Similarly, some participants without psychotic behaviorswill develop delusions and hallucinations in the future.Confirmation in a longitudinal sample would be helpful.

In conclusion, our findings suggest that psychoticsymptoms in AD may be caused by genetic alterationsin cytokine signaling, specifically involving IL-1�, thusconfirming earlier work in schizophrenia. Further re-search into the biological pathways leading to symptomgeneration, particularly with reference to cytokine-neurotransmitter interactions, is needed.

References1. Sweet RA, Nimgaonkar VL, Devlin B, Jeste DV. Psychotic

symptoms in Alzheimer disease: evidence for a distinct pheno-type. Mol Psychiatry 2003;8:383–392.

2. Williams J, Spurlock G, Holmans P, et al. A meta-analysis andtransmission disequilibrium study of association between thedopamine D3 receptor gene and schizophrenia. Mol Psychiatry1998;3:141–149.

3. Gaughran F. Immunity and schizophrenia: autoimmunity, cy-tokines, and immune responses. Int Rev Neurobiol 2002;52:275–302.

4. Tarkowski E, Liljeroth AM, Minthon L, et al. Cerebral patternof pro- and anti-inflammatory cytokines in dementias. BrainRes Bull 2003;61:255–260.

5. Katila H, Hanninen K, Hurme M. Polymorphisms of theinterleukin-1 gene complex in schizophrenia. Mol Psychiatry1999;4:179–181.

6. Zanardini R, Bocchio-Chiavetto L, Scassellati C, et al. Associ-ation between IL-1beta -511C/T and IL-1RA (86bp)n repeatspolymorphisms and schizophrenia. J Psychiatr Res 2003;37:457–462.

7. Ma SL, Tang NL, Lam LC, Chiu HF. Lack of association ofthe interleukin-1beta gene polymorphism with Alzheimer’s dis-ease in a Chinese population. Dement Geriatr Cogn Disord2003;16:265–268.

8. Ehl C, Kolsch H, Ptok U, et al. Association of an interleukin-1beta gene polymorphism at position �511 with Alzheimer’sdisease. Int J Mol Med 2003;11:235–238.

9. McKhann G, Drachman D, Folstein M, et al. Clinical diagno-sis of Alzheimer’s disease: report of the NINCDS-ADRDAWork Group under the auspices of Department of Health andHuman Services Task Force on Alzheimer’s Disease. Neurology1984;34:939–944.

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. Reisberg B. Functional assessment staging (FAST). Psychophar-macol Bull 1988;24:653–659.

12. Kaufer DI, Cummings JL, Christine D, et al. Assessing the im-pact of neuropsychiatric symptoms in Alzheimer’s disease: theNeuropsychiatric Inventory Caregiver Distress Scale. J AmGeriatr Soc 1998;46:210–215.

13. Nishimura M, Mizuta I, Mizuta E, et al. Influence ofinterleukin-1beta gene polymorphisms on age-at-onset of spo-radic Parkinson’s disease. Neurosci Lett 2000;284:73–76.

14. Araujo DM, Cotman CV, Differential effects of interleukin-1beta and interleukin-2 on glia and hippocampal neurons in cul-ture. Int J Dev Neurosci 1995;13:201–212.

15. Wichers M, Maes M. The psychoneuroimmuno-patho-physiology of cytokine-induced depression in humans. IntJ Neuropsychopharmacol 2002;5:375–388.

16. El-Omar EM, Carrington M, Chow WH, et al. The role ofinterleukin-1 polymorphisms in the pathogenesis of gastric can-cer. Nature 2001;412:99.

17. Hassanain M, Zalcman S, Bhatt S, Siegel A. Interleukin-1 betain the hypothalamus potentiates feline defensive rage: role ofserotonin-2 receptors. Neuroscience 2003;120:227–233.

18. Zakzanis KK, Andrikopoulos J, Young DA, et al. Neuropsycho-logical differentiation of late-onset schizophrenia and dementiaof the Alzheimer’s type. Appl Neuropsychol 2003;10:105–114.

19. Harrison PJ. The neuropathology of schizophrenia. A criticalreview of the data and their interpretation. Brain 1999;122:593–624.

20. Schurhoff F, Krebs MO, Szoke A, et al. Apolipoprotein E inschizophrenia: a French association study and meta-analysis.Am J Med Genet 2003;119B:18–23.

Sequence-Dependent andIndependent InhibitionSpecific for Mutant Ataxin-3by Small Interfering RNAYi Li, MD,1 Takanori Yokota, MD,1

Ryusuke Matsumura, MD,2 Kazunari Taira, PhD,3

and Hidehiro Mizusawa, MD1

In Machado–Joseph disease (MJD) gene, there is a C/Gpolymorphism immediately after the CAG repeat; the ex-panded CAG repeat tract is exclusively followed by C,whereas about half of wild-type alleles are followed by G.Using this C/G polymorphism, we have engineered thesmall interfering RNA (siRNA) which decreased the ex-pression of mutant ataxin-3, Q79C, by 96.0%, whereasthere was minimal reduction on that of the wild type,Q22G (5.9%). Furthermore, unexpectedly, the expressionof another wild-type allele, Q22C, was also much lesssuppressed (22.5%) by this siRNA possibly due to differ-ence of the secondary structure of the target RNA. This isthe first report of sequence-independent discriminationof mutant and wild-type alleles by siRNA.

Ann Neurol 2004;56:124–129

From the 1Department of Neurology and Neurological Science,Graduate School, Tokyo Medical and Dental University, Tokyo;2Department of Neurology, Nara Medical University, Nara; 3De-partment of Chemistry and Biotechnology, School of Engineering,University of Tokyo, Tokyo, Japan.

Received Nov 24, 2003, and in revised form Feb 25, 2004. Ac-cepted for publication Apr 2, 2004.

Published online Jun 28, 2004, in Wiley InterScience(www.interscience.wiley.com). DOI: 10.1002/ana.20141

Address correspondence to Dr Yokota, Department of Neurology,Tokyo Medical and Dental University, 1-5-45 Yushima, Bunkyo-ku, Tokyo 113-8519, Japan. E-mail: tak-yokota.nuro@tmd.ac.jp

124 © 2004 American Neurological AssociationPublished by Wiley-Liss, Inc., through Wiley Subscription Services

Machado–Joseph disease (MJD) is an autosomal dom-inant neurodegenerative disorder that is characterizedclinically by cerebellar ataxia, pyramidal and extrapyra-midal signs, peripheral neuropathy, and ophthalmople-gia. The number of CAG in MJD1 gene repeats is be-tween 13 and 36, whereas in patients this range isexpanded from 62 to 84.1 The pathogenosis of MJD isconsidered to be caused by “gain of toxic function” ofmutant MJD protein.2,3 Therefore, a most effectiveand simple gene therapeutic approach for MJD re-quires the reduction of the aberrant mutant protein.Furthermore, it might be needed to reduce mutantataxin-3 selectively, leaving wild-type protein intact,because the wild-type MJD1 gene product ataxin-3should have an important role in cell survival, such asquality control of endoplasmic reticulum4 and DNArepair.5

RNA interference (RNAi) is the process of sequence-specific, posttranscriptional gene silencing, initiated bydouble-stranded RNA (dsRNA). This has a multistepprocess that involves the generation of 21 to 23-nucleotide small interfering RNA (siRNA), resulting indegradation of the homologous RNA. The siRNA islong enough to mediate gene-specific suppression butshort enough to evade adverse effects of long dsRNA inmammalian cells6 and is expected to be a powerful toolfor gene therapy of human diseases.

We found that the CAG repeat tract in the MJD1gene is followed by C or G and there is extreme bias ofthis C/G polymorphism between mutant and normalMJD1 alleles; mutant alleles have exclusively the(CAG)nC, whereas normal alleles have both (CAG)nGand (CAG)nC in a similar frequency7,8 (Fig 1). Here,we engineered siRNAs to cleave selectively mutantMJD RNA targeting the sequence including this C/Gpolymorphism.

Materials and MethodsPlasmid Constructs and Small Interfering RNADesignThe expression plasmids of ataxin-3 were kindly given by DrAkira Kakizuka.4 The MJD1 cDNA in the plasmid was atruncated fragment including either 22 (normal, pCMX HA-Q22C) or 79 (expanded, pCMX HA-Q79C) repeats ofCAG, and was hemagglutinin (HA)-fused on the N termi-nus. Cytosine after CAG repeat in pCMX HA-Q22C waschanged to guanine (pCMX HA-Q22G) using theQuikChange site-directed mutagenesis system (Stratagene, LaJolla, CA). Furthermore, these MJD cDNAs also were sub-cloned into pIRES-hrGFP-1a (pGFP-Q22G, pGFP-Q22Cand pGFP-Q79C; Stratagene).

Four different siRNAs were designed to target the mutantMJD RNA at the junction of CAG repeat and 3� terminalregion (see Fig 1B). A single nucleotide alternation at C/Gpolymorphism was placed in siRNA at 5th, 8th, 11th, or14th position from the 5� end of the sense strand.

Transfections and Western BlottingTo see the effect of siRNAs for MJD RNA on the expressionof ataxin-3, we cotransfected both siRNA and ataxin-3-expression plasmids to human embryonic kidney cell line

Fig 1. Design of small interfering RNA (siRNA) sequencestargeting the C/G polymorphism in the Machado–Joseph dis-ease (MJD)–1 gene (A) Schema of human MJD1 gene show-ing C/G polymorphism just downstream from CAG repeat (toppanel). The distribution of C/G polymorphism and the CAGrepeat length in 56 Japanese MJD patients (bottom panel).All expanded alleles have (CAG)nC, whereas the normal al-leles have (CAG)nC and (CAG)nG in similar frequencies(0.46 and 0.54). Arrows indicate the CAG repeat numbers ofataxin-3 expression vector constructs used in this study as wildtype and mutant. (B) Sequences of siRNAs used to target(CAG)nC in MJD1 mRNA. Bold characters indicate the siteof C/G polymorphism. Uppercase letters indicate deoxyribonu-cleotides.

Li et al: siRNA for MJD 125

293T cells (293T) cells. siRNAs were annealed and trans-fected according to our previously reported method.9 Four-microgram expression vectors and 10 or 25nM siRNA werecotransfected to 293T cells in six-well culture plates by Lipo-fectamine Plus reagent (Life Technologies, Rockville, MD).

Twenty-four hours after transfection, cells were harvestedby TNG buffer (50mM Tris-HCl, 150mM NaCl, 1% Tri-ton X-100) with protease inhibitor cocktail (Roche, Mann-heim, Germany), separated on 15% sodium dodecyl sulfatepolyacrylamide gels, and transfered onto polyvinylidene di-fluoride membrane (Bio-Rad, Richmond, CA). Rat mono-clonal anti–HA antibody (Roche) was reacted and detectedusing an enhanced chemiluminescence detection kit (Amer-ham Pharmacia Biotech, Arlington Heights, IL).

Assessment of Cell DeathTo assess the effect of siRNA on the cell death caused byexpression of mutant ataxin-3, we cotransfected Neuro2acells in 24-well culture plates with pCMX-Q79C (1�g/well)and siRNA (100nM/well). At 48 hours after transfection, thecell death of Neuro2a cells was determined by assaying withboth trypan blue exclusion method and measurement of cy-toplasmic lactate dehydrogenase (LDH) activity with the Cy-totox 96 nonradioactive cytotoxicity assay (Promega, Madi-son, WI).

Statistical analysis were performed using the Student’s ttest and single-factor analysis of variance followed by Fisher’sprotected least-significant difference post hoc test.

ResultsThe mutant ataxin-3, Q79C, was most effectively sup-pressed by 25nM siRNA MJD3 and was decreased by96.0% on signal intensity of Western blotting com-pared with that in control (Fig 2A). In contrast, theexpression of the wild-type ataxin-3 Q22G was mod-erately suppressed by siRNA MJD4, but not clearly in-fluenced by siRNA MJD1, -2, or -3. Therefore, siRNAMJD3 best recognized one nucleotide alternation be-tween C and G and suppressed specifically expressionof Q79C, with almost no effect on that of Q22G. Ef-fect of siRNA MJD3 on Q79C expression was dosedependent, but suppression of Q22G was similar at 10and 25nM concentrations of siRNA (see Fig 2B).

Unexpectedly, siRNA MJD3 did not significantlydecrease (22.5%) the expression of another wild-typeallele, Q22C, although the target sequence in(CAG)79C and (CAG)22C were the same (see Fig2B). These suppression effects were confirmed by thereduction in fluorescence of GFP which was bicistroni-cally expressed by internal ribosomal entry site (IRES)with ataxin-3 (pGFP-Q22C or pGFP-Q79C) usingDsRed fluorescence as a control for transfection effi-ciency (see Fig 2C).

Next, by applying siRNA MJD3, we attempted todecrease the cell toxicity induced by mutant ataxin-3.Forty-eight hours after transfection, expression of themutant MJD (Q79C) was toxic to Neuro2a cells, butthose of the wild-type ataxin-3 (Q22G and Q22C) did

not influence cell survival. siRNA MJD3 could mark-edly suppress the toxicity of mutant ataxin-3, decreas-ing cell death by 62.8% in LDH assay and by 75.9%in trypan blue exclusion method (Fig 3).

DiscussionThis is the report of the siRNA which can discriminatemutant and wild-type type RNAs of MJD1 gene: thesiRNA MJD3 is highly effective for suppressing mu-tant ataxin-3, Q79C, with a negligible effect on onewild-type ataxin-3, Q22G, observed in half of JapaneseMJD patients, and with much less effect on the otherwild-type allele, Q22C, in another half patients. ThesiRNA has been reported to suppress expression of themutant androgen receptor RNA including expandedCAG repeat and rescue polyglutamine-mediated cyto-toxicity, but it cleaved wild-type RNA with normalCAG repeat in a similar way.10 During preparation ofthis article, Miller and his colleagues11 reported the ef-fective siRNA targeting the C/G polymorphism inMJD1 gene, which suppressed the Q166C expressionwith only modest effect on Q28G expression, but theydid not examine another wild-type allele Q28C.

Diversing effects of mismatch between siRNA andtarget sequences have been reported. The central singlemutation, mismatch in siRNA at 10th and 11th posi-tion, produces marked decrease of cleavage siRNA ac-tivity.13 The mismatch in the 5�end has negligible ef-fect on siRNA cleavage activity compared with morecentrally located mutation. This bias might be linkedto the proposed existence of a “ruler” region in the

Fig 2. Specific effect of the small interfering RNA (siRNA) onthe expression of mutant MJD1. (A) Selection of best siRNAto discriminate C/G polymorphism. Western blot analysis ofbest siRNA for targeting mutant ataxin-3 among siRNAMJD1-4. 293T cells were cotransfected with ataxin-3 expres-sion plasmids (pCMX HA-Q79C and pCMX HA-Q22G) andsiRNAs. Tubulin loading controls also are shown. Right panelindicates quantitation of signal intensities. Each percentagesuppression was determined by the band intensity with trans-fection of siRNA MJD3 shuffle as a control. siRNA MJD3suppressed most Q79C expression, and siRNA MJD1-3 didnot influence the Q22G expression. Values are the mean andSEM. (B) Comparison of the effect of siRNA MJD3 onQ79C, Q22G, and Q22C expression. Western blot analysis ofthe effect of siRNA MJD3 at 10 and 25nM on Q79C,Q22G, and Q22C expression. The results of siRNA(�) weremade with siRNA MJD3 shuffle. Right panel indicates quan-titation of signal intensities. Almost no suppression on Q22Gexpression (5.9%) and mild suppression on Q22C expression(22.5%) were noted by siRNA MJD3 at 25nM siRNA incontrast with robust suppression on Q79C expression (96.0%).Bottom panel shows the GFP images of 293T cells cotrans-fected with pGFP-Q79C/pGFP-Q22C/pGFP-Q22G andsiRNA MJD3. pDsRed also was transfected as a control oftransfection efficiency (�200).

‹

126 Annals of Neurology Vol 56 No 1 July 2004

Figure 2

-

Li et al: siRNA for MJD 127

siRNA that is used by the RNA-induced silencingcomplex (RISC) complex to “measure” the target RNAfor cleavage .13 In four siRNAs examined in our study,siRNAs (siRNA MJD2 and -3) with central mismatchhave better discrimination than that with outside mis-match (siRNA MJD1 and -4), and this result is con-sistent with the previously reported results.

Surprisingly, there was much difference between theeffect of siRNA on Q22C and Q79C, although theirtarget sequences of siRNAs were the same. This is thefirst report to our knowledge of sequence-independentdiscrimination of mutant and wild-type alleles bysiRNA. One possible reason for this difference is thatnot all RNA sequences are equally accessible tosiRNAs: some sequences might be buried within thesecondary structure of target RNAs especially whenthey are highly folded.15 We also experienced that thebest target site of siRNA for the highly folded RNAwas almost same as that for ribozyme, which cleavageefficiency is much influenced by the secondary struc-ture of target RNA.9 The target site of C/G polymor-phism is just downstream from the CAG repeat, whichrepresents a tight stem form in their secondary struc-ture on a computer prediction. Although we have no

data indicating that siRNA MJD3 is more accessible to(CAG)79C than (CAG)22C, a change of secondarystructure of the MJD RNA due to a large difference ofthe CAG repeat length might affect the efficiency ofsiRNA MJD3. Another possible explanation is thatthere is a RNA-binding protein preferentially binding(CAG)22C over (CAG)79C, which interferes the ac-cess of siRNA MJD3 to (CAG)22C RNA. The RNA-binding protein to MJD RNA has not been found, butCUG repeat in myotonic dystrophy protein kinase(DMPK) RNA binds the proteins including CUG-binding protein in repeat length-dependent manner.15

Moreover, existence of a RNA-binding protein attach-ing CAG repeat was suggested in Huntington’s dis-ease.16 Further studies are needed to make clear themechanism for the difference of the siRNA effect on(CAG)79C and (CAG)22C.

We are in the process of making adeno-associatedvirus and transgenic mice expressing siRNA MJD3 toinvestigate the efficacy of siRNAs in vivo. Although aless invasive delivery method for introducing siRNAsinto a larger area of cerebellum would be needed forclinical feasibility, the efficiency of our siRNA to spe-cifically reduce the expression and toxicity of mutant

Fig 3. Effect of small interfering RNA (siRNA) MJD3 on the mutant ataxin-3–induced toxicity in mammalian cells. siRNA MJD3rescues the cell death induced by overexpression of Q79C, Q22C, and Q22G on lactate dehydrogenase (LDH) assay and trypanblue exclusion method (***p � 0.0001, **p � 0.01). Overexpression of Q22G or Q22C did not show toxicity. pcDNA3.1 plas-mid (Invitrogen, La Jolla, CA) was used as mock. Values are the mean and SEM. The result of siRNA(�) was made with siRNAMJD3 shuffle.

128 Annals of Neurology Vol 56 No 1 July 2004

ataxin-3 in cells suggests that this mRNA-targeting ap-proach by siRNA might provide effective therapy forMJD.

This work was supported by grants from the Ministry of Education,Science and Culture of Japan (14570582; T.Y.) and from the Min-istry of Health, Labor and Welfare of Japan (H14-KOKORO-018;T.Y.), and 21st Century Center of Excellence (COE) Program Fel-lowship (L.Y.).

We thank Dr A. Kakizuka for providing expression vectors of MJDgene.

References1. Kawaguchi Y, Okamoto T, Taniwaki M, et al. CAG expansions

in a novel gene for Machado-Joseph disease at chromosome14q32.1. Nat Genet 1994;8:221–228.

2. Ross CA. When more is less: pathogenesis of glutamine repeatneurodegenerative diseases. Neuron 1995;15:493–496.

3. Ikeda H, Yamaguchi M, Satoshi S, et al. Expanded polyglu-tamine in the Machado-Joseph disease protein induced celldeath in vitro and in vivo. Nat Genet 1996;13:196–202.

4. Kobayashi T, Tanaka K, Inoue K, Kakizuka A. Functional AT-Pase activity of p97/valosin-containing protein (VCP) is re-quired for the quality control of endoplasmic reticulum in neu-ronally differentiated mammalian PC12 cells. J Biol Chem2002;277:47358–47365.

5. Wang G, Sawai N, Kotliarova S, et al. Ataxin-3, the MJD1gene product, interacts with the two human homologs of yeastDNA repair protein RAD23, HHR23A and HHR23B. HumMol Genet 2000;9:1795–1803.

6. Elbashir S, Harborth J, Lendeckel W, et al. Duplexes of 21nucleotide RNAs mediate RNA interference in cultured mam-malian cells. Nature 2001;411:494–498.

7. Matsumura R, Takayanagi T, Murata K, et al. Relationship of(CAG)nC configuration to repeat instability of the Machado-Joseph disease gene. Hum Genet 1996;98:643–645.

8. Gaspar C, Lopes-Cendes I, Hayes S, et al. Ancestral origins ofthe Machado-Joseph disease mutation: a worldwide haplotypestudy. Am J Hum Genet 2001;68:523–528.

9. Yokota T, Sakamoto N, Enomoto Y, et al. Inhibition of intra-cellular hepatitis C virus replication by synthetic and vector-derived small interfering RNAs. EMBO Rep 2003;4:602–608.

10. Caplen NJ, Taylor JP, Statham VS, et al. Rescue ofpolyglutamine-mediated cytotoxicity by double-stranded RNA-mediated RNA interference. Hum Mol Genet 2002;11:175–184.

11. Miller VM, Xia H, Marrs GL, et al. Allele-specific silencing ofdominant disease genes. Proc Natl Acad Sci USA 2003;100:7195–7200.

12. Elbashir SM, Martinez J, Patkaniowska A, et al. Functionalanatomy of siRNAs for mediating efficient RNAi in Drosophilamelanogaster embryo lysate. EMBO J 2001;20:6877–6888.

13. Holen T, Amarzguioui M, Wiiger MT, et al. Positional effectsof short interfering RNAs targeting the human coagulation trig-ger tissue factor. Nucleic Acids Res 2002;30:1757–1766.

14. Yoshinari K, Miyagishi M, Taira K, et al. Effect on RNAi ofthe tight structure, sequence and position of the targeted re-gion. Nucleic Acids Res 2004;32:691–699.

15. Philips AV, Timchenko L, Cooper TA. Disruption of splicingregulated by a CUG-binding protein in myotonic dystrophy.Science 1998;280:737–741.

16. McLaughlin BA, Spencer C, Eberwine J. CAG trinucleotideRNA repeats interact with RNA-binding proteins. Am J HumGenet 1996;59:561–569.

Is Benign Rolandic EpilepsyGenetically Determined?Lata Vadlamudi, MBBS, FRACP,1

A. Simon Harvey, MD, MBBS, FRACP,1,2

Mary M. Connellan, MPH,1

Roger L. Milne, Grad Dip Clin Epi,3

John L. Hopper, PhD,3

Ingrid E. Scheffer, PhD, MBBS, FRACP,1 andSamuel F. Berkovic, MD, FRACP1

Benign rolandic epilepsy (BRE) is considered to be a ge-netically determined idiopathic partial epilepsy. We stud-ied twins with BRE and compared the concordance witha twin sample of idiopathic generalized epilepsy (IGE).All eight BRE pairs (six monozygous [MZ], two dizygous[DZ]) were discordant. MZ pairwise concordance was 0(95% confidence interval [CI], 0–0.4) for BRE comparedwith 0.7 (95% CI, 0.5–0.9) for 26 IGE MZ pairs. Ourdata suggest that conventional genetic influences in BREare considerably less than for IGE, and other mechanismsneed to be explored.

Ann Neurol 2004;56:129–132

Benign rolandic epilepsy (BRE) is the commonest id-iopathic childhood epilepsy syndrome,1 characterizedby onset usually between 5 and 10 years, unilateral sen-sorimotor seizures, normal neurological development,and the electroencephalogram (EEG) trait of centro-temporal spikes (CTS).2 CTS occur in 1.6% of asymp-tomatic healthy children between 1 and 15 years,3 andit is estimated that only approximately 9% of childrenwith CTS actually have seizures.4 CTS are found in avariety of epilepsies, not just BRE, and are the arche-type of an age-dependent EEG trait, which disappearsby midadolescence.5

BRE is classified as an idiopathic partial epilepsy,“idiopathic” loosely implying “genetic” predisposition.However, a striking finding from family studies ofBRE probands is that few family members have BREor other idiopathic partial epilepsies.6 This contrasts

From the 1Epilepsy Research Centre, Department of Medicine (Neu-rology), University of Melbourne, Austin Health; 2Children’s Epi-lepsy Program, Department of Neurology, RCH; and 3Centre for Ge-netic Epidemiology, University of Melbourne, Melbourne, Australia.

Received Dec 12, 2003, and in revised form Feb 12, 2004. Ac-cepted for publication Apr 6, 2004.

Published online Jun 28, 2004, in Wiley InterScience(www.interscience.wiley.com). DOI: 10.1002/ana.20153

Address correspondence to Dr Berkovic, Epilepsy Research Centre,First Floor, Neurosciences Building, Repatriation Campus, AustinHealth, Banksia Street, Heidelberg West, Victoria, Australia 3081.E-mail: s.berkovics@unimelb.edu.au

© 2004 American Neurological Association 129Published by Wiley-Liss, Inc., through Wiley Subscription Services