorsal thalamic nu-clei, the cingulate gyrus, and orbitofrontal gyrus, lead-ing to interruption of thalamocortical limbic circuitsinvolved in the control of the sleep-wake cycle.5,7

Ascending serotonergic projections arising in medianraphe nuclei densely innervate the limbic thalamus andrelated cortical regions. The brain serotonergic systemis considered to play an integral role in the regulationof sleep homeostasis,8 autonomic functions, and con-trol of locomotor activities.9 To investigate whether al-terations of the serotonergic system are present in FFI,we immunohistochemically visualized and evaluatedquantitatively serotonin-synthesizing neurons in themedian raphe.

Materials and MethodsWe used formol-fixed, paraffin-embedded brain tissue ob-tained at autopsy from 8 FFI affected subjects: 4 members ofthe Austrian FFI family10 and 4 cases from 3 different, pre-viously published kindreds.2,11 All cases were methionine ho-mozygous at codon 129. Eight sex- and age-matched indi-viduals without neurological disease served as controls(details summarized in the Table). Additionally, 4 sporadicand 2 familial subjects with Creutzfeldt-Jakob disease (1 re-ported previously12) were investigated as disease controls.The following brain regions were examined: upper pons, in-cluding the dorsal raphe nucleus (DRN) and superior centralnucleus (SCN), and medulla oblongata with raphe obscurusnucleus (RON). In 2 FFI cases, only the medulla oblongataand in another only the pons could be examined.

Immunohistochemistry was performed on 4-mm-thick sec-tions using a monoclonal antibody against tryptophan hy-droxylase (PH8; Pharmingen, San Diego, CA; 1:500) to vi-sualize serotonin (5-HT)-synthesizing neurons.13 Sections fortryptophan hydroxylase labeling were boiled 10 minutes incitrate-buffered saline (pH 6.0). As secondary system, weused a ChemMate detection kit (Dako, Glostrup, Denmark)as described previously.14 Immunostained sections werecounterstained with cresyl violet.

Total numbers of neurons and numbers of tryptophan hy-droxylase–immunoreactive (TH1) neurons were quantifiedon two consecutive sections across the upper pons and me-dulla with a 40-mm interval by counting cell bodies in sixcontinuous fields per section with a 403 objective, includingfour fields next to the median line and two more lateralfields in SCN, and six fields next to the median line inRON. These fields covered most or all of the area of thenucleus. Only nucleated cells or cells with their completesoma visible were counted. Relative numbers of TH1/TH2

neurons were evaluated in PH8-stained sections by countingpositive and negative neurons in the same fields. Mean values of12 fields, respectively, were entered into statistical evaluation.Since sections of the pons from some brains contained only cau-dal parts of the DRN, quantitation was not done there.

From the *Institute of Neurology, University of Vienna, and Aus-trian Reference Center of Human Prion Diseases, and †Departmentof Clinical Pathology, University of Vienna, Vienna, Austria; and‡Division of Neuropathology, Institute of Pathology, Case WesternReserve University, Cleveland, OH.

Received Dec 28, 1999, and in revised form Jun 12, 2000. Ac-cepted for publication Jun 15, 2000.

Address correspondence to Prof Budka, Institute of Neurology, Uni-versity of Vienna, AKH 04J, Wahringer Gurtel 18-20. POB 48.A-1097 Wien, Austria.

BRIEF COMMUNICATIONS

788 Copyright © 2000 by the American Neurological Association

The significance of the difference of mean regional cellnumbers and the ratio of TH1 to TH2 neurons were testedusing a Kruskal-Wallis one-way test followed by the exactversion of a two-tailed Mann-Whitney U test. For all tests,p # 0.05 was considered as the level of significance.

ResultsControlsHematoxylin and eosin (H&E)-stained sectionsshowed no histopathological changes. TH immunohis-tochemistry visualized cell bodies and axons of seroto-nergic neurons in raphe nuclei. Morphology of neu-rons was similar in all cases, appearing as either smallor large cells with multipolar or fusiform-shapedperikarya (Fig 1). In SCN, 38.9% 6 4.8 SEM wereTH1 and 61.1% 6 4.8 SD were TH2. In RON,34.6% 6 3.1 SEM of the neurons were TH1, whereas65.4% 6 3.1 SD were TH2 (Fig 2).

Creutzfeldt-Jakob DiseaseH&E sections yielded mild to moderate gliosis andspongiform change in the tegmentum. Quantificationof neurons showed no loss of total cell numbers and nochanged ratio of TH1 to TH2 neurons in SCN andRON. In SCN, 42.4% 6 4.4 SEM (p 5 0.818 vscontrols) of neurons were TH1 and 57.6% 6 4.4SEM were TH2. In RON, 40.5% 6 3.4 SEM of theneurons were TH1 (p 5 0.366 vs controls) and59.5% 6 3.4 SEM were TH2 (Fig 2).

Fatal Familial InsomniaH&E sections showed moderate gliosis in median rapheand other brainstem nuclei in all cases. There was nomacroscopical or microscopical atrophy or shrinkage inthe investigated nuclei. Quantification demonstrated noloss of total neuronal numbers in RON and SCN (datanot shown). In SCN, 62.2% 6 5.6 SEM of neuronswere TH1 and 37.8% 6 5.6 SEM were TH2 (p 50.026 vs controls; p 5 0.041 vs Creutzfeldt-Jakob dis-ease [CJD]). In RON, 56.9% 6 4.0 SEM of the neu-rons were TH1 and 43.1% 6 4.0 SEM were TH2

(p 5 0.002 vs controls; p 5 0.008 vs CJD; see Fig 2).In sum, we observed an increase of 59.6% and 64.4% ofTH1 neurons in SCN and RON, respectively.

DiscussionThe main clinical features of FFI—namely, loss of cy-clic sleep patterns, flattening of circadian and neuroen-docrine rhythmicity, and unbalanced sympathetic hy-peractivity—indicate a profound disturbance of controlof homeostasis. The pathophysiological role of the thal-amus has been emphasized based on the invariable andsevere degeneration of anteroventral and mediodorsalthalamic nuclei during the disease, while other areasclassically known to control autonomic functions, suchas the hypothalamus and several regions of the brain-stem, appear morphologically spared by the diseaseprocess.7

Table. Patients’ Clinical Data

Kindred ID No. Sex Age (yr) Cause of Death Mutation Reference

FFIFFI A IV-5 M 26 Pneumonia D178N M129M 10FFI A IV-8 M 34 Pneumonia D178N M129M 10FFI A III-5 F 58 Pneumonia D178N M129M 10FFI A IV-13 F 21 Pneumonia D178N M129M 10FFI-2 IV-26 M 52.5 NA D178N M129M 11FFI-2 IV-27 M 52.5 NA D178N M129M 11FFI-II III-1 M 55 NA D178N M129M 2FFI-10 PNA M 53 NA D178N M129M 11

Controls C1 M 35 CardiomyopathyC2 M 36 Accidental shot injuryC3 F 55 Idiopathic pulmonary fibrosisC4 F 27 Pneumonia after cesarean sectionC5 M 52 Coronary bypass operationC6 M 55 Acid ingestionC7 M 46 LymphogranulomaC8 M 62 Acute lymphoblastic leukemia

Creutzfeldt-Jakob CJD1 M 74 Pneumonia NA NAdisease CJD2 M 62 NA NA NA

CJD3 M 60 Pneumonia NA NACJD4 F 63 Status epilepticus NA NACJD5 M 31 NA E200K M129MCJD6 F 66 NA E200K V129V

FFI A 5 affected members of the Austrian FFI family; NA 5 not available.

Brief Communication: Wanschitz et al: Serotonin, Sleep, and FFI 789

The serotonergic system has been implicated in theregulation of the sleep-wake cycle and circadian rhyth-micity.15 It has been demonstrated that serotonin sup-presses REM sleep and ponto-geniculo-occipital wavesby inhibition of cholinergic laterodorsal tegmental andpedunculo-pontine tegmental nuclei.16 Atonia of anti-gravity muscles is a fundamental feature of REM sleep.Neurophysiological studies in pontine-lesioned catsshowed increased activity of 5-HT neurons duringREM sleep associated with high muscle tone of anti-gravity muscles and overt behavior,17 resembling the

REM sleep abnormalities in FFI. Moreover, experi-mental data document a sympathoexcitatory role of5-HT, causing exaggerated cardiovascular responses.18

In this study, we demonstrate increased numbers ofTH1 neurons and a changed ratio of TH1 to TH2

neurons in the pons (SCN) and medulla (RON) ofFFI patients compared with controls. These changeswere consistently observed in FFI cases from 4 differ-ent families. There is no decrease in total cell numbers;thus, relative increase of TH1 neurons in FFI cannotbe explained by loss of TH2 neurons. This observation

Fig 1. Immunostaining for tryptophan hydroxylase in raphe obscurus nucleus of the medulla (A–C) and superior central nucleus ofthe pons (D–F) in control (A, D), fatal familial insomnia (B, E), and Creutzfeldt-Jakob disease (C, F) brains. Note the increase ofnumbers of TH1 neurons in fatal familial, insomnia (B, E). Asterisks indicate midline. (Original magnification: A–C, 350;D–F, 3200; anti-PH8, cresyl violet counterstain.)

Fig 2. Box-plot analysis of the percentage of TH1 neurons in raphe obscurus nucleus (RON) and superior central nucleus (SCN) offamilial fatal insomnia (FFI) (A), corresponding controls (B), and Creutzfeldt-Jakob disease (CJD) (C). Differences between FFIand controls and FFI and CJD are significant in RON (p 5 0.002 and p 5 0.008, respectively; Mann-Whitney U test) andSCN (p 5 0.026 and p 5 0.041, respectively, Mann-Whitney U test). There is no statistical difference between CJD and corre-sponding controls.

790 Annals of Neurology Vol 48 No 5 November 2000

more likely reflects a recruitment of TH1 neuronsfrom a pool of “silent” raphe neurons with the poten-tial to synthesize serotonin or the ability to activate se-rotonin production. Because we have not observed sim-ilar changes in the serotonergic system of sporadic andfamilial CJD or cases with a selective ischemic thalamiclesion (unpublished data of our group), the changedratio of TH1 to TH2 neurons is unlikely to occursecondary to the thalamic damage in FFI. The imbal-ance between TH1/TH2 neurons indicates a distur-bance in the homeostasis of serotonin metabolism.These observations provide evidence for an enhancedserotonergic neurotransmission as possible functionalsubstrate for some symptoms in FFI.

Several serotonin-related disorders show no detect-able morphological abnormalities, such as mood disor-ders.19 Investigation of the TH1/TH2 ratio in thesedisorders may provide deeper insight into the patho-physiology of these disorders and contribute to the for-mation of a rationale for adequate treatment.

This work has been supported by the European Union Biomed-2Concerted Actions “Human transmissible spongiform encephalopa-thies (prion diseases): neuropathology and phenotypic variation”and “European Centralized Facility for Human TSEs” (projectleader, H. Budka).

Special thanks are due to G. Paul Amminger for helpful suggestions.

References1. Medori R, Tritschler HJ, LeBlanc A, et al. Fatal familial insom-

nia, a prion disease with a mutation at codon 178 of the prionprotein gene. N Engl J Med 1992;326:444–449

2. Goldfarb LG, Petersen RB, Tabaton M, et al. Fatal familialinsomnia and familial Creutzfeldt-Jakob disease: disease pheno-type determined by a DNA polymorphism. Science 1992;258:806–808

3. Manetto V, Medori R, Cortelli P, et al. Fatal familial insomnia:clinical and pathologic study of five new cases. Neurology1992;42:312–319

4. Guentchev M, Wanschitz J, Voigtlander T, et al. Selective neu-ronal vulnerability in human prion diseases. Fatal familial in-somnia differs from other types of prion diseases. Am J Pathol1999;155:1453–1457

5. Lugaresi E, Medori R, Montagna P, et al. Fatal familial insom-nia and dysautonomia with selective degeneration of thalamicnuclei. N Engl J Med 1986;315:997–1003.

6. Lugaresi E, Tobler I, Gambetti P, Montagna P. The pathophys-iology of fatal familial insomnia. Brain Pathol 1998;8:521–526

7. Benarroch EE, Stotz-Potter EH. Dysautonomia in fatal familialinsomnia as an indicator of the potential role of the thalamus inautonomic control. Brain Pathol 1998;8:527–530

8. Jouvet M. Sleep and serotonin: an unfinished story. Neuropsy-chopharmacology 1999;21(Suppl):24S–27S

9. Gerson SC, Baldessarini RJ. Motor effects of serotonin in thecentral nervous system. Life Sci 1980;27:1435–1451

10. Almer G, Hainfellner JA, Brucke T, et al. Fatal familialinsomnia: a new Austrian family. Brain 1999;122:5–16

11. Parchi P, Castellani R, Cortelli P, et al. Regional distribution ofprotease-resistant prion protein in fatal familial insomnia. AnnNeurol 1995;38:21–29

12. Hainfellner JA, Parchi P, Kitamoto T, et al. A novel phenotypein familial Creutzfeldt-Jakob disease: prion protein gene E200Kmutation coupled with valine at codon 129 and type 2protease-resistant prion protein. Ann Neurol 1999;45:812–816

13. Haan EA, Jennings IG, Cuello AC, et al. Identification of se-rotonergic neurons in human brain by a monoclonal antibodybinding to all three aromatic amino acid hydroxylases. BrainRes 1987;426:19–27

14. Guentchev M, Groschup MH, Kordek R, et al. Severe, earlyand selective loss of a subpopulation of GABAergic inhibitoryneurons in experimental transmissible spongiform encephalopa-thies. Brain Pathol 1998;8:615–623

15. Jouvet M. Biogenic amines and the states of sleep. Science1969;163:32–41

16. Horner RL, Sanford LD, Annis D, et al. Serotonin at the lat-erodorsal tegmental nucleus suppresses rapid-eye-movementsleep in freely behaving rats. J Neurosci 1997;17:7541–7552

17. Trulson ME, Jacobs BL, Morrison AR. Raphe unit activity dur-ing REM sleep in normal cats and in pontine lesioned cats dis-playing REM sleep without atonia. Brain Res 1981;226:75–91

18. Bell AA, Butz BL, Alper RH. Cardiovascular responses pro-duced by microinjection of serotonin-receptor agonists into theparaventricular nucleus in conscious rats. J Cardiovasc Pharma-col 1999;33:175–180

19. Maes M, Meltzer HY. The serotonin hypothesis of depression.New York, NY: Raven Press, 1995

Brief Communication: Wanschitz et al: Serotonin, Sleep, and FFI 791

Inhibition ofCyclooxygenase-2 ProtectsMotor Neurons in anOrganotypic Model ofAmyotrophic LateralSclerosisDaniel B. Drachman, MD,and Jeffrey D. Rothstein, MD, PhD

The pathogenesis of motor neuron loss in amyotrophiclateral sclerosis (ALS) is thought to involve bothglutamate-mediated excitotoxicity and oxidative damagedue to the accumulation of free radicals and other toxicmolecules. Cyclooxygenase-2 (COX-2) may play a keyrole in these processes by producing prostaglandins,which trigger astrocytic glutamate release, and by induc-ing free radical formation. We tested the effects ofCOX-2 inhibition in an organotypic spinal cord culturemodel of ALS. The COX-2 inhibitor (SC236) providedsignificant protection against loss of spinal motor neu-rons in this system, suggesting that it may be useful inthe treatment of ALS.

Drachman DB, Rothstein JD. Inhibition ofcyclooxygenase-2 protects motor neurons in an

organotypic model of amyotrophic lateralsclerosis. Ann Neurol 2000;48:792–795

There is increasing evidence that glutamate-mediatedexcitotoxicity plays an important role in the pathogen-esis of amyotrophic lateral sclerosis (ALS).1,2 Gluta-mate released by presynaptic neurons normally stimu-lates motor neurons, and its synaptic actions arerapidly terminated by specific glutamate transporterson astrocytic and neuronal membranes surrounding thesynapse. Accumulation of glutamate in the synapticspace can exert neurotoxic effects mediated by the en-try of Ca21 in motor neurons and the production ofvarious toxic molecules.2 Impairment of glutamate up-take by the astrocytic transporters is thought to play animportant role in the pathogenesis of ALS.1,2 It is nowclear that astrocytes not only take up glutamate but

that they also synthesize and release substantialamounts of glutamate. Astrocytic glutamate release maybe stimulated by prostaglandins via a calcium-dependent pathway,3,4 and this triggers additional glu-tamate release from the astrocytes by a positive feed-back mechanism. Prostaglandins are produced withinthe central nervous system by the enzymatic action ofcyclooxygenase-2 (COX-2), which catalyzes the synthe-sis of prostaglandins from arachidonic acid. Inhibitorsof COX can powerfully inhibit astrocytic glutamate re-lease.3 Thus, COX-2 inhibitors might have a therapeu-tic effect in ALS by inhibiting astrocytic release ofglutamate. Furthermore, the fact that COX-2 playsa pivotal role in inflammatory processes in the centralnervous system5 has led us to anticipate that its inhibi-tion could have additional therapeutic benefit in ALS.

To evaluate the therapeutic effect of COX-2 inhibi-tion, we used an in vitro model of glutamate excito-toxicity.6 Mammalian spinal cord slices are maintainedin an organotypic culture system in which the motorneurons normally remain intact for more than 3months. When astroglial transport of glutamate is in-hibited by threo-hydroxyaspartate (THA), however, thespinal cord sections exhibit persistent elevation of glu-tamate levels and undergo gradual loss of motor neu-rons. This model system has been used in the preclin-ical evaluation of agents of potential interest for thetreatment of ALS.7 In the present study, we used aselective COX-2 inhibitor (SC236; Monsanto/Searle,St Louis, MO), which is homologous to celecoxib, adrug currently in use for the treatment of arthritis. Ourfindings show that the addition of SC236 to THA-treated spinal cord cultures resulted in highly signifi-cant protection against loss of motor neurons.

Material and MethodsOrganotypic spinal cord cultures were prepared from lumbarspinal cords of 8-day-old rat pups sectioned transversely into350-mm slices and cultured on Millicell CM (MilliporeCorp, Bedford, MA) semipermeable culture inserts at 37°Cwith 5% CO2 and 95% humidity. Under these conditions,more than 95% of cultures retain cellular organization, withsurvival of motor neurons in excess of 3 months. Culturemedia consisted of HEPES-buffered minimum essential me-dium (50%) with 25% heat-inactivated horse serum and25% Hanks’ balanced salt solution supplemented withD-glucose (25.6 mg/ml) and glutamine (0.2 M). Seven daysafter culture preparation, THA was added to experimentalcultures at 150 mM, which produces excitotoxic damage tomotor neurons within 3 to 4 weeks. SC236 was added asindicated to achieve final concentrations from 0 to 100 mM.Parallel spinal cord cultures with no drugs added, THAalone, SC236 alone, and SC236 plus THA were run simul-taneously. Experiments at each concentration of SC236 wererepeated three to five times. A total of 25 to 47 cultures wereused for each treatment. The medium, with THA andSC236 at indicated concentrations, was changed twice aweek. The cultures were maintained for 4 weeks and pre-

From the Department of Neurology, The Johns Hopkins School ofMedicine, Baltimore, MD.

Received May 1, 2000, and in revised form Jun 14. Accepted forpublication Jun 15, 2000.

Address correspondence to Dr Drachman, Department of Neurol-ogy, Johns Hopkins School of Medicine, Meyer Building 5-119,600 North Wolfe Street, Baltimore, MD 21287–7519.

792 Copyright © 2000 by the American Neurological Association

pared as described below for quantification of motor neu-rons.

Motor neurons were identified by two immunocytochem-ical markers: SMI-32 stains nonphosphorylated neurofila-ments, which are abundant in motor neuron cell bodies, andIslet-1 is a motor neuron–specific marker. Cultures werefixed with 4% paraformaldehyde in 0.1 M phosphate bufferfor 30 minutes and then permeabilized with cold methanolor 0.1% Triton X-100 for SMI-32 and Islet-1 staining, re-spectively. Incubation with SMI-32 antibodies (1:8,000) orIslet-1 antibodies (1:100) was performed overnight at 4°C.After incubation with biotinylated horse anti-mouse antibod-ies and avidin-biotin complex reagents (Vector Labs, Burlin-game, CA), color was developed with diaminobenzidine.

To quantify surviving motor neurons, whole-mount cul-tures immunostained by SMI-32 were used. The microsco-pist was unaware of the treatments used. Motor neuronswere identified by three criteria: immunostaining with SMI-32, size greater than 25 mm, and localization to the ventralgray region of spinal cord (Fig 1). The motor neuron countusing SMI-32 was verified by Islet-1 staining in some cases.

ResultsTreatment with THA produces gradual motor neurontoxicity attributable to excessive extracellular glutamate.This mimics, in part, the loss of glutamate transportthat occurs in sporadic ALS patients. We used a con-centration of THA (150 mM) that produces motorneuron damage in 3 to 5 weeks.

Addition of THA alone resulted in a highly signifi-cant loss of motor neurons compared with control un-treated cultures (see Figs 1 and 2). SC236 added to theTHA-treated cultures potently and significantly pro-tected motor neurons against excitotoxic damage at allconcentrations tested (see Figs 1 and 2). The apparentlack of dose response is unexplained but could reflect aplateau effect. Nevertheless, this COX-2 inhibitor pre-vented the excitotoxic loss of motor neurons. We alsonoted a moderate decrease of motor neurons in culturestreated with SC236 alone.

DiscussionThese experiments were undertaken to test the hypoth-esis that inhibition of COX-2 activity could preventexcitotoxic motor neuron loss in an ALS-like organo-typic spinal cord culture model. This hypothesis wasbased on the key role of COX-2 in the production ofprostaglandins and the demonstrated effects of prosta-glandins in stimulating astroglial glutamate release andproducing inflammatory changes as well as a variety oftoxic molecules. Our findings show that COX-2 inhi-bition had a highly significant protective effect on mo-tor neurons in this model system.

Different forms of COX exist in two isoforms:COX-1, which is constitutively expressed at extremelylow levels in the nervous system but is not inducible,and COX-2, which is present throughout the centralnervous system8,9 and is highly inducible. The differentforms of COX catalyze the rate-limiting steps in thesynthesis of prostaglandins.8,10 COX-2 is present inneurons and astrocytes11,12 as well as in macrophagesand microglia.13 COX-2 has been demonstrated in theanterior horns of the spinal cord14–17 and specificallyin motor neurons of rat spinal cord by immunocyto-chemistry.14,16 COX-2 is an inducible enzyme8,11 thatincreases in the brain after synaptic activity, seizures, orischemia9 and in the spinal cord after trauma to thecord16 or peripheral noxious stimuli such as inflamma-tory arthritis.15 Proinflammatory cytokines, includinginterleukin-1b, upregulate the expression of COX-2 inastrocytes as well as in other cells.8,11 Of particular in-terest, interleukin-1b is markedly elevated in spinalcords of a mutant superoxidase dismutase 1G93A mousemodel of ALS and plays a role in apoptotic neuronalcell death in that model.18

It is likely that the effectiveness of COX-2 inhibitionin preventing excitotoxic motor neuron death in ourmodel system is due to the interruption of multipleprocesses, which can interact in a vicious cycle to pro-duce progressive damage to motor neurons and astro-cytes. One such mechanism is the prostaglandin-induced release of glutamate from astrocytes andsecondarily from neurons, which induces excitotoxicdamage. Excitotoxicity of the released glutamate is fur-ther enhanced because of the decreased buffering ca-

Fig 1. Administration of cyclooxygenase-2 (COX-2) inhibitorSC236 significantly protected against glutamate-mediated tox-icity. Spinal cord organotypic cultures were incubated with theglutamate transporter inhibitor threo-hydroxyaspartate (THA)for 4 weeks in vitro along with the COX-2 inhibitor SC236as indicated. THA alone produced a large loss of motor neu-rons. SC236 protected against THA-mediated toxicity over awide concentration range. **p , 0.01 when compared withTHA-treated cultures. *p , 0.05 when compared with THA-treated cultures.

Brief Communication: Drachman and Rothstein: COX-2 Inhibition in a Model of ALS 793

pacity of astrocytes due to impairment of astrocyticglutamate transport, actual loss of astrocytes,1 and po-tentiation of excitotoxic effects of glutamate by overex-pression of COX-2 in neurons.10 In ALS, any of thestimuli or inflammatory molecules noted above couldinitiate upregulation of COX-2. The process may beself-perpetuating, with damage and dysfunction of neu-rons and astrocytes spreading contiguously in wideningareas as is commonly observed clinically in ALS.

COX-2 inhibitors also interrupt inflammatory pro-cesses that result in the production of free radicals andother toxic molecules. Inhibition of COX-2 has beenshown to protect against the effects of global ischemiaor reperfusion injury in the brain and trauma to thespinal cord.16,19 There is a recent report that a solubleaspirin analog may delay the onset of motor deficits inmutant superoxidase dismutase 1 mice.20

The present results suggest that COX-2 inhibitioncould have therapeutic effects in ALS by altering thecascade of pathogenic processes that otherwise cause re-lentless progression of motor neuron damage. Inhibi-tion of COX-2 could break the pathogenetic cycle by(1) preventing prostaglandin-mediated release of gluta-mate from astrocytes and (2) interrupting an importantpathway that otherwise leads to the production of freeradicals, toxic and proinflammatory molecules whichdamage neurons and glia and upregulate the activity ofCOX-2. The availability of COX-2 inhibitors thatreadily penetrate the central nervous system and arerelatively nontoxic facilitates therapeutic trials in ani-mal models of ALS.

The SC236 was supplied by Monsanto Searle, which also providedpartial support for these studies.

Fig 2. Cyclooxygenase-2 (COX-2) inhibitor SC236 protected against threo-hydroxyaspartate (THA)–induced chronic glutamate tox-icity. Organotypic spinal cord cultures were maintained for 4 weeks, and motor neurons were identified by staining neurofilamentswith SMI-32. (A) Control culture with no THA or SC236. Labeled motor neuron cell bodies in the ventral horn are within thedotted circle labeled VH. (B) SC236-treated cultures (without THA). (C) THA-treated culture showing loss of motor neurons bilat-erally in the ventral horns and loss of motor axons. (D) Culture treated with SC236 (10 mM) and THA showing protection ofmotor neurons. DH 5 dorsal horn; VH 5 ventral horn; arrow 5 motor axons.

794 Annals of Neurology Vol 48 No 5 November 2000

We thank Dr K. Andreasson for helpful discussions throughout thesestudies and Carol Coccia for carrying out highly skilled technical work.

References1. Rothstein JD, Martin LJ, Kuncl RW. Decreased glutamate

transport by the brain and spinal cord in amyotrophic lateralsclerosis. N Engl J Med 1992;326:1464–1468 (Comments)

2. Shaw PJ, Ince PG. Glutamate, excitotoxicity and amyotrophiclateral sclerosis. J Neurol 1997;244(Suppl 2):S3–14

3. Bezzi P, Carmignoto G, Pasti L, et al. Prostaglandins stimulatecalcium-dependent glutamate release in astrocytes. Nature1998;391:281–285

4. Sanzgiri RP, Araque A, Haydon PG. Prostaglandin E(2) stim-ulates glutamate receptor-dependent astrocyte neuromodulationin cultured hippocampal cells. J Neurobiol 1999;41:221–229

5. Boldyrev AA, Carpenter DO, Huentelman MJ, et al. Sources ofreactive oxygen species production in excitotoxin-stimulatedcerebellar granule cells. Biochem Biophys Res Commun 1999;256:320–324

6. Rothstein JD, Jin L, Dykes-Hoberg M, Kuncl RW. Chronicinhibition of glutamate uptake produces a model of slow neu-rotoxicity. Proc Natl Acad Sci USA 1993;90:6591–6595

7. Corse AM, Bilak MM, Bilak SR, et al. Preclinical testing ofneuroprotective neurotrophic factors in a model of chronic mo-tor neuron degeneration. Neurobiol Dis 1999;6:335–346

8. Williams CS, DuBois RN. Prostaglandin endoperoxide synthase:why two isoforms? Am J Physiol 1996;270:G393–400

9. Kaufmann W, Andreasson K, Isakson P, Worley P. Cyclooxy-genases and the central nervous system. Prostaglandins 1997;54:601–624

10. Kelley KA, Ho L, Winger D, et al. Potentiation of excitotox-icity in transgenic mice overexpressing neuronal cyclooxy-genase-2. Am J Pathol 1999;155:995–1004

11. O’Banion MK, Miller JC, Chang JW, et al. Interleukin-1 beta in-duces prostaglandin G/H synthase-2 (cyclooxygenase-2) in primarymurine astrocyte cultures. J Neurochem 1996;66:2532–2540

12. Hirst WD, Young KA, Newton R, et al. Expression of COX-2by normal and reactive astrocytes in the adult rat central ner-vous system. Mol Cell Neurosci 1999;13:57–68

13. Tomimoto H, Akiguchi I, Wakita H, et al. Cyclooxygenase-2 isinduced in microglia during chronic cerebral ischemia in hu-mans. Acta Neuropathol (Berl) 2000;99:26–30

14. Goppelt-Struebe M, Beiche F. Cyclooxygenase-2 in the spinalcord: localization and regulation after a peripheral inflammatorystimulus. Adv Exp Med Biol 1997;433:213–216

15. Ebersberger A, Grubb BD, Willingale HL, et al. The intraspi-nal release of prostaglandin E2 in a model of acute arthritis isaccompanied by an up-regulation of cyclooxygenase-2 in thespinal cord. Neuroscience 1999;93:775–781

16. Resnick DK, Graham SH, Dixon CE, Marion DW. Role ofcyclooxygenase 2 in acute spinal cord injury. J Neurotrauma1998;15:1005–1013

17. Willingale HL, Gardiner NJ, McLymont N, et al. Prostanoidssynthesized by cyclo-oxygenase isoforms in rat spinal cord andtheir contribution to the development of neuronal hyperexcit-ability. Br J Pharmacol 1997;122:1593–1604

18. Li M, Ona VO, Guegan C, et al. Functional role of caspase-1and caspase-3 in an ALS transgenic mouse model. Science2000;288:335–339

19. Nakayama M, Uchimura K, Zhu RL, et al. Cyclooxygenase-2inhibition prevents delayed death of CA1 hippocampal neuronsfollowing global ischemia. Proc Natl Acad Sci USA 1998;95:10954–10959

20. Barneoud P, Curet O. Beneficial effects of lysine acetylsalicy-late, a soluble salt of aspirin, on motor performance in a trans-genic model of amyotrophic lateral sclerosis. Exp Neurol 1999;155:243–251

Down’s Syndrome IsAssociated with Increased8,12-iso-iPF2a-VI Levels:Evidence for EnhancedLipid Peroxidation In VivoDomenico Pratico, MD,* Luigi Iuliano, MD,†Giulia Amerio, MD,‡ Lina X. Tang, BSc,*Joshua Rokach, MD,§ Giuseppe Sabatino, MD,‡and Francesco Violi MD†

Postmortem and in vitro studies have shown that oxida-tive stress plays a role in the pathogenesis of many of theclinical features of Down’s syndrome. The isoprostane8,12-iso-iPF2a-VI is a specific marker of lipid peroxida-tion. We found elevated levels of this isoprostane in urinesamples of subjects with Down’s syndrome comparedwith those of matched controls, which correlated with theduration of the disease. These results suggest that in-creased in vivo lipid peroxidation is a prominent compo-nent early in the course of Down’s syndrome.

Pratico D, Iuliano L, Amerio G, Tang LX,Rokach J, Sabatino G, Violi F. Down’s syndrome

is associated with increased 8,12-iso-iPF2a-VIlevels: evidence for enhanced lipid peroxidation in

vivo. Ann Neurol 2000;48:795–798

Caused by the triplication of chromosome 21, Down’ssyndrome (DS) is one of the most common humancytogenetic abnormalities.1 Those with DS suffer froma wide range of symptoms that are considered to occurthrough damage caused by overexpression of normalgene products encoded by extra gene copies present onthis chromosome.2 Human copper/zinc-superoxide dis-mutase (SOD1) is one of these genes, and increasedlevels of this enzyme as a result of gene dosage havebeen shown in DS subjects.3 The increased gene dos-age for SOD1 has been proposed to contribute to theprecocious dementia of Alzheimer’s type and mentalretardation that occurs as part of the syndrome.4 High

From *The Center for Experimental Therapeutics, University ofPennsylvania, Philadelphia, PA; †Institute of Clinica Medica I, Uni-versity “La Sapienza,” Rome, and ‡Children’s Hospital of the Uni-versity of Chieti, Chieti, Italy; and §Claude Pepper Institute andDepartment of Chemistry, Florida Institute of Technology, Mel-bourne, FL.

Received Apr 3, 2000, and in revised form Jun 28. Accepted forpublication Jun 28, 2000.

Address correspondence to Dr Pratico, Center for ExperimentalTherapeutics, University of Pennsylvania, BRB II/III, Room 812,421 Curie Boulevard, Philadelphia, PA 19104.

Copyright © 2000 by the American Neurological Association 795

SOD1 activity results in the accumulation of hydrogenperoxide, which leads to the formation of an excess ofhydroxyl radicals, the most reactive oxygen species(ROS), through a Fenton-type reaction and to subse-quent oxidative stress with loss of cellular functionthrough damage to macromolecules.4

Despite the fact that some studies have associatedDS with increased levels of markers of oxidative stress,there are shortcomings to these investigations. Theyhave used markers that are not specific and also haveused mostly tissue assays and in vitro cellular systemsrather than changes in plasma or urine as potential bi-omarkers of oxidant stress in living patients with DS.5,6

F2-isoprostanes are isomers of the prostaglandin F2a

produced by ROS attack on polyunsaturated fatty ac-ids. They are chemically stable, sensitive, and specificquantitative biomarkers of lipid peroxidation in vitroand in vivo.7

Here, we measured levels of 8,12-iso-iPF2a-VI, oneof the most abundant F2-isoprostanes in human be-ings,7,8 in the urine of young living subjects with DSand compared those levels with the levels in the urineof matched controls.

Subjects and MethodsThe participants were randomly recruited from the Chil-dren’s Hospital of the University of Chieti, Chieti, Italy. Pa-tients were all karyotyped at birth. Each participating familywith a DS child was asked to provide a urine sample. Con-trol samples were from non-DS siblings of the patients orfrom healthy nonrelated children attending the clinic. Thestudy was approved by the Ethics Committee of the Institu-tional Review Board of the hospital where it was conducted.Informed and signed consent was obtained from a parent orguardian of each participant. The urine samples were col-lected in plastic tubes early in the morning and brought tothe clinic within 4 hours of collection. An aliquot (1 ml) wasstored for creatinine determination. Immediately afterward,the samples were frozen in a coded fashion at 280°C andkept frozen until analysis.

8,12-iso-iPF2a-VI AssaySamples were spiked with internal standard [4H2]-8,12-iso-iPF2a-VI, extracted, purified, and finally assayed by gaschromatography and mass spectrometry as previously de-scribed.7,8 The intra-assay and interassay variability was64% and 65%, respectively. All assays were performed in acoded fashion.

Statistical AnalysisComparisons between groups were performed by nonpara-metric one-way ANOVA (Kruskall-Wallis test) with the useof Dunn’s posttest. Correlation was studied by linear regres-sion analysis. Statistical significance was set at p , 0.05.

ResultsThirty-three subjects with DS and 33 control subjectsparticipated in the study; their characteristics are

shown in Table 1. No significant difference in age orgender was observed between controls and DS patients.Study participants stopped taking any vitamins or anti-oxidants 4 weeks before the study. Three patients weretaking substitutive hormonal therapy (thyrotropin-stimulating hormone) for a mild form of hypothyroid-ism, which was well controlled at the time of the study(not shown).

A scatterplot diagram of urine 8,12-iso-iPF2a-VI lev-els for the 2 groups of patients is shown in Figure 1.Control subjects had a median value of 1.20, with arange of 0.25 to 2.5 ng/mg of creatinine. DS patientshad significantly higher levels than controls, with a me-dian of 1.97 and a range of 0.75 to 4.25 ng/mg ofcreatinine (p , 0.01). This significant difference per-sisted, excluding the patients taking thyrotropin-stimulating hormone and the ones with congenitalheart disease (data not shown). Interestingly, 15 (45%)of the DS subjects had levels of 8,12-iso-iPF2a-VIhigher than 2 ng/mg of creatinine (mean 6 2 SDhigher than that of control subjects). Creatinine valuesbetween the 2 groups did not differ (p 5 0.9). Therewas no difference between male and female DS sub-jects (p 5 0.8). Next, we determined whether the in-crease in 8,12-iso-iPF2a-VI observed in DS subjectswas correlated with the duration of the disease. A sig-nificant correlation was observed between those levelsand the age of the DS subjects (r2 5 0.35, p 5 0.002)(Fig 2); however, the same was not found in controls(data not shown).

DiscussionIn this article, we demonstrate that young DS patientshave significantly higher urinary levels of 8,12-iso-iPF2a-VI, a sensitive and specific marker of lipid per-oxidation, than age- and sex-matched controls. Thisfinding supports the hypothesis that increased lipidperoxidation and oxidative stress occur in vivo and areimportant components early in the life of DS subjects.Furthermore, the increased levels significantly corre-

Table. Clinical Characteristics of Down’s Syndrome andControl Subjects

Down’sSyndromeSubjects

ControlSubjects

Number 33 33Age (yr) 6 (range,

1–15)6 (range,1–15)

Sex (male/female) 21/12 20/13Associated disease

Congenital heart disease 4 0Hypothyroidism 3 0

796 Annals of Neurology Vol 48 No 5 November 2000

lated with the age of the patients, indicating that theyalso correlate with the duration of the disease.

Although DS was described over 100 years ago, andthe relation between trisomy 21 and Down’s pheno-type has been known for more than three decades, lit-tle is still known about the mechanisms responsible forthe clinical manifestation of the disease.9 Elevated lev-els of normal products encoded by genes residing inthis chromosome are thought to alter various biochem-ical pathways and their homeostasis, giving rise to thespecific pathologies affecting these subjects. TheSOD-1 gene is located on this chromosome, and it isknown that DS patients have elevated levels of this en-zyme and activity.2 This can produce an excess in ROSand a subsequent imbalance in the antioxidant enzy-matic pathways.3,4 Based on this observation, it is be-lieved that the DS fetus is situated in an environmentof increased oxidative stress early in the course of itslife. Indeed, DS fetal brains have increased lipid per-

oxidation, glycoxidation, and DNA oxidation prod-ucts.5,10 In addition, the augmented SOD-1 activity,by increasing the amount of ROS, mediates degenera-tion and apoptosis of DS cortical neurons in vitro.11

Thus, cellular damage may begin accumulating inutero and continue during the life span of the individ-ual with DS, contributing to the phenotypes associatedwith aging changes such as the Alzheimer’s type ofneuronal pathology.12,13 Another possible mechanismthat can underlie oxidative brain damage in DS is thepresence on chromosome 21 of the gene that codifiesfor the amyloid precursor protein.14 This gene dosageeffect would create an increased concentration of amy-loid b (Ab) in the brain of these patients. It is knownthat amyloid Ab can directly produce ROS by bindingdirectly to the receptor of advanced glycation endproducts15 and that accumulation of intracellular Ab,similar to the case in Alzheimer’s disease patients, in-duces cellular oxidative stress.16 Interestingly, we andothers have reported that F2-isoprostane formation isincreased in Alzheimer’s disease patients.17,18

Although it is clear that oxidative stress might play arole in DS, there have been few in vivo data to supportthis hypothesis. This is because most of the studieshave been performed in postmortem DS brains. Sec-ond, measuring oxidative stress in vivo has been a dif-ficult challenge, because some techniques give nonspe-cific results.7 This could explain some conflicting datain the literature and the fact that the role of oxidativestress in DS has remained controversial.19,20

In the present study, by using a reliable, specific, andsensitive quantitative index, we provided evidence thatlipid peroxidation is a prominent component early inthe course of the syndrome in living patients with DS.The temporal increase in lipid peroxidation observed inDS subjects may reflect accumulation over time ofthese products, which could derive from overproduc-tion and/or reduced clearance.

Quantification of this noninvasive specific marker ofin vivo lipid peroxidation may be especially helpful intesting the hypothesis that antioxidant therapy could beefficacious in slowing down the evolution of this disease.

This work was in part supported by a grant from the AmericanFederation for Aging Research (P99143).

We thank the patients and their families who made this researchpossible.

References1. Hook EB. In: de la Cruz FF, Gerald PS, eds. Trisomy 21

(Down’s syndrome), research perspective. Baltimore: UniversityPark Press, 1981:3–68

2. Coyle JT, Oster-Granite ML, Gearhart JD. The neurobiologicconsequences of Down syndrome. Brain Res Bull 1986;16:773–787

Fig 1. Urinary 8,12-iso-iPF2a-VI levels in Down’s syndrome(Œ) and control subjects (f ) (n 5 33 in each group).

Fig 2. Linear regression analysis of urinary 8,12-iso-iPF2a-VIlevels versus age in Down’s syndrome subjects (r2 5 0.35,p 5 0.002 ).

Brief Communication: Pratico et al: Lipid Peroxidation in DS 797

3. Groner Y, Elroy-Stein O, Avraham KB, et al. Down syndromeclinical symptoms are manifested in transfected and transgenicmice over-expressing the human Cu/Zn-superoxide dismutasegene. J Physiol (Lond) 1990;84:53–77

4. De Haan JB, Wolvetang EJ, Cristiano F, et al. Reactive oxygenspecies and their contribution to pathology in Down syndrome.Adv Pharmacol 1997;38:379–402

5. Odetti P, Angelini G, Dapino D, et al. Early glycoxidationdamage in brains from Down’s syndrome. Biochem BiophysRes Commun 1998;243:849–851

6. Peled-Kamar M, Lotem J, Okon E, et al. Thymic abnormalitiesand enhanced apoptosis of thymocytes and bone marrow cellsin transgenic mice over-expressing Cu/Zn-superoxide dismu-tase: implication for Down’s syndrome. EMBO J 1995;14:4985–4993

7. Pratico D. F2-isoprostanes: sensitive and specific non-invasiveindices of lipid peroxidation in vivo. Atherosclerosis 1999;147:1–10

8. Li H, Lawson JA, Reilly M, et al. Quantitative high perfor-mance liquid chromatography/tandem mass spectrometry anal-ysis of the four classes of F2-isoprostanes in human urine. ProcNatl Acad Sci USA 1999;96;13381–13386

9. Patterson DH. The causes of Down syndrome. Sci Am 1987;257:42–49

10. Brooksbank BWL, Martinez M. Lipid abnormalities in thebrain in adult Down’s syndrome and Alzheimer’s disease. MolChem Neuropathol 1989;11:157–185

11. Busciglio J, Yanker BA. Apoptosis and increased generation ofreactive oxygen species in Down’s syndrome neurons in vitro.Nature 1995;378:776–779

12. Iannello RC, Crack PJ, de Haan JB, Kola I. Oxidative stressand neural dysfunction in Down syndrome. J Neural Trans1999;57(Suppl):S257–267

13. Burger PC, Vogel FS. The development of the pathologicchanges of Alzheimer’s disease and senile dementia in patientswith Down’s syndrome. Am J Pathol 1973;73:457–468

14. Weideman A, Konig G, Bunke D, et al. Identification, biogen-esis, and localization of precursors of Alzheimer’s disease A4amyloid protein. Cell 1989;57:115–126

15. Yan SD, Chen X, Fu J, et al. RAGE and amyloid-beta peptideneurotoxicity in Alzheimer’s disease. Nature 1996;382:685–691

16. Yan SD, Fu J, Soto C, Chen X, et al. An intracellular proteinthat binds amyloid beta peptide and mediates neurotoxicity inAlzheimer’s disease. Nature 1997;389:689–695

17. Pratico D, Lee VM-Y, Trojanowski JQ, et al. Increased F2-isoprostanes in Alzheimer’s disease: evidence of enhanced lipidperoxidation in vivo. FASEB J 1998;12:1777–1783

18. Montine TJ, Markesbery WR, Morrow JD, Roberts LJ. Cere-brospinal fluid F2-isoprostane levels are increased in Alzheimer’sdisease. Ann Neurol 1998;44:410–413

19. Seidl R, Greber S, Schuller E, et al. Evidence against increasedoxidative DNA-damage in Down syndrome. Neurosci Lett1997;235:137–140

20. Jovanovic SV, Clements D, MacLeod K. Biomarkers of oxida-tive stress are significantly elevated in Down syndrome. FreeRadic Biol Med 1998;25:1044–1048

PABP2 Polyalanine TractExpansion CausesIntranuclear Inclusions inOculopharyngeal MuscularDystrophyVijayalakshmi Shanmugam, PhD,* Patrick Dion, PhD,*Daniel Rochefort, MSc,* Janet Laganiere, BSc,*Bernard Brais, MD, PhD,†and Guy A. Rouleau, MD, PhD*

Oculopharyngeal muscular dystrophy is caused by expan-sion of a (GCG)n trinucleotide repeat in the poly(A)binding protein 2 (PABP2) gene. The pathological hall-mark of oculopharyngeal muscular dystrophy is the ac-cumulation of intranuclear inclusions in muscle fibers.To test whether the polyalanine expansion of PABP2 di-rectly leads to the formation of the nuclear aggregates,both normal and expanded PABP2 cDNAs were ex-pressed in COS-7 cells. We find that expression of mu-tated PABP2 protein is sufficient for its accumulation asintranuclear inclusions.

Shanmugam V, Dion P, Rochefort D, Laganiere J,Brais B, Rouleau GA. PABP2 polyalanine tract

expansion causes intranuclear inclusions inoculopharyngeal muscular dystrophy.

Ann Neurol 2000;48:798–802

Oculopharyngeal muscular dystrophy (OPMD) is a he-reditary disease that affects all skeletal muscles, espe-cially those responsible for eyelid elevation and swal-lowing. OPMD may lead to severe dysphagia withmalnutrition and aspiration pneumonia, which can befatal.1,2 It has a worldwide distribution but is particu-larly frequent in French Canadians.1–3 The OPMDgene, which was mapped to chromosome 14q.11.2-q134 and positionally cloned,5 is the human homo-logue of the bovine poly(A) binding protein type II(PABP2) gene.6 The 306–amino acid protein has apredicted molecular mass of 32.8 kd.6 PABP2 is a nu-clear protein7 that is involved in the polyadenylation ofmRNA.8,9 The normal PABP2 gene has a (GCG)6

trinucleotide repeat coding for a polyalanine stretch at

From the *Centre for Research in Neuroscience, McGill University,and the Montreal General Hospital; and †Centre de Recherche duCHUM, Hopital Notre-Dame, Universite de Montreal, Montreal,Quebec, Canada.

Received Apr 4, 2000, and in revised form Jul 5. Accepted for pub-lication Jul 7, 2000.

Address correspondence to Dr Rouleau, L7-224, Montreal GeneralHospital, 1650 Cedar Avenue, Montreal, Quebec, H3G 1A4, Canada.

798 Copyright © 2000 by the American Neurological Association

the 59 end. The disease occurs when a small expansionof (GCG)6 to (GCG)8–13 occurs.5 Because there arethree GCA and one GCG codon adjacent to the GCGtract, the normal PABP2 protein has a 10-alanine stretchwhereas the mutant proteins have 12 to 17 alanines.

In 1980, Tome and Fardeau described clusters ofintermediate-sized tubular filaments in the nuclei ofOPMD skeletal muscle, intranuclear inclusions, a fea-ture that has since served as the pathological hallmarkof this condition.10 It has been proposed that these nu-clear filaments might be “nuclear toxic” and cause celldeath.2,5 In vitro studies have shown that polyalaninepolymers as small as 12 amino acids form b-pleatedsheet complexes,11 which are extremely insoluble.12

The presence of intranuclear inclusions in OPMD sug-gests that the expanded polyalanine tract of mutantPABP2 protein may play a direct role in disease patho-genesis by accumulating as insoluble aggregates. Inter-estingly, a similar pathogenetic model has been pro-posed for diseases caused by expanded polyglutamineencoding CAG tracts.

To determine whether expression of the mutatedPABP2 protein alone is sufficient for formation ofintranuclear inclusions, both normal and mutated(GCG-13 repeat) PAPB2 fused to enhanced green flu-orescent proteins was transfected into COS-7 cells. Weshow that mutated proteins, and not the normal pro-tein, oligomerize and form aggregates leading to in-tranuclear inclusions in a time-dependent manner. Fur-thermore, we show that the aggregates contain thecomplete PABP2 protein. This model for OPMD willhelp to study the mechanisms of toxicity of mutantPABP2 protein in vitro.

Materials and MethodsAntiseraA 15-mer peptide corresponding to the N-terminus of hu-man PABP2 protein (GRGSGPGRRRHLVPG) was used toraise antisera, termed anti-NterPABP2, using standard pro-tocols.13 We also used antisera raised against bovine PABP2(Dr Wahle, University of Giessen) and EGFP (Clontech,Palo Alto, CA).

TransfectionsCOS-7 cells were seeded at 1.5 3 105 cells per well in six-wellplates containing sterile coverslips. In each experiment, 2.0 mgof plasmid DNA was transfected using lipofectamine (GibcoBRL) in OPTI-MEM, using the manufacturer’s protocol.

ImmunocytochemistryCells grown on coverslips were fixed in 4% paraformalde-hyde in purified buffered saline (PBS), mounted, permeabil-ized with 0.05% Triton-X, blocked with 10% normal goatserum for 1 hour, and incubated with primary antibody(1:1,000) at room temperature overnight. Cells were rinsedin PBS and incubated with biotinylated secondary antibody(1:500). Amplification was performed using the ABC kit

(Vector; Burlingame, CA) and products visualized using theVIP kit (Vector).

Western BlottingCells from 100-mm plates transfected with various constructswere scraped, collected, and lysed in Laemmli sample buffer.Proteins were quantitated using the Bio-Rad (Hercules, CA)protein assay. One hundred micrograms of total proteinfrom each sample were electrophoresed on a 10% polyacryl-amide gel and transblotted to nitrocellulose membrane. Im-munodetection was performed with immunopurified anti-NterPABP2 (1:1,000), anti-bovine PABP2 (1:1,000), andanti-GFP (1:10,000). Horseradish peroxidase–conjugatedsecondary antibodies were used (1:10,000) and visualized bychemiluminescence (Renaissance: NEN, Boston, MA).

Plasmid ConstructionThe longest PABP2 cDNA available (;950 bp) did not havethe complete 59 sequence. The 59 fragment of the cDNA wasremoved and replaced with a genomic RsaI and Bsu36I frag-ment (complete 59 end) of either the normal or mutant (13GCGs) PABP2. The insert was digested with HindIII andAccI and cloned in-frame into pEGFP-N1 (Clontech) withthe EGFP at the 39 end of the PABP2 gene.

ResultsCellular Distribution of PABP2Normal and expanded (GCG)13 PABP2 cDNA fusedin-frame to EGFP and driven by the cytomegaloviruspromoter were transiently transfected into COS-7 cellsand the protein detected as green fluorescence at regu-lar time points (12, 24, 48, and 72 hours posttransfec-tion). At 12 and 24 hours, in both constructs, therewas diffuse nuclear fluorescence, consistent with previ-ous studies of PABP2.7,14 At 48 and 72 hours post-transfection, normal PABP2-transfected cells continuedto show diffuse nuclear fluorescence (Fig 1b). Withmutant PABP2, however, the fluorescence was now ex-clusively observed as nuclear punctate patches (see Fig1a). Transfection with pEGFP-N1 vector alone re-sulted in fluorescence throughout the cell.

Protein ExpressionExpression of PABP2-EGFP in transiently transfectedCOS-7 cells was analyzed using anti-NterPABP2 andanti-GFP antisera. Equal amounts of protein lysatesfrom normal and mutant PABP2, as well as pEGFP-N1or mock transfected cells (Fig 2, lanes 1–4) were ana-lyzed. No difference in migration was observed betweennormal and mutant protein, although the latter hadseven additional alanine residues. Figure 2A shows thepresence of endogenous PABP2 protein migrating as a50-kd band in all lanes and the expected 80-kd fusionprotein in transfected cells. Similar results were obtainedusing immunopurified anti-NterPABP2 and bovine anti-bovine PABP2 (data not shown). The same blot wasstripped and reprobed with polyclonal EGFP antibody.

Brief Communication: Shanmugam et al: OPMD Nuclear Inclusions 799

Both normal and mutant PABP2-EGFP fusion proteinswere detected as an 80-kd band (see Fig 2b, lanes 1 and2), confirming our previous observations using antibod-ies against PABP2. Cells transfected with the vectoralone (see Fig 2b, lane 3) detected a 30-kd band, thepredicted size of EGFP protein. Mock transfected cellsshowed no bands using this antibody. Similar observa-tions were made with cell extracts from 24, 48, 72, and96 hours. At 96 hours and later, the fluorescence beganto decrease as cells were detaching from the plate.

COS-7 cells transfected with normal or mutant con-structs were immunostained at different time points us-ing anti-NterPABP2. At 24 and 48 hours, both con-structs showed heavy nuclear staining. At 72 hours,compact nuclear inclusions were observed only in cellstransfected with the mutant PABP2 construct (Fig 3a),whereas cells transfected with the normal construct (seeFig 3b) still showed diffuse nuclear staining. Cellstransfected with pEGFP-N1 vector alone (see Fig 3c)or mock transfected (see Fig 3d) were used as controls.Cells transfected with mutant construct showed onemajor intranuclear inclusion with a few cells also hav-ing two to three smaller inclusions. The inclusion pat-tern was consistent in all transfected cells. These dataare also consistent with our previous observations, inwhich compact fluorescent nuclear patches were seenunder fluorescent microscopy at 72 hours in the cellstransfected with the mutant PABP2.

DiscussionOur data suggest that the expanded polyalanine tractfrom the PABP2 protein as seen in OPMD patients issufficient to cause nuclear aggregates in cultured mam-

Fig 1. Distribution of the normal and expanded PABP2-EGFP fusion protein in transfected COS-7 cells. Cells transfected with theconstructs were observed for the expression of the fluorescent fusion protein at 72 hours posttransfection: (a) cells transfected with themutated PABP2 cDNA; (b) cells transfected with normal PABP2 (both the normal and the expanded PABP2 cDNAs have EGFPin frame); and (c and d) cells transfected with pEGFP-N1 vector alone and mock transfected cells, respectively. (Magnification,3100.)

Fig 2. Western blot analysis of COS-7 cell extracts transfectedwith the normal and expanded PABP2 constructs. Equalamounts of cell extracts from COS-7 cells transfected with variousconstructs were immunodetected using immunopurified anti-PABP2 polyclonal antiserum or polyclonal EGFP antiserum. (A)Lane 1, lysates of COS-7 cells containing the expanded PABP2construct; lane 2, lysates expressing the normal PABP2 construct;and lanes 3 and 4, cell lysates of COS-7 cells transfected withvector alone and mock transfected cells, respectively. (A) Cellsprobed with anti-PABP2 serum. (B) The same blot was strippedof the original antibody and reprobed with anti-EGFP serum.

800 Annals of Neurology Vol 48 No 5 November 2000

malian cells. We provide here an in vivo model for theformation of nuclear inclusions in OPMD.

A small increase in the polyalanine repeat from(GCG)6 to (GCG)13 causes aggregation in our system,suggesting that a threshold exists for polyalanine tract–induced formation of inclusions. It remains unclearhow aggregation occurs, however. We may speculatethat the expansion of the alanine tract beyond a criticalnumber may lead to formation of degradation resistantoligopeptides that aggregate and subsequently recruitnormal PABP2 protein. An alternate model would bethat mutant PABP2 protein is abnormally folded, re-sists degradation, and aggregates as insoluble inclu-sions. The fact that normal PABP2 protein formsdimers and oligomers6 is consistent with either model.Further experiments are needed to differentiate be-tween these two models.

Our results show that expression of the PABP2cDNA bearing an extra seven GCG codons requires 2to 3 days before formation of nuclear inclusions (seeFig 1a). This delay in inclusion formation may resultfrom the time needed to obtain a critical quality ofmutant protein. Alternatively, spontaneous conforma-tional changes in the protein induced by the polyala-nine tracts may take time to occur,15 eventually leadingto protein-protein aggregation.

Immunostaining transfected cells using PABP2 anti-bodies raised against the N-terminus shows nuclear in-clusions only with the mutant protein (see Fig 3a). Inaddition, inclusions also contain EGFP, which is at-

tached to the C-terminus, suggesting that the inclusioncontains intact PABP2 protein. These results are consis-tent with the identification of intact filamentous aggre-gates of PABP2 in muscle fibers from OPMD pa-tients.16 We propose a model in which proteinconformation is altered by the expanded polyalaninetract, leading to abnormal protein-protein interactions.Over time these proteins accumulate as insoluble nuclearaggregates, leading to premature cell death. While spec-ulative, this model is supported by the fact that nuclearinclusions are also a common phenomenon in polyQ/CAG triplet disorders, which are also thought to be ini-tiated by altered protein conformation17 and proteinmisfolding.18 These aggregates are resistant to degrada-tion and may prevent ubiquitin-mediated proteolytic de-struction by disrupting the proteasome machinery.19

The fact that we see a similar phenomenon in a polyA/GCG triplet disease suggests that the nucleus may beparticularly susceptible to aggregate formation.

This work was supported by the Muscular Dystrophy Association(USA) and the Federation Foundation of Greater Philadelphia. DrRouleau is supported by the Medical Research Council of Canada. DrBrais is a chercheur-boursier of the Fonds de Recherche en Sante duQuebec.

We thank Dr Elmar Wahle for providing the bovine PABP2 antibody.

References1. Bouchard J-P, Brais B, Tome FMS. Oculopharyngeal muscular

dystrophy. In: Emery AHE, ed. Neuromuscular disorders: clin-

Fig 3. Immunocytochemical detection of intranuclear inclusions in COS-7 cells transfected with normal and expanded PABP2cDNAs. Immunostaining was performed with immunopurified anti-PABP2 antiserum: (a) COS-7 cells transfected with theexpanded PABP2; (b) cells transfected with normal PABP2; (c) cells transfected with pEGFP-N1 vector alone; (d) mocktransfected cells. (Magnification, 3100.)

Brief Communication: Shanmugam et al: OPMD Nuclear Inclusions 801

ical and molecular genetics. Chichester, UK: John Wiley &Sons, 1998:157–179

2. Brais B, Rouleau GA, Bouchard J-P, et al. Oculopharyngealmuscular dystrophy. Semin Neurol 1999;19:59–66

3. Tome FMS, Fardeau M. In: Engel AG, Franzini-Armstrong C,eds. Myology. New York: McGraw-Hill, 1994:1233–1245

4. Brais B, Xie Y, Sanson M, et al. The oculopharyngeal musculardystrophy locus maps to the region of the cardiac a and bmyosin heavy chain genes on chromosome 14q11.2–q13. HumMol Genet 1995;4:429–434

5. Brais B, Bouchard J-P, Xie Y, et al. Short GCG expansions inthe PABP2 gene cause oculopharyngeal muscular dystrophy.Nat Genet 1998;18:164–167

6. Nemeth A, Krause S, Blank D, et al. Isolation of genomic andcDNA clones encoding bovine poly(A) binding protein II. NuclAcids Res 1995;23:4034–4041

7. Krause S, Fakan S, Weis K, Wahle E. Immunodetection ofpoly(A) binding protein II in the cell nucleus. Exp Cell Res1994;214:75–82

8. Bienroth S, Keller W, Wahle E. Assembly of a processive mes-senger RNA polyadenylation complex. EMBO J 1993;12:585–594

9. Wahle E. A novel poly(A)-binding protein acts as a specificityfactor in the second phase of messenger RNA polyadenylation.Cell 1991;66:759–768

10. Tome FMS, Fardeau M. Nuclear inclusions in oculopharyngealmuscular dystrophy. Acta Neuropathol 1980;49:85–87

11. Blondelle SE, Forood B, Houghten RA, Perez-Paya E. Poly-alanine-based peptides as models for self-associated b-pleatedsheet complexes. Biochemistry 1997;36:8393–8400

12. Forood B, Perez-Paya E, Houghten RA, Blondelle SE. Forma-tion of an extremely stable polyalanine B-sheet macromolecule.Biochem Biophy Res Comm 1995;211:7–13

13. Harlow E, Lane D. Antibodies. Cold Spring Harbor, NY: CSHLaboratory Press, 1988

14. Calodo A, Kutay U, Kuhn U, et al. Deciphering the cellularpathway for transport of poly(A)-binding protein II. RNA2000;6:245–256

15. Onodera O, Burke JR, Miller SE, et al. Oligomerisation ofexpanded-polyglutamine domain fluorescent fusion proteins incultured mammalian cells. Biochem Biophys Res Comm 1997;238:599–605

16. Calado A, Tome FMS, Brais B, et al. Nuclear inclusions inmuscle fibres from patients with oculopharyngeal muscular dys-trophy are filamentous aggregates of poly(A) binding protein 2.Congres International de Myologie, Nice, France. Resumes descommunications. Myology 2000;465:117

17. Perez MK, Paulson HL, Pendse SJ, et al. Recruitment and therole of nuclear localization in polyglutamine-mediated aggrega-tion. J Cell Biol 1999;143:1457–1470

18. Cummings CJ, Mancini MA, Antalffy BA, et al. Chaperone sup-pression of aggregation and altered subcellular proteosome localiza-tion imply misfolding in SCA1. Nat Genet 1998;19:148–154

19. Johnston JA, Ward CL, Kopito RR. Aggresomes: a cellular re-sponse to misfolded proteins. J Cell Biol 1997;143:1883–1898

High Mortality in NipahEncephalitis Is Associatedwith Presence of Virus inCerebrospinal FluidKaw Bing Chua, FRCP,* Sai Kit Lam, FRCPath,*Chong Tin Tan, FRCP,† Poh Sim Hooi, DipMLT,*Khean Jin Goh, MRCP,† Nee Kong Chew, MRCP,†Kay Sin Tan, MRCP,† Adeeba Kamarulzaman, FRACP,†and Kum Thong Wong, FRCPath‡

During the outbreak of Nipah virus encephalitis in Ma-laysia, stored cerebrospinal fluid (CSF) samples from 84patients (27 fatal and 57 nonfatal cases) were cultured forthe virus. The virus was isolated from 17 fatal cases and1 nonfatal case. There were significant associations be-tween CSF virus isolation and mortality as well as clinicalfeatures associated with poor prognosis. In addition,there was a positive linear correlation of CSF virus isola-tion with age. There was no significant association be-tween CSF virus isolation and the character of the CSF,presence of Nipah-specific antibody in the serum or CSF,duration of illness before collection of samples, or sex orethnicity of the patients. This study suggests that highviral replication in the central nervous system may be animportant factor for high mortality.

Chua KB, Lam SK, Tan CT, Hooi PS, Goh KJ,Chew NK, Tan KS, Kamarulzaman A,

Wong KT. High mortality in Nipahencephalitis is associated with presence

of virus in cerebrospinal fluid.Ann Neurol 2000;48:802–805

A novel paramyxovirus, now named Nipah virus, wasfirst isolated from the cerebrospinal fluid (CSF) of apatient from Sungei Nipah village in March 1999 andsubsequently identified as the etiological agent respon-sible for the outbreak of a fatal encephalitis in Malay-sia.1–4 The infection subsequently spread to neighbor-ing Singapore, where it involved abattoir workers.5,6

This virus is antigenically related to, but distinct from,Hendra virus, which was first isolated in September1994 during an outbreak of a severe respiratory illnessthat killed 14 horses and 2 humans in Queensland,Australia.7–12 The outbreak of Nipah virus encephalitis

From the Departments of *Medical Microbiology, †Medicine, and‡Pathology, Faculty of Medicine, University of Malaya, Kuala Lum-pur, Malaysia.

Received Apr 24, 2000, and in revised form Jun 29. Accepted forpublication Jul 7, 2000.

Address correspondence to Prof C. T. Tan, Neurology Laboratory,University Hospital, Kuala Lumpur 59100, Malaysia.

802 Copyright © 2000 by the American Neurological Association

started in September 1998 and reached its peak inMarch 1999, with 268 human cases reported and 105fatalities. The virus also caused a respiratory and cen-tral nervous system (CNS) illness in pigs, and most ofthe human cases occurred in individuals who had oc-cupational exposure to infected pigs.1–3 The spread ofthe disease was due to the movement of infected pigsfrom farm to farm. The outbreak was controlled onlyafter successful isolation and identification of the etio-logical agent followed by the culling of a million pigsin the infected areas.

Between February and June 1999, 110 patients sus-pected of Nipah encephalitis were admitted to the Uni-versity of Malaya Medical Centre (UMMC), KualaLumpur.13 From the CSF of 18 patients, Nipah viruswas initially isolated from 5 fatal cases but from noneof the patients who survived. In this current study,Nipah virus isolation was further attempted on CSFsamples to determine the possible relation betweenmortality and virus isolation.

Patients and MethodsPatientsPatients included in the study were defined to have Nipahvirus encephalitis if they:

1. Had evidence of encephalitis:a. Clinical features of fever, headache, altered senso-

rium or focal neurological signs; orb. Abnormal CSF findings, more than 6 leukocytes/ml,

or protein levels above 0.45 g/L; orc. Characteristic brain magnetic resonance scan findings

2. Came from known outbreak areas3. Had direct or close contact with pigs or other infected

animals

The clinical criteria of Nipah virus infection wheneverpossible was supported by the presence of positive cross-reacting antibodies against Hendra virus antigen in the bloodor CSF. Microscopic, biochemical, and serological investiga-tions were performed on all the CSF obtained. Serum andCSF samples were tested by IgM capture and enzyme-linkedimmunosorbent assay (ELISA) using culture cell slurry de-rived from Vero E6 cells infected with Hendra virus aftertreatment by gamma radiation. Detection of specific IgG wasby indirect ELISA using cell lysate derived from Hendra-infected Vero E6 cells. Pooled serum samples derived fromlocal healthy blood donors were used as the background neg-ative control.

Virus IsolationVirus isolation was carried out in a modified biosafety level 3laboratory. CSF stored in a 220°C freezer for 3 to 4 monthswas thawed and 100 ml of each sample was inoculated intoeach well of a 24-well culture plate preseeded with 1 ml of1 3 105 Vero cells (ATCC, CCL81) in Eagle’s minimal es-sential growth medium. The culture plate was sealed withadhesive tape, incubated at 37°C, and examined daily for the

presence of characteristic syncytial formation of multinucle-ated giant cells. Infected cells were harvested, washed twicewith sterile phosphate buffered saline, and centrifuged; thecells were fixed onto Teflon-coated slides with cold acetoneand were ultraviolet irradiated. The identity of the viruswas confirmed by indirect immunofluorescence using anti-Hendra hyperimmune mouse ascitic fluid as primary anti-body and goat anti-mouse fluorescein isothiocyanate-conju-gated IgG as secondary antibody.

Statistical AnalysisThe x2 test and x2 for linear trend were used to analyze theassociation between the isolation rate of Nipah virus fromCSF and the clinical outcome of the disease, the presence ofcertain clinical features related to poor prognosis, and theNipah serological results. All significant results were based onp , 0.05.

ResultsNinety-one patients were defined by the above criteriato have Nipah encephalitis. However, stored CSF sam-ples for virus isolation were available from only 84 pa-tients (27 fatal and 57 nonfatal cases). Nipah virus wasisolated from the CSF of 17 fatal cases and 1 nonfatalcase (Table 1). Positive virus isolation from CSF wassignificantly associated with mortality (x2 5 37.21;p 5 0.0000).

The mean age of the patients included in this studywas 37.5 years (age range, 13–68 years). The positiveisolation rate correlated with the increasing age of pa-tients (x2 for linear trend 5 6.96; p 5 0.006). Theisolation of the virus also significantly correlated withclinical features that indicate severe CNS involvementand poor prognosis. These include segmental myoclo-nus, altered conscious level (Glasgow Coma Scale score,15), the need for ventilatory support, tachycardia,and hypertension13 (see Table 1).

However, virus isolation was not significantly associ-ated with sex, ethnicity, or a history of Japanese en-cephalitis vaccination (see Table 1). Positive virus iso-lation was also not associated with abnormal CSFbiochemistry, cell count (Table 2), or duration of ill-ness at the time of CSF collection (x2 for lineartrend 5 0.06; p 5 0.861). The presence of anti–Nipahvirus antibodies in the CSF or serum did not appear toinfluence the virus isolation (see Table 1).

DiscussionThis study shows that positive virus isolation from theCSF of patients with Nipah encephalitis is associatedwith high mortality, a finding similar to Japanese en-cephalitis.14 Positive virus isolation also correlated withpoor clinical prognostic features. The rate of positiveviral isolation increased with age, and this was consis-tent with increased mortality risk in older patients.These findings differed from another study thatshowed that Nipah virus isolation from urine and

Brief Communication: Chua et al: Mortality and Virus Isolation in Nipah Encephalitis 803

throat secretions was not associated with poor clinicaloutcome or age.15

One of the main pathological findings in Nipah en-cephalitis was disseminated microinfarction secondary

to vasculitis-induced thrombosis of the cerebral bloodvessels.4 MRI of the brain in acute, nonfatal cases alsoshowed disseminated small, discrete high-signal lesionson T2-weighted and fluid attenuated inversion recov-

Table 1. Virus Isolation, Demographic, Clinical, and Serological Profiles of Patients with Nipah Encephalitis

Characteristic Features No. Total

Virus Isolation (no.) Statistical Test

Positive Negative x2 Valuea p Value

MortalityFatal 27 17 10 37.21 0.0000Nonfatal 57 1 56

SexMale 69 17 52 Fisher exact 0.1740Female 15 1 14

RaceChinese 70 18 52 4.58 0.1012Indian 13 0 13Other 1 0 1

JE vaccinationYes 65 13 52 Fisher exact 0.5400No 19 5 14

HeadachePresence 55 8 47 3.38 0.0661Absence 29 10 19

MyoclonusPresence 27 10 17 4.47 0.0344Absence 57 8 49

VentilatedYes 44 17 27 14.18 0.0002No 40 1 39

ComaPresence 49 17 32 10.47 0.0012Absence 35 1 34

TachycardiaPresence 34 17 17 24.92 0.0000Absence 50 1 49

HypertensionPresence 35 15 20 14.25 0.0002Absence 49 3 46

CSFAbnormal 66 13 53 0.17 0.5206Normal 18 5 13

Anti-Nipah antibody (CSF)Positive 14 2 12 Fisher exact 1.0000Negative 36 6 30

Anti-Nipah antibody (serum)Positive 39 4 35 Fisher exact 0.4891Negative 30 5 25

ax2 value: after Yates corrected value.

JE 5 Japanese encephalitis; CSF 5 cerebrospinal fluid.

Table 2. Cerebrospinal Fluid of Patients with Nipah Encephalitis

No. of PatientsWhite Cell Counta

(cells/ml)Proteina

(g/L)

Viral isolation negative 18 49 (0–842) 0.72 (0.29–2.15)Viral isolation positive 66 45 (0–700) 0.72 (0.27–2.12)p value 0.931 0.977

aMean values are shown, with the range in parentheses.

Table 1. Virus Isolation, Demographic, Clinical, and Serological Profiles of Patients with Nipah Encephalitis

Characteristic Features No. Total

Virus Isolation (no.) Statistical Test

Positive Negative x2 Valuea p Value

MortalityFatal 27 17 10 37.21 0.0000Nonfatal 57 1 56

SexMale 69 17 52 Fisher exact 0.1740Female 15 1 14

RaceChinese 70 18 52 4.58 0.1012Indian 13 0 13Other 1 0 1

JE vaccinationYes 65 13 52 Fisher exact 0.5400No 19 5 14

HeadachePresence 55 8 47 3.38 0.0661Absence 29 10 19

MyoclonusPresence 27 10 17 4.47 0.0344Absence 57 8 49

VentilatedYes 44 17 27 14.18 0.0002No 40 1 39

ComaPresence 49 17 32 10.47 0.0012Absence 35 1 34

TachycardiaPresence 34 17 17 24.92 0.0000Absence 50 1 49

HypertensionPresence 35 15 20 14.25 0.0002Absence 49 3 46

CSFAbnormal 66 13 53 0.17 0.5206Normal 18 5 13

Anti-Nipah antibody (CSF)Positive 14 2 12 Fisher exact 1.0000Negative 36 6 30

Anti-Nipah antibody (serum)Positive 39 4 35 Fisher exact 0.4891Negative 30 5 25

ax2 value: after Yates corrected value.

JE 5 Japanese encephalitis; CSF 5 cerebrospinal fluid.

Table 2. Cerebrospinal Fluid of Patients with Nipah Encephalitis

No. of PatientsWhite Cell Counta

(cells/ml)Proteina

(g/L)

Viral isolation negative 18 49 (0–842) 0.72 (0.29–2.15)Viral isolation positive 66 45 (0–700) 0.72 (0.27–2.12)p value 0.931 0.977

aMean values are shown, with the range in parentheses.

804 Annals of Neurology Vol 48 No 5 November 2000

ery (FLAIR) images, which seemed to correspond tothe microinfarction seen in fatal cases, indicating vas-culitis could be a key event. However, there has beenno correlation between these lesions with depth ofcoma and outcome of patients.16 Clinically, it was pre-viously shown that poor prognostic signs in Nipah en-cephalitis, such as altered conscious level, impaireddoll’s eye reflex, abnormal pupils, hypertension, tachy-cardia, vomiting, and segmental myoclonus, indicatedsevere brainstem involvement.13 Yet there was no pre-disposition of microinfarction in the brainstem in fatalcases pathologically.4 Hence, microinfarction alonemay not be sufficient to account for the high mortalityand the associated high virus isolation rate.

Direct neuronal involvement in the brainstem maybe important in the pathogenesis of fatal Nipah en-cephalitis. Pathologically, there is evidence of direct vi-ral attack with neuronal inclusion body and positiveimmunohistochemical staining.4,6,17 Clinically, thereare characteristic features, such as segmental myoclo-nus, hypotonia, and areflexia, that indicate a predilec-tion for certain groups of neurons, including those inthe brainstem.13

Pathologically, meningitis is usually mild in Nipahencephalitis.4 This is consistent with the observationthat 25% of the initial CSF was normal.13 There isalso lack of correlation between virus isolation and ab-normal CSF. Thus, viral isolation from the CSF prob-ably does not reflect extent of viral replication in thecells lining the meninges. On the other hand, there isinflammation in the brain parenchyma adjacent to sub-arachnoid space (Wong KT, unpublished observation),which would allow virus in the parenchyma to reachthe CSF. This study showed an association betweenCSF virus isolation and mortality. This support by di-rect neuronal invasion with high viral replication maybe important in the pathogenesis of severe disease.

The presence of anti–Nipah virus antibodies in theCSF did not influence the isolation of the virus, incontrast to the study on virus isolation in urine andthroat secretions in which positive viral isolation corre-lated with the presence of specific antibodies.15 Thismay suggest that humoral immune response in theCNS probably plays a minor role in the disease processand recovery. The appearance of antibodies to Nipahvirus appears to be dependent on the viremic phase,which is associated with diffuse viral spread throughoutthe body, but is unrelated to the presence of virus inthe CNS. The latter, however, is more important indetermining the morbidity and mortality.

We thank Dr T. G. Ksiazek and the Centers for Disease Control inAtlanta for carrying out the serological testing of the Nipah virusand the supply of reagents for the continuation of serological testing.

References1. Centers for Disease Control and Prevention, Atlanta. Outbreak

of Hendra-like virus—Malaysia and Singapore, 1998–1999.MMWR 1999;48:265–269

2. Centers for Disease Control and Prevention, Atlanta. Outbreakof Nipah virus—Malaysia and Singapore, 1999. MMWR 1999;48:335–337

3. Centers for Disease Control and Prevention, Atlanta. Outbreakof Hendra-like virus—Malaysia and Singapore, 1998–1999.JAMA 1999;281:1787–1788

4. Chua KB, Goh KJ, Wong KT, et al. Fatal encephalitis due toNipah virus among pig-farmers in Malaysia. Lancet 1999;354:1257–1259

5. Lee KE, Umapathi T, Tan CB, et al. The neurological mani-festations of Nipah virus encephalitis, a novel paramyxovirus.Ann Neurol 1999;46:428–432

6. Paton NI, Leo YS, Zaki SR, et al. Outbreak of Nipah virusinfection among abattoir workers in Singapore. Lancet 1999;354:1253–1256

7. Murray K, Selleck P, Hooper P, et al. A morbillivirus thatcaused fatal disease in horses and humans. Science 1995;268:94–97

8. O’Sullivan JD, Allworth AM, Paterson DL, et al. Fatal enceph-alitis due to novel paramyxovirus transmitted from horses. Lan-cet 1997;349:93–95

9. Selvey LA, Wells RM, McCormack JG, et al. Infection of hu-mans and horses by a newly described morbillivirus. Med JAust 1995;162:642–645

10. Rogers RJ, Douglas IC, Baldock FC, et al. Investigation of asecond focus of equine morbillivirus infection in coastalQueensland. Aust Vet J 1996;74:243–244

11. Murray K, Eaton B, Hooper P, et al. Flying foxes, horses, andhumans: a zoonosis caused by a new member of the Paramyxo-viridae. In: Scheld WM, Armstrong D, Hughes JM, et al, eds.Emerging Infection 1. Washington DC: ASM Press, 1998:43–58

12. McCormack JG, Allworth AM, Selvey LA, Selleck PW. Trans-missibility from horses to humans of a novel paramyxovirus,equine morbillivirus (EMV). J Infect 1999;38:22–23

13. Goh KJ, Tan CT, Chew NK, et al. Clinical features of Nipahencephalitis among pig farmers in Malaysia. N Engl J Med2000;342:1229–1235

14. Burke DS, Lorsomrudee W, Leake CJ, et al. Fatal outcome inJapanese encephalitis. Am J Trop Med Hyg 1985;34:1203–1210

15. Chua KB, Lam SK, Goh KJ, et al. The presence of Nipah virusin respiratory secretions and urine of patients during an out-break of Nipah virus encephalitis in Malaysia. J Infect (In press)

16. Ahmad Sarji S, Abdullah BJJ, Goh KJ, et al. MR imaging fea-tures of Nipah encephalitis. AJR 2000;175:437–442

17. Chua KB, Bellini WJ, Rota PA, et al. Nipah virus: a recentlyemergent deadly paramyxovirus. Science 2000;288:1432–1435

Brief Communication: Chua et al: Mortality and Virus Isolation in Nipah Encephalitis 805

Variant Alzheimer’s Diseasewith Spastic Paraparesis andCotton Wool Plaques IsCaused by PS-1 Mutationsthat Lead to ExceptionallyHigh Amyloid-bConcentrationsHenry Houlden,*† Matt Baker,† Eileen McGowan,†Patrick Lewis,† Mike Hutton,† Richard Crook,†Nicholas W. Wood,* Samir Kumar-Singh,‡Jennian Geddes,§ Michael Swash,§ Francesco Scaravilli,*Janice L. Holton,* Tammaryn Lashley,* Taisuke Tomita,\Tadafumi Hashimoto,\ Auli Verkkoniemi,¶Hannu Kalimo,# Mirja Somer,** Anders Paetau,¶Jean-Jacques Martin,†† Christine Van Broeckhoven,‡Todd Golde,† John Hardy,† Matti Haltia,¶and Tamas Revesz*

We describe 3 new families affected by Alzheimer’s diseasewith spastic paraparesis. In affected individuals, includingthe earliest known patient with this clinical syndrome,neuropathological examination revealed large “cottonwool” plaques similar to those we have previously de-scribed in a Finnish family. In the families in which DNAwas available, presenilin-1 mutations were observed.Transfection of cells with these mutant genes caused ex-ceptionally large increases in secreted Ab42 levels. Further-more, brain tissue from individuals with this syndromehad very high amyloid-b concentrations. These findingsdefine the molecular pathogenesis of an important sub-group of Alzheimer’s disease and have implications for thepathogenesis of the disease in general.

Houlden H, Baker M, McGowan E,Lewis P, Hutton M, Crook R, Wood NW,

Kumar-Singh S, Geddes J, Swash M,

Scaravilli F, Holton JL, Lashley T, Tomita T,Hashimoto T, Verkkoniemi A, Kalimo H,

Somer M, Paetau A, Martin J-J,Van Broeckhoven C, Golde T, Hardy J,

Haltia M, Ravesz T. Variant Alzheimer’s diseasewith spastic paraparesis and cotton wool plaques

is caused by PS-1 mutations that lead toexceptionally high amyloid-b concentrations.

Ann Neurol 2000;48:806–808

The molecular causes of two distinct diseases character-ized by the combination of progressive dementia andspastic paraparesis have recently been defined: (1) avariant of Alzheimer’s disease (AD) with spastic para-paresis and large, nonneuritic “cotton wool” plaquescaused by a variant of the PS1D9 mutation1,2 and (2)familial British dementia caused by a mutation in anovel gene, BRI.3,4 These discoveries led us to examineour sample archives to look for families with similarclinical syndromes and to subject these to neuropatho-logical and molecular analysis. During this search, weidentified a new branch of the familial British demen-tia kindred3,4 (Houlden, 2000, unpublished data). Wealso identified a further 3 families in which the clinicalfeatures were reminiscent of those in the original Finn-ish family.1,2,5 Here, we report on our pathological ex-amination of these new families and compared thiswith the original Finnish families, and we report ouridentification of presenilin-1 mutations within them.We also report analysis of the effects of these mutationson Ab production.

MethodsThe family material is summarized in Table 1.

Standard neuropathological examination was carried outon fixed specimens. Amyloid-b (Ab) staining (Fig 1) of tis-sue sections was with the Dako anti-Ab antibody (Carpen-teria, CA), which does not discriminate between Ab40 andAb42. Initially, analysis was carried out to define whetherthese samples had either of the two variants of the PS1D9mutation.1,2,6 When this mutation was not found, the wholepresenilin-1 gene was sequenced.7 Wild-type and D9 PS1cDNA constructs (Invitrogen, Carlsbad, CA) were cloned asdescribed previously.8 The mutation causing deletion ofcodons 83 and 84 (DelIM; see later) was introduced into thewild-type PS1 cDNA in pcDNA3 (Invitrogen) by site-directed mutagenesis using the Transformer mutagenesis sys-tem (Clontech, Palo Alto, CA). The P436Q mutation wasintroduced by polymerase chain reaction (PCR) mutationamplification of the PS1 gene using the Roche High FidelityPCR kit (Roche, Indianapolis, IN).9 The amplified sequencewas purified and cloned into the pAG3 vector. All constructswere transfected into H4 glioma cells using Fugene transfec-tion reagent (Roche).

H4 glioma cells were cultured in Optimem media with1% penicillin/streptomycin and 10% fetal bovine serum(Gibco, Rockville, MD) with neomycin (800 mg/ml) to se-

From the *Department of Clinical Neurology and Neuropathology,Institute of Neurology, Queen Square, and §Department of MorbidAnatomy, Royal London Hospital, London, UK; †Laboratory ofNeurogenetics, Mayo Clinic Jacksonville, Jacksonville, FL; ‡Depart-ment of Molecular Genetics, Flanders Interuniversity Institute forBiotechnology, and ††Department of Neurology, University Hospi-tal Antwerpen, Born Bunge Foundation, University of Antwerp,Antwerpen, Belgium; \Department of Neuropathology and Neuro-science, Graduate School of Pharmaceutical Sciences, University ofTokyo, Tokyo, Japan; ¶Departments of Neurology and Pathology,Helsinki University Central Hospital and University of Helsinki,and **Department of Clinical Genetics, The Family Federation ofFinland, Helsinki; and #Department of Pathology, Turku Univer-sity Hospital, Turku, Finland.

Received May 16, 2000, and in revised form Jul 10. Accepted forpublication Jul 14, 2000.

Address correspondence to Dr Hardy, Center for Neuroscience,Mayo Clinic Jacksonville, Jacksonville, FL 32224.

806 Copyright © 2000 by the American Neurological Association

lect for delIM and D9 PS1 expression or hygromycin (800mg/ml) to select for P436Q PS1 expression. After 2 weeksunder selection, cells were transfected with APP695 K/M670/1N/L cDNA in pAG3 overnight. The cells were incubated for24 hours then were lysed in 200 ml of 1% triton in TBS withproteinase inhibitor complete cocktail (Roche).

Media samples were assayed for Ab40, Ab42, and secretedAPP (sAPP). The Ab40 and Ab42 were captured using theBan50 antibody (transfection studies) or BNT77 (brain amy-loid studies) and detected (using BA27 HRP and BC05HRP antibodies, respectively, each of which is specific to oneof the two peptides) with peroxidase substrate/solution(Kirkegaard and Perry Laboratories, Gaithersburg, MD).sAPP was assayed using a cELISA based around the 207 N

terminal APP antibody (B. Greenberg). Ab42 as a percentageof Ab40 was then calculated.9

Results and DiscussionAll of the families in which spastic paraparesis was aprominent feature, including that family (Family CO)in which this clinical feature was first described,10

shared the characteristic pathology of cotton woolplaques previously described in the family FINN2.This, therefore, represents a variant of AD. In all thecases in which DNA was available, presenilin-1 muta-tions were found: family FINN2 is the one in whichwe have previously described the PS1D9 mutation1,2,5;in family EB, an unusual and novel deletion of codons83 and 84 removing isoleucine and methionine fromexon 4 (the DelIM mutation); and in family D, thepreviously reported P436Q mutation.11 These find-ings, together with previous reports, are of considerableinterest. First, it is clear that not all cases of the PS1D9mutation necessarily have spastic paraparesis. Thus, insome cases with both the Finnish variant of the muta-tion and in others,1,5,6 dementia is the presenting fea-ture. In some cases with dementia as the presentingfeature, it seems as though the pathology is of moretypical AD with more neuritic, rather than cottonwool, plaques,2,5,12 although this conclusion remains

Table 1. Clinical Summary

Family GenerationsAffected(Path) First Clinical Presentation (n)

Mean Age of Onset ofSpastic Paraplegia, yr (range)

Finn 2(D9) 4 22 (5) Spastic paraparesis (9) 50 (45–55)Dementia (5)

D (P436Q) 2 3 (1) Spastic paraparesis 29 (20–42)EB (DelIM) 3 5 (1) Spastic paraparesis 36 (34–38)CO 3 5 (1) Spastic paraparesis (2)

Dementia (2) 31 (26–38)Spastic paraparesis/dementia (1)

Table 2. Shift in Ab42/40 due to SpasticParaparesis Mutations

PS1 MutationMean Ab42/Ab40(SD)

P436Q 0.34a (0.05)D9 0.19a (0.02)DelIM 0.21a (0.02)M139V 0.15 (0.01)PS1wt 0.08 (0.01)

The shift in Ab42/Ab40 due to spastic paraparesis mutations is sig-nificantly different to that shown by FAD M139V when analyzedby one-way ANOVA: ap , 0.001. In this system, the M139V mu-tation has had the largest effect on Ab42/40 ration of the followingmutations (E280G, A285V, M146L, L286V [Golde, unpublisheddata]). All mutations were significantly different to PS1 wild-typevalues, as we and others have previously shown.

Fig. Photomicrograph of serial sections from Family EB show-ing cotton wool plaques and amyloid angiopathy (arrow). (A,haematoxylin and eosin, 3200; B, Ab immunohistochemistry,3200).

Brief Communication: Houlden et al: Spastic Paraparesis Variant of AD 807

provisional until more extensive neuropathological in-vestigations are performed.

These genetic and pathologic data suggest that thepathogenic route to disease can run down two distinctpaths—either a standard AD route (dementia first,with neuritic plaques at end stage) or the unusual pathwe describe here (spastic paraparesis first with cottonwool plaques at end stage) (see Fig). This dichotomysuggests some form of threshold effect in the patho-genic pathways. Most presenilin mutations (and othercauses of AD) fail to exceed this threshold, but some ofthese mutations have a different effect. We and othershad previously noted that the PS1D9 mutation had aparticularly large effect on Ab42 production in trans-fected cells.8,13 This observation led us to investigatewhether this is a more general effect of spastic parapa-resis–causing mutations (Table 2). Examination of thedata from transfection experiments revealed that all thespastic paraparesis-causing mutations have a larger ef-fect on Ab42 production than those mutations thatlead to the more typical dementia phenotype, althoughthe presenilin active site mutation, G384A, has a largeeffect on Ab42 production without spastic paraparesisphenotype.14,15 This suggests that part of the pathogen-esis of the spastic paraparesis variant relates to theamount of production of Ab42, although it is likely thatthere are also other additional genetic and nongeneticfactors. Examination of Ab42 levels in the brain of in-dividuals with this syndrome confirms that it is associ-ated with particularly large quantities of Ab (Table 3).

These data are provocative: a simple reading of theamyloid cascade hypothesis16 would be the predictionthat those mutations that had the largest effects onAb42 production would be associated with the lowestage at onset of disease. This is not the case: rather,those mutations that have the largest effect on Ab pro-duction are associated with a different clinical andpathological phenotype.

This work was supported by the Mayo Foundation, the NIH/NIAthrough the ADRC, Fund for Scientific Research Flanders, the Bel-gian DWTC Interuniversity Attractionpoles, the Wellcome Trust,and the Academy of Finland (grant number 48173).

Thanks to Takeshi Iwatsubo, MD, and Patrick Cras, MD, for theirhelp on this project.

References1. Crook R, Verkkoniemi A, Perez-Tur J, et al. A variant of Alz-

heimer’s disease with spastic paraparesis and unusual plaques dueto deletion of exon 9 of presenilin 1. Nat Med 1998;4:452–455

2. Prihar G, Verkkoniemi A, Perez-Tur J, et al. Alzheimer diseasePS-1 exon 9 deletion defined. Nat Med 1999;5:1090

3. Worster-Drought C, Greenfield JG, McMenemey WH. A formof familial presenile dementia with spastic paralysis (including thepathological examination of a case). Brain 1940;63:237–254

4. Vidal R, Frangione B, Rostagno A, et al. A stop-codon muta-tion in the BRI gene associated with familial British dementia.Nature 1999;399:776–781

5. Verkkoniemi A, Somer M, Rinne JO, et al. Variant Alzheimer’sdisease with spastic paraparesis. Neurology 2000;54:1103–1109

6. Perez-Tur J, Froelich S, Prihar G, et al. A mutation in Alzhei-mer’s disease destroying a splice acceptor site in the presenilin-1gene. Neuroreport 1995;7:297–301

7. Hutton M, Busfield F, Wragg M, et al. Complete analysis ofthe presenilin 1 gene in early onset Alzheimer’s disease. Neuro-report 1996;7:801–805

8. Mehta ND, Refolo LM, Eckman C, et al. IncreasedAbeta42(43) from cell lines expressing presenilin 1 mutations.Ann Neurol 1998;43:256–258

9. Murphy MP, Hickman LJ, Eckman CB, et al. g-Secretase, evi-dence for multiple proteolytic activities and influence of mem-brane positioning of substrate on generation of amyloid beta pep-tides of varying length. J Biol Chem 1999;274:11914–11923

10. Van Bogaert PL, Maere M, De Smedt E. Sur les formes famil-iales precoces de la maladie d’Alzheimer. Monatsschr PsychNeurol 1940;102:249–301

11. Taddei K, Kwok JB, Kril JJ, et al. Two novel presenilin-1 mu-tations (Ser169Leu and Pro436Gln) associated with very earlyonset Alzheimer’s disease. Neuroreport 1998;9:3335–3339

12. Mann DM, Iwatsubo T, Cairns NJ, et al. Amyloid beta protein(Ab) deposition in chromosome 14-linked Alzheimer’s disease:predominance of Abeta42(43) Ann Neurol 1996;40:149–156

13. Citron M, Westaway D, Xia W, et al. Mutant presenilins ofAlzheimer’s disease increase production of 42-residue amyloidbeta-protein in both transfected cells and transgenic mice. NatMed 1997;3:67–72

14. Martin JJ, Gheuens J, Bruyland M, et al. Early-onset Alzheimer’sdisease in 2 large Belgian families. Neurology 1991;41:62–68

15. De Jonghe C, Cras P, Vanderstichele H, et al. Evidence thatAb42 plasma levels in presenilin-1 mutation carriers do not allow for the prediction of their clinical phenotype. NeurobiolDis 1999;6:280–287

16. Hardy J. Amyloid, the presenilins and Alzheimer’s disease.Trends Neurosci. 1997;20:154–159

17. Hosoda R, Saido TC, Otvos L Jr, et al. Quantification of mod-ified amyloid beta peptides in Alzheimer disease and Down syn-drome brains. J Neuropathol Exp Neurol 1998;57:1089–1095

Table 3. Ab Levels in Brains of Cases with Alzheimer’s Disease (AD)/Spastic Paraparesis Variant

Family Mutation Ab40 (mg/g) Ab42(43) (mg/g) Ab42(43)/Ab40

Sporadic ADa

(n 5 14)1.66 3.14 1.89

EB DelIM 56.18 35.34 0.63D P436Q 20.05 35.21 1.76

aData taken from Hosoda et al.17 Sporadic AD cases processed at the same time as the spastic paraparesis cases showed levels similar to thesehistorical controls.

808 Annals of Neurology Vol 48 No 5 November 2000

Increased 8,12-iso-iPF2a-VIin Alzheimer’s Disease:Correlation of a NoninvasiveIndex of Lipid Peroxidationwith Disease SeverityDomenico Pratico, MD,*Christopher M. Clark, MD,†§Virginia M.-Y. Lee, PhD,‡John Q. Trojanowski, MD, PhD,‡§Joshua Rokach, PhD,\and Garret A. FitzGerald, MD*

The isoprostane 8,12-iso-iPF2a-VI is a sensitive and spe-cific marker of in vivo lipid peroxidation. We found el-evated levels in the urine, blood, and cerebrospinal fluidof patients with Alzheimer’s disease (AD) that correlatedwith measures of cognitive and functional impairment,established biomarkers of AD pathology (cerebrospinalfluid tau and amyloid) and the number of apolipoproteinE «4 alleles. These results suggest that 8,12-iso-iPF2a-VIis a useful biomarker of oxidative damage in AD.

Pratico D, Clark CM, Lee VM-Y,Trojanowski JQ, Rokach J, FitzGerald GA.

Increased 8,12-iso-iPF2a-VI in Alzheimer’s disease:correlation of a noninvasive index of lipid

peroxidation with disease severity.Ann Neurol 2000;48:809–812

Alzheimer’s disease (AD) is a neurodegenerative disor-der that impairs cognition and behavior.1 Neuropatho-logically, the disease is characterized by the presence ofsenile plaques, neurofibrillary tangles, and neuron loss.2

While there is argument about the “cause” of brain de-generation in AD, most agree that it is associated withan unusual form of inflammation that is accompaniedby oxidative damage that may result, as the brain isrich in lipids, in widespread lipid peroxidation (LP).3

Isoprostanes are chemically stable prostaglandin iso-mers formed by free radical peroxidation of polyunsat-urated fatty acids, which are stored in tissues, circulate

in plasma, and are excreted in urine.4 These propertiesmake them reliable quantitative in vivo LP biomarkers.We and others have shown that isoprostane levels incerebrospinal fluid (CSF) obtained postmortem fromthe lateral ventricle of AD patients are higher than incontrol subjects.5,6 However, it is known that thepathological changes in AD begin years before a clini-cal diagnosis can be made. It is during this preclinicalphase that therapies to slow, halt, or reverse the patho-logical process will be most effective. Thus, there is acompelling need to identify, in this phase, AD patientswith increased LP by using biomarkers that can be ac-quired noninvasively.

We undertook this study to address the questionwhether living patients with AD have enhanced iso-prostane generation in CSF, blood, and urine comparedwith matched healthy individuals, and whether their lev-els correlate with the severity of cognitive impairmentsand other biomarkers of AD brain degeneration.

Patients and MethodsPatientsThis study was reviewed and approved by the InstitutionalReview Board of the University of Pennsylvania. Subjectswere recruited from the university’s Alzheimer’s Disease Cen-ter Memory Disorders Clinic (MDC) between May 1998and September 1999. Informed consent was obtained fromall participants and their caregivers. The clinical diagnosis ofprobable or possible AD was based on the National Instituteof Neurological and Communicative Diseases and Stroke–Alzheimer’s Disease and Related Disorders Association crite-ria.7 As part of their routine cognitive assessment, all patientsreceived the Consortium to Establish a Registry for Alzhei-mer’s Disease psychometric battery, which assesses memory,language, and constructional praxis.8 The Dementia SeverityRating Scale (DSRS) and Mini-Mental State Examination(MMSE) assessments were performed to evaluate the clinicalseverity of the disease.9,10

Extensive laboratory and clinical examinations were per-formed, including magnetic resonance imaging, to excludeother causes of dementia. Subjects were excluded if they hadan acute infectious or inflammatory disease, hepatic chronicdisease, alcoholism, or cancer, or if they were treated withvitamins, since they all affect F2-isoprostane biosynthesis.4

Thirty-five consecutive patients were enrolled; urine andblood samples were obtained from all of them. Two weekslater, 14 patients and 10 control subjects agreed to have alumbar puncture for CSF and a second urine sample collec-tion. There were no significant clinical or demographic differ-ences between the patients who did or did not undergo lum-bar puncture. Age- and sex-matched control subjects wererecruited from a cohort of cognitively normal individuals fol-lowed by the Alzheimer’s Disease Center and from cognitivelynormal spouses of AD patients attending the MDC.

Isoprostane AnalysisSamples were analyzed by gas chromatography/mass spec-trometry, as previously described.4,11 The intraassay and in-

From the *Center for Experimental Therapeutics, †Department ofNeurology, ‡Center for Neurodegenerative Disease Research, andthe §Alzheimer’s Disease Center, University of Pennsylvania, Phila-delphia, PA; and \Claude Pepper Institute, and Department of Bio-chemistry, Florida Institute of Technology, Melbourne, FL.

Received Feb 7, 2000, and in revised form Jun 12. Accepted forpublication Jul 14, 2000.

Address correspondence to Dr Pratico, Center for Experimental Ther-apeutics, Dept of Pharmacology, University of Pennsylvania, BRB II/III, Room 812, 421, Curie Boulevard, Philadelphia, PA 19104.

Copyright © 2000 by the American Neurological Association 809

terassay variability for urine, plasma, and CSF, is 64% and65%, 65.5% and 64%, 65.0% and 64.5%, respectively.Blood, anticoagulated with EDTA, was centrifuged to obtainplasma; CSF, collected visually free of blood contamination,was sedimented at 1,500 rpm; both were stored at 280°C.All assays were performed without knowledge of the clinicaldiagnosis of the patient.

CSF Tau, CSF Ab1–40 , and Ab1–42

Tau protein levels were measured by sandwich ELISA usingthe Innotest hTAU-Antigen kit (Innogenetics, Belgium).12

Ab1–40 and Ab1–42 levels were measured by a previouslywell-characterized sandwich ELISA.13,14 The Ab and tausandwich ELISA have detection limits of less than 1 fento-mole of synthetic Ab and less than 75 pg/ml of tau per sam-ple. Both assays were performed in duplicate, and the varia-tion between samples in the duplicate assays was less than10% for each. They were performed without knowledge ofthe clinical diagnosis of the patient.

Apolipoprotein E GenotypeDNA was extracted from peripheral leukocytes and apolipopro-tein E (ApoE) genotyping was performed as previously de-scribed,12 without knowledge of the patient’s clinical diagnosis.

Statistical AnalysisComparisons among groups were performed by nonparametricone-way analysis of variance (Kruskall-Wallis test) with the useof Dunn’s posttest. Correlation was studied by linear regres-sion analysis. Statistical significance was set at p , 0.05.

ResultsTable 1 presents the characteristics of the AD patientsand controls studied here. ApoE ε2, ε3, and ε4 alleledistribution in the AD patients was as follows: ε2/ε3,n 5 3; ε3/ε3, n 5 8; ε4/ε3, n 5 12; ε4/ε4, n 5 12.No significant difference among groups was observedfor plasma cholesterol and triglyceride levels (data notshown). Patients with probable or possible AD had uri-nary 8,12-iso-iPF2a-VI levels significantly greater thancontrols (4.93 6 0.42, 4.18 6 0.56 vs 1.77 6 0.17ng/mg creatinine, respectively) (p , 0.0001, for both)(Fig, top). A similar pattern was observed in plasma(0.68 6 0.05, 0.67 6 0.08 vs 0.18 6 0.01 ng/ml,respectively; p , 0.0001) (see Fig center). Subse-quently, CSF was obtained from 10 patients withprobable AD, 4 with possible AD, and 10 normal con-trols (see Fig, bottom). At this time, a second urinesample was collected, and the urinary 8,12-iso-iPF2a-VI levels measured in these samples did not dif-fer significantly from the ones initially obtained fromthe same subjects (not shown). In both groups of ADpatients CSF 8,12-iso-iPF2a-VI levels were significantly(p , 0.0001) higher than in controls (see Fig, bottom;Table 2). A direct correlation was observed betweenCSF and urinary levels of 8,12-iso-iPF2a-VI and be-tween CSF and plasma 8,12-iso-iPF2a-VI. The coeffi-

cient of correlation (r2) for each was 0.55 and 0.64(both p , 0.001), respectively.

AD patients had impaired cognitive function asshown by the MMSE and DSRS assessments (see Ta-ble 1). A direct correlation was observed betweenDSRS and CSF 8,12-iso-iPF2a-VI levels (r2 5 0.22,p 5 0.02), whereas an inverse correlation was obtainedbetween MMSE scores and CSF 8,12-iso-iPF2a-VI lev-els (r2 5 20.15, p 5 0.04). CSF tau protein was ele-vated, while the percentage ratio between CSF Ab1–40

and Ab1–42 was lower in AD patients than in matchedcontrols (see Table 2). CSF tau protein and CSF 8,12-iso-iPF2a-VI levels were directly correlated (r2 5 0.43,p , 0.0001), whereas 8,12-iso-iPF2a-VI levels were in-versely correlated with the percentage of CSF Ab1–42

(r2 5 20.32, p 5 0.04). Finally, we investigated therelation between the number of apoE ε4 alleles and thelevels of 8,12-iso-iPF2a-VI in CSF. Subjects with twoε4 alleles had significantly higher CSF 8,12-iso-iPF2a-VI levels than those with one or no ε4 alleles(74 6 7.4 [p 5 0.04, versus both], 49 6 9, 50 6 5.7,pg/ml, respectively).

DiscussionIn this study, we show for the first time that comparedwith matched controls, patients with a clinical diagno-sis of AD have increased CSF, plasma, and urinary lev-els of 8,12-iso-iPF2a-VI, a reliable marker of in vivoLP.4 The finding that urinary and circulating levels ofthis specific isoprostane significantly correlate with thelevels in CSF of AD patients adds further credence tothe hypothesis that LP occurs early in the course of thisdementing disorder, thereby implicating it as contrib-utor to brain degeneration in AD. Significantly, it is

Table 1. Demographic and Clinical Characteristics ofAlzheimer’s Disease (AD) Patients and Controls

Probable AD(n 5 25)

Possible AD(n 5 10)

Controls(n 5 25)

Age (yr)Mean (SE) 76 (1.4) 75 (1.8) 74.5 (2)Range 58–97 68–90 57–94

F/M ratio 23/2 7/3 19/6% White 80 70 100Education

(% , 9 yr)2 1 0

Smokers 3 1 0MMSE

Mean (SE) 17 (1.4) 19 (2.2) 29 (0.3)Range 4–27 10–29 27–30

DSRSMean (SE) 18 (1.7) 16 (1) NTRange 5–39 10–20

Results are expressed as mean, standard error (SE), and range.

MMSE 5 Mini-Mental State Examination; DSRS 5 Dementia Se-verity Rating Scale; NT 5 not tested.

810 Annals of Neurology Vol 48 No 5 November 2000

plausible to infer that the assessment of 8,12-iso-iPF2a-VI in the urine may provide a convenient, non-invasive and reliable approach to study LP and oxida-tive damage in AD since both have been implicatedmechanistically in its pathogenesis.3,15

Another important finding we report here is the cor-relation between the levels of CSF 8,12-iso-iPF2a-VIand the severity of the dementia in AD patients asmeasured by the MMSE and DSRS scales. Further,since it has been demonstrated that CSF tau increases

and Ab1–42 decreases with progression of the disease inAD patients,16,17 it is notable that we observed a directcorrelation of CSF 8,12-iso-iPF2a-VI levels and CSFtau and an inverse correlation of these levels with CSFAb1–42 in AD patients. Thus, elevations in the level of8,12-iso-iPF2a-VI not only reflect an increase in centralnervous system oxidative stress but also correlate withthe progression of the disease.

The risk of development of late-onset AD has beenlinked to the polymorphism of the ApoE gene, in par-ticular the ε4 allele.18 We found that subjects homozy-gous for ε4 allele had higher CSF 8,12-iso-iPF2a-VIlevels than those with one or no ε4 alleles. Thus, thissuggests that apoE ε4 allele could influence the re-sponse of the brain to oxidant injury in AD.19

Because the neurological changes of AD are thoughtto begin insidiously well before a clinical diagnosis canbe made, it would be attractive to be able to identifyAD patients with increased LP by using reliable assaysto noninvasively measure biomarkers of LP. Althoughthe sample size in our study was small and follow-upstudies of larger cohorts are needed, we observed nooverlap between CSF and plasma 8,12-iso-iPF2a-VIlevels of AD patients and controls. This strongly sug-gests that quantification of this specific isoprostanemay be useful in early detection of AD. Finally, its cor-relation with clinical severity and other biomarkers ofthe disease could enable this marker to be exploited tomonitor the response to antioxidant therapies in AD.

This work was supported by grants from the American Federationfor Aging Research (P99143), the National Institutes of Health (HL61364, M01RR00040 and AG10124), and a Pilot Grant from theADC at the University of Pennsylvania. Dr Lee is the John M.

Fig. Urinary (top), plasma (center), and cerebrospinal fluid(CSF) (bottom) 8,12-iso-iPF2a-VI levels in patients withprobable (pro) or possible (pos) Alzheimer’s disease (AD) andage- and sex-matched controls (Con).

Table 2. Demographics, Cerebrospinal Fluid (CSF) 8,12-iso-iPF2a-VI, CSF Tau Protein, and CSF Ab1–42 PercentageLevels in Alzheimer’s Disease (AD) Patients and Controls

AD (n 5 14) Controls (n 5 10)

Age (yr)Mean (SE) 74 (1.3) 74 (2)Range 64–82 60–82

Ratio F/M 11/3 7/38,12-iso-iPF2a-VI

(pg/ml)Mean (SE) 66a (4.6) 25 (3.3)Range 43–105 6–38

CSF tau (pg/ml)Mean (SE) 770b (70) 320 (65)Range 300–1,500 170–460

CSF Ab1–42 (%)Mean (SE) 5.3b (0.7) 8.0 (0.8)Range 2.4–7.8 5.4–16.7

Results are expressed as mean, standard error (SE), and range.ap , 0.0001.bp , 0.01.

Brief Communication: Pratico et al: Lipid Peroxidation in Alzheimer’s Disease 811

Ware Professor of Alzheimer’s disease. Dr FitzGerald is the Robi-nette Foundation Professor of Cardiovascular Medicine.

We thank the patients and all their families who made this researchpossible.

References1. Clark CM. Clinical manifestation and diagnostic evaluation of

patients with Alzheimer’s disease. In: Clark CM, TrojanowskiJQ, eds. Neurodegenerative dementias: clinical features andpathological mechanisms. New York: McGraw-Hill, 2000:95–114

2. Martin JB. Molecular basis of the neurodegenerative disorders.N Engl J Med 1999;340:1970–1980

3. Markesbery WR, Carney JM. Oxidative alterations in Alzhei-mer’s disease. Brain Pathol 1999;9:133–146

4. Pratico D. F2-isoprostanes: sensitive and specific non-invasiveindices of lipid peroxidation in vivo. Atherosclerosis 1999;147:1–10

5. Pratico D, Lee VM-Y, Trojanowski JQ, Rokach J, FitzGeraldGA. Increased F2-isoprostanes in Alzheimer’s disease: evidenceof enhanced lipid peroxidation in vivo. FASEB J 1998;12:1777–1783

6. Montine TJ, Markesberry WR, Roberts LJ, Morrow JD. Cere-brospinal fluid F2-isoprostane levels are increased in Alzheimer’sdisease. Ann Neurol 1998;4:410–413

7. Radebaugh TS, Buckholz NS, Khachaturian ZS. Fishersymposium: strategies for the prevention of Alzheimer’s dis-ease—overview of research planning meeting III. Alzheimer’sDis Assoc Disord 1996;10(Suppl 1):15

8. Welsh KA, Butters N, Mohs RC, et al. The Consortium toEstablish a Registry for Alzheimer’s Disease (CERAD). Part V.A normative study of the neuropsychological battery. Neurol-ogy 1994;44:609–614

9. Clark CM, Ewbank DC. Performance of the dementia severityrating scale: a caregiver questionnaire for rating severity in Alz-heimer’s disease. Alzheimer’s Dis Assoc Disord 1996;10:31–39

10. Folstein MF, Folstein SE, McHugh PR. “Mini-mental state”: apractical method for grading the cognitive state of patients forthe clinician. J Psychiatr Res 1975;12:189–198

11. Li H, Lawson JA, Reilly M, et al. Quantitative high perfor-mance liquid chromatography/tandem mass spectrometry anal-ysis of the four classes of F2-isoprostanes in human urine. ProcNatl Acad Sci USA 1999;96:13381–13386

12. Arai H, Terajima M, Miura M, et al. Tau in cerebrospinalfluid: a potential diagnostic marker in Alzheimer’s disease. AnnNeurol 1995;38:649–652.

13. Turner RS, Suzuki N, Chyung AS, et al. Amyloids b40 and b42

are generated intracellularly in cultured human neurons andtheir secretion increases with maturation. J Biol Chem 1996;271:8966–8970

14. Gravina SA, Ho L, Eckman CB, et al. Amyloid b protein (Ab)in Alzheimer’s disease brain: biochemical and immunocyto-chemical analysis with antibodies specific for forms ending atAb40 or Ab42(43). J Biol Chem 1995;270:7013–7016

15. Roberts LJ, Montine TJ, Markesberry WR, et al. Formation ofisoprostane-like compounds (neuroprostanes) in vivo from do-cosahexaenoic acid. J Biol Chem 1998;273:13605–13612

16. Kanai M, Matsubara E, Isoe K, et al. Longitudinal study ofcerebrospinal fluid levels of Tau, Ab1–40, and Ab1–42(43) inAlzheimer’s disease: a study in Japan. Ann Neurol 1998;44:17–26

17. Nakamura T, Shoji M, Harigaya Y, et al. Amyloid b proteinlevels in cerebrospinal fluid are elevated in early-onset Alzhei-mer’s disease. Ann Neurol 1994;36:903–911

18. Corder EH, Saunders AM, Strittmatter WJ, et al. Gene dose of

apolipoprotein E type 4 and the risk of Alzheimer’s disease inlate onset families. Science 1993;261:921–923

19. Pratico D, Rokach J, Tangirala RT. Brains of aged apolipopro-tein E–deficient mice have increased levels of F2-isoprostanes,in vivo markers of lipid peroxidation. J Neurochem 1999;73:736–741

Intranuclear Inclusions inOculopharyngeal MuscularDystrophy Contain Poly(A)Binding Protein 2Mark W. Becher, MD,* Joyce A. Kotzuk, BS,*Larry E. Davis, MD,†\ and David G. Bear, PhD‡§

Intranuclear inclusions are one of the ultrastructural hall-marks of oculopharyngeal muscular dystrophy (OPMD),a disorder caused by small polyalanine (GCG) expansionsin the gene that codes for a ubiquitous nuclear proteincalled poly(A) binding protein 2 (PABP2). We studiedOPMD skeletal muscle and found that 1.0 to 10.0% ofmyocyte nuclei contained discreet PABP2 immunoreac-tive intranuclear inclusions, providing the first direct ev-idence of the relation between the proposed gene forOPMD and the pathology of OPMD.

Becher MW, Kotzuk JA, Davis LE, Bear DG.Intranuclear inclusions in oculopharyngeal

muscular dystrophy contain poly(A) bindingprotein 2. Ann Neurol 2000;48:812–815

Oculopharyngeal muscular dystrophy (OPMD) is anautosomal dominant myopathy that presents in thefourth to sixth decade and is characterized by ptosis,dysphagia, and dysphonia often followed by limitationof eye movements, proximal limb weakness, and facialweakness1 (reviewed in Proceedings of the First In-ternational Symposium on OPMD2). Pathologically,skeletal muscle from OPMD patients is not mark-edly abnormal, although occasional myofibers contain“rimmed vacuoles” in the sarcoplasm that appear to be

From the Departments of *Pathology, †Neurology, Neuroscience,and Microbiology, and ‡Cell Biology and Physiology, and §CancerCenter, University of New Mexico, Health Sciences Center, andiNew Mexico VA Health Care System, Albuquerque, NM.

Received May 15, 2000, and in revised form Jul 17. Accepted forpublication Jul 21, 2000.

Address correspondence to Dr Becher, Neuropathology Laboratory,University of New Mexico HSC, CRTC-B93, 915 Camino de Sa-lud NE, Albuquerque, NM 87131.

812 Copyright © 2000 by the American Neurological Association

autophagocytic structures.3 Many years ago, it wasfound by ultrastructural studies that some OPMD skel-etal muscle nuclei contain intranuclear inclusions ofunknown significance.4 These inclusions are reportedlyfound in less than 10% of myocyte nuclei from af-fected individuals and, at the ultrastructural level, arecomposed of collections of tubulofilamentous struc-tures (8.5 nm external diameter and 3.0 nm internaldiameter) that displace the chromatin to the peripheryof the nucleus.4 The nature of these tubulofilaments isunknown, although some inclusions have been shownto exhibit immunoreactivity to ubiquitin antibodies.5

OPMD was linked to chromosome 14q11.2-q13 in1995.6 Three years later, Brais and colleagues foundthat a polyalanine (GCG)6 repeat in the N-terminus ofthe poly(A) binding protein 2 (PABP2) gene in thisregion was expanded to (GCG)8–13 in a large series ofOPMD pedigrees.7 PABP2 is a ubiquitous 33-kd nu-clear protein with an acidic region near the amino ter-minus, a basic carboxy terminus, and a single RNA-binding domain near the center.8

Thus, OPMD is a member of the growing numberof disorders that have a very restricted neurologicalphenotype and are caused by expansion of a trinucle-otide repeat sequence.9 Nearly all of the previouslycharacterized triplet repeat disorders have been shownto develop intranuclear proteinaceous aggregates com-posed, at least in part, of the mutant gene product10,11

(reviewed in Kaytor and Warren12), which are a novelmorphological abnormality previously unheard of inany known cellular mechanism of disease. The signifi-cance of these intranuclear aggregates remains un-clear.13 It is not known whether they are detrimentalto cellular function, are protective in some manner, orrepresent a reactive process of sequestering abnormalproteins in an intranuclear compartment, perhaps aspart of a dynamic process of protein degradation.14 Inthis study, we sought to determine whether nuclei ofskeletal muscle from patients with OPMD containedthe proposed gene product, PABP2, in an aggregateformation.

Subjects and MethodsSubjectsArchival skeletal muscle biopsy tissues from 4 unrelated adultpatients with the clinical manifestations of OPMD from theRio Grande Valley region of New Mexico were studied(UNM Human Research Review Committee approval #00-228). Samples were obtained from the deltoid or lower ex-tremity muscles. Control skeletal muscle was obtained fromarchival surgical pathology adult cases without neurologicalsymptoms.

Tissue Preparation and PABP2 ImmunocytochemistryFresh skeletal muscle was fixed in 10% neutral-buffered for-malin at the time of the original biopsy and embedded in

paraffin. Immunocytochemistry was performed using a rabbitpolyclonal antibody to PABP2 (gift of Dr Elmar Wahle,University of Halle, Germany)15 that was further purified ona protein A sepharose column (Pharmacia, Piscataway, NJ)and standard immunocytochemical methods with microwaveantigen retrieval. Control samples were identically preparedand sectioned. Briefly, 4-mm deparaffinized sections wereblocked with 3% H2O2 in methanol for 15 minutes, micro-waved for 5 minutes in citrate buffer at pH 6.0, subjected toprimary antibody at 1:1,000 in Tris-buffered saline (TBS)with 0.5% normal goat serum overnight at room tempera-ture, and labeled with avidin-biotin complex reagents (Vec-tor, Burlingame, CA) and 3,39-diaminobenzidine chroma-gen. Sections treated with TBS or pooled rabbit IgG in placeof the primary antibody served as technique controls. Afterthe chromagen reaction, all sections were briefly stained inhematoxylin to provide a nuclear counterstain. Hematoxylinand eosin (H&E)–stained sections were present on each caseto review the diagnostic light microscopic findings.

ResultsNone of the 4 OPMD cases had markedly abnormalskeletal muscle on H&E stain. Rare atrophic myocytesand occasional rimmed vacuoles were present. Myocytenuclei had intact nuclear membranes by light micros-copy and lacked chromatin or intranuclear abnormali-ties. Samples from control subjects were normal. Im-munoreactivity to the PABP2 antibody was restrictedto the nucleus in all cases. Strongly immunoreac-tive round intranuclear inclusions were found in all4 OPMD cases (Fig) and none of the control sub-jects. Inclusion-bearing myocyte nuclei ranged from1.0–10.0% (Table). Most nuclei with inclusions con-tained a single aggregate, although up to three distinctinclusions were identified in single nuclei (see Fig, B).Inclusions were approximately 2 to 3 mm in greatestdimension and myocyte nuclei were 8 to 10 mm indiameter. Given that 4-mm paraffin sections were stud-ied, we evaluated approximately 40% of each nucleus.Therefore, it could be calculated that the actual num-ber of inclusion-bearing myocyte nuclei could be up to2.5 times higher than observed (ie, up to 25%). Noinclusions were identified in control subject myocytenuclei. Fine particulate and diffuse nuclear stainingwith PABP2 was identified in many myocyte nuclei inboth OPMD cases and control subjects. Techniquecontrol slides (buffer or rabbit IgG as primary anti-body) were blank (see Fig). Intranuclear inclusions inOPMD cases were distinct from fine particulate or dif-fuse nuclear staining and there was no predilection to-ward inclusions being present in nuclei with or withoutadditional PABP2 immunoreactivity. The diffuse nu-clear immunoreactivity in control subjects is consistentwith what has been shown by others in mammaliannuclei.15

Brief Communication: Becher et al: Intranuclear Inclusions in OPMD 813

DiscussionThese studies show that the intranuclear inclusions ofthe first polyalanine repeat disorder to be identified,OPMD, which were previously best seen by electronmicroscopy, do indeed contain PABP2, the predictedgene product that harbors a triplet repeat expansion inaffected individuals. It is possible that PABP2 antibod-ies will have diagnostic utility for the evaluation ofmuscle biopsies at the light microscopic level from sus-pected OPMD patients in the future.

With the exception of the androgen receptor of spi-nobulbar muscular atrophy16 and the a subunit of aPurkinje cell–specific calcium channel in spinocerebel-lar ataxia type 6,17 the protein products of the majorityof the triplet repeat disorder genes are of unknownfunction. PABP2 is a well-characterized nuclear protein

that binds to newly created poly(A) tails after at least10 adenine residues are in place, and it is thought thatadditional PABP2 molecules assemble as the poly(A)tail elongates.18 PABP2 has two known regulatoryfunctions in polyadenylation. In conjunction withcleavage and polyadenylation specificity factor (CPSF),PABP2 enhances the processivity and rate of poly(A)tail elongation by poly(A) polymerase. PABP2 is alsorequired to terminate the synthesis of the poly(A) tailafter the polymerization of 250 to 300 adenine resi-dues.18 Recently, Keller and colleagues19 have shownthat PABP2-poly(A) complexes can form 7-nm linearfilaments, as well as 21-nm oligomeric particles thatincluded 200 to 300 adenine residues. Thus, the nor-mal aggregation of PABP2 monomers may play a rolein both elongation and termination of poly(A) synthe-sis. The PABP2 filament diameter is similar to the di-mensions of the filaments aggregated within OPMDintranuclear inclusion bodies.4 PABP2 is found pre-dominately in the nucleus, although it may shuttle be-tween the nucleus and the cytoplasm.15,20 PABP2 alsolikely has a role in protein trafficking within the nu-cleus or to nuclear pores.15 These functions may becritical to our understanding of the pathogenesis of dis-ease in OPMD.

Compared with other triplet repeat disorders inwhich pathogenic expansions are in the tens to hun-dreds of codons,9 the expansion in OPMD is remark-ably small. How the small polyadenine expansion inPABP2 causes disease is unknown. Perhaps the intrin-sic aggregation properties of PABP2 are altered enoughto significantly increase its normal tendency to formoligomeric complexes.

The mechanism by which trinucleotide repeat ex-pansion mutations in PABP2 lead to pathogenesis inOPMD is unknown. In addition, the relation betweenprotein aggregation and the etiology of the triplet re-peat disorders remains an enigma. Secondary pathologyin these disorders needs to be examined. For example,structural and genomic mitochondrial abnormalities

Fig. Poly(A) binding protein 2 intranuclear inclusions in nu-clei of oculopharyngeal muscular dystrophy myocytes. Distinctintranuclear inclusions (arrows in A and C), as well as par-ticulate nuclear immunoreactivity, are identified with antibod-ies to PABP2. Most inclusion-bearing nuclei have a singleaggregate, although occasional nuclei contain two (B) or threeinclusions. PABP2 inclusions are slightly larger than nucleoli(arrowhead in B) and affect both peripheral differentiatedmyocyte nuclei (A and B) and myocytes in clusters (C). Neuro-logically normal control subject myocyte nuclei lack inclusions,although, as expected, many nuclei have similar appearingdiffuse, finely particulate PABP2 immunoreactivity (D). Nu-clei in technique control sections (rabbit IgG in place of pri-mary antibody) are negative (E). Bar 5 10 mm. (Polyclonalantibody to PABP2 with light hematoxylin counterstain.)

Table. PABP2 Staining in Myocyte Nuclei fromOculopharyngeal Muscular Dystrophy Cases and Controls

Case No.

Total No.of NucleiCounted

Nuclei withDiffuse and/orParticulateStaining

IntranuclearInclusions

OPMD 1 1,065 165 (15.5%) 24 (2.3%)OPMD 2 1,041 337 (32.4%) 36 (3.5%)OPMD 3 1,082 744 (68.8%) 11 (1.0%)OPMD 4 1,084 129 (11.9%) 108 (10.0%)CTL 1 1,055 416 (39.4%) 0CTL 2 1,059 18 (1.7%) 0CTL 3 1,092 642 (58.8%) 0CTL 4 1,100 818 (74.4%) 0

814 Annals of Neurology Vol 48 No 5 November 2000

have been identified in OPMD, yet their significanceand relationship to cellular dysfunction, age-relatedchanges, or mutant gene products are unknown.21 Sev-eral hypotheses remain to be explored, such as thefactors that impart cell specificity, proteins associatedwith gene product proteins, mitochondrial dysfunction,ubiquitin/proteasome/molecular chaperone abnormali-ties, the mechanism of protein aggregation, and the re-cently developed concept of perturbations in “proteinclearance.”22 Given the large body of knowledge aboutthe protein product of the OPMD gene and the evi-dence contained in this communication linking theproposed gene product to the pathology of OPMD forthe first time, we expect that the study of OPMD andPABP2 will lead to a better understanding of themechanism of intranuclear protein aggregation in thisand other triplet repeat disorders in the near future.

We thank Elmar Wahle for the generous gift of PABP2 antiserum.This work was supported in part by grants from NIH NINDS(M.W.B.), National Science Foundation (D.G.B.), and MuscularDystrophy Association (D.G.B.).

References1. Brais B, Rouleau GA, Bouchard J-P, et al. Oculopharyngeal

muscular dystrophy. Semin Neurol 1999;19:59–662. Bouchard J-P, Brais B, Tome FM, eds. Proceedings of the first

international symposium on oculopharyngeal muscular dystro-phy. Neuromusc Disord 1997;7:S1–S106

3. Tome FMS, Fardeau M. Oculopharyngeal Muscular Dystro-phy. In: Engel AG, Franzini-Armstrong C, eds. Myology, 2nded. New York: McGraw-Hill, 1994:1233–1245

4. Tome FMS, Fardeau M. Nuclear inclusions in oculopharyngealdystrophy. Acta Neuropathol 1980;49:85–87

5. Askanas V, Serdaroglu P, Engel WK, Alvarez RB. Immunolo-calization of ubiquitin in muscle biopsies of patients with in-clusion body myositis and oculopharyngeal muscular dystrophy.Neurosci Lett 1991;130:73–76

6. Brais B, Xie YG, Sanson M, et al. The oculopharyngeal mus-cular dystrophy locus maps to the region of the cardiac alphaand beta myosin heavy chain genes on chromosome 14q11.2-q13. Hum Mol Genet 1995;4:429–434

7. Brais B, Bouchard JP, Xie YG, et al. Short GCG expansions in

the PABP2 gene cause oculopharyngeal muscular dystrophy.Nat Genet 1998;18:164–167

8. Nemeth A, Krause S, Blank D, et al. Isolation of genomic andcDNA clones encoding bovine poly(A) binding protein II. Nu-cleic Acids Res 1995;23:4034–4041

9. La Spada AR, Paulson HL, Fischbeck KH. Trinucleotide repeatexpansion in neurological disease. Ann Neurol 1994;36:814–822

10. DiFiglia M, Sapp E, Chase KO, et al. Aggregation of Hunting-tin in neuronal intranuclear inclusions and dystrophic neuritesin brain. Science 1997;277:1990–1993

11. Becher MW, Kotzuk JA, Sharp AH, et al. Intranuclear neuro-nal inclusions in Huntington’s disease and dentatorubral andpallidoluysian atrophy: correlation between the density of inclu-sions and IT15 CAG triplet repeat length. Neurobiol Dis 1998;4:387–397

12. Kaytor MD, Warren ST. Aberrant protein deposition and neu-rological disease. J Biol Chem 1999;274:37507–37510

13. Sisodia SS. Nuclear inclusions in glutamine repeat disorders: arethey pernicious, coincidental, or beneficial? Cell 1998;95:1–4

14. Saudou F, Finkbeiner S, Devys D, Greenberg ME. Huntingtinacts in the nucleus to induce apoptosis but death does not cor-relate with the formation of intranuclear inclusions. Cell 1998;95:55–66

15. Krause S, Fakan S, Weis K, Wahle E. Immunodetection ofpoly(A) binding protein II in the cell nucleus. Exp Cell Res1994;214:75–82

16. La Spada AR, Wilson EM, Lubahn DB, et al. Androgen recep-tor gene mutations in X-linked spinal and bulbar muscular at-rophy. Nature 1991;352:77–79

17. Zhuchenko O, Bailey J, Bonnen P, et al. Autosomal dominantcerebellar ataxia (SCA6) associated with small polyglutamine ex-pansions in the alpha 1A-voltage-dependent calcium channel.Nat Genet 1997;15:62–69

18. Wahle E, Ruegsegger U. 39-End processing of pre-mRNA ineukaryotes. FEMS Microbiol Rev 1999;23:277–295

19. Keller RW, Kuhn U, Aragon M, et al. The nuclear poly(A)binding protein, PABP2, forms an oligomeric particle coveringthe length of the poly(A) tail. J Mol Biol 2000;297:569–583

20. Chen Z, Li Y, Krug RM. Influenza A virus NS1 protein targetspoly(A)-binding protein II of the cellular 39-end processing ma-chinery. EMBO 1999;18:2273–2283

21. Lezza AM, Cormio A, Gerardi P, et al. Mitochondrial DNAdeletions in oculopharyngeal muscular dystrophy. FEBS Lett1997;418:167–170

22. Orr HT, Zoghbi HY. Reversing neurodegeneration: a promiseunfolds. Cell 2000;101:1–4

Brief Communication: Becher et al: Intranuclear Inclusions in OPMD 815