BRIEF COMMUNICATIONS

(ioESM) is still the gold standard for the detection of Language Area Localization these essential language areas.2 Noninvasive functional neuroimaging techniques are emerging as promising with Three-Dimensional 4,5 clinical tools for this purpose. Functional Magnetic Functional magnetic resonance imaging (fMRI) is a completely noninvasive preoperative method of deter- Resonance Imaging Matches mining whole-brain functional neuroanatomy.6,7 It en- ables localization of brain function with high spatial Intrasulcal Electrostimulation and temporal resolution by monitoring the hemody- in Broca’s Area namic changes that are coupled with neuronal activity. We report successful identification of Broca’s area G. J. M. Rutten, MD,* P. C. van Rijen, MD, PhD,* within a sulcus 5 to 8 mm below the cortical surface C. W. M. van Veelen, MD, PhD,* and N. F. Ramsey, PhD† using three-dimensional (3D) fMRI and ioESM in a 14-year-old girl with a left inferior frontal desmoplastic ganglioglioma. In this study, intraoperative electrocortical stimulation mapping (ioESM), the current gold standard for the lo- Patient and Methods calization of critical language areas, is compared with Case Report functional magnetic resonance imaging (fMRI) in a 14- The 14-year-old patient has suffered from medically intrac- year-old girl with medically intractable epilepsy caused table complex partial seizures since she was 4 years old. by a tumor in the region of Broca’s area. Prior to the Computed tomography and MRI revealed a calcified, intra- operation, four different fMRI tasks that target inferior axial, noncontrast-enhancing lesion (diameter, 3 cm) within frontal language areas were applied. Prior to the resec- the left inferior frontal gyrus (Broca’s area) and within part tion, ioESM as well as fMRI detected no language areas of the left middle frontal gyrus. The tumor had remained at the exposed cortical area. After removal of the tumor, unchanged in successive MRI scans over a period of 3 years. a unique opportunity presented itself, where ioESM Electroencephalography suggested a left anterior frontal ictal could be performed in the depth of a now exposed and onset and a causal relationship with the lesion. The patient intact gyrus. One specific locus that was indicated to be a occasionally experienced a few minutes of postictal speech critical language area by multiple-task fMRI was targeted. problems. A neurological examination showed no abnormal- IoESM selectively confirmed the location of this language ities. A neuropsychological examination showed no language area to within an estimated 3 mm. We propose that the deficits. Hand preference was predominantly left-sided (Ϫ0.5 combined use of different fMRI tasks increases the sen- on the Edinburgh Handedness Inventory).8 An intracarotid sitivity and specificity for the detection of essential lan- sodium amobarbital test (Wada test) showed a left hemi- guage areas. sphere dominance for language functions.9 Surgical removal of the tumor was decided on, with ioESM of the speech area Rutten GJM, van Rijen PC, van Veelen CWM, under local anesthesia. Ramsey NF. Language area localization with three-dimensional functional magnetic resonance imaging matches intrasulcal electrostimulation in fMRI Broca’s area. Ann Neurol 1999;46:405–408 fMRI was done 1 month prior to surgery using a 3D BOLD technique (navigated PRESTO sequence), implemented on a Philips ACS-NT 1.5-T scanner (Philips Medical Systems, 10,11 Neurosurgical resective procedures in the language- Best, The Netherlands) with standard hardware. Four vi- sually presented language tasks were used in the scanner: pic- dominant hemisphere must avoid the classic language ture naming, verb generation, repetitive reciting of the days areas of Broca (posterior part of the inferior frontal gy- of the week, and verbal fluency.4,12 The patient was in- rus) and Wernicke (perisylvian region of the tem- structed to use covert speech. During the control condition, poroparietal lobe) to prevent persistent language defi- the patient had to fixate her eyes on a small cross hair. Each 1–3 cits. Despite considerable practical disadvantages, task-control combination took about 10 minutes of imaging intraoperative electrocortical stimulation mapping time. A conventional T1-weighted MRI scan completed the session. Functional and anatomical scans were meticulously registered to each other prior to the statistical analysis. Sta- tistical maps were obtained on a voxel by voxel basis for each 10 From the Departments of *Neurosurgery and †Psychiatry, Univer- task (with reference to the control condition). sity Hospital Utrecht, Utrecht, The Netherlands. None of the individual tasks yielded activation at the cor- Received Jan 20, 1999, and in revised form Mar 29 and Apr 21. tical surface (Fig 1). A common area of activation for all four Accepted for publication Apr 23, 1999. tasks was found on the bank of a gyrus located directly Address correspondence to Dr Rutten, Department of Neuro- cranial-posterior to the tumor at a depth of 5 to 8 mm from surgery, University Hospital Utrecht, PO Box 85500, 3508 GA, the cortical surface (see Fig 1). The volume of this area was Utrecht, The Netherlands. 43 mm3 (ie, the volume of 1 functional voxel).

Copyright © 1999 by the American Neurological Association 405 Fig 1. Composite result of multiple-task functional MRI (fMRI). Sagittal slices of the left hemisphere are shown. The most lateral slice is located in the bottom left corner, and the most medial slice is located in the upper right corner. Significant voxels (p Ͻ 0.05, Bonferroni corrected ) are colored and superimposed on anatomical images (in-plane resolution, 2 mm). The red voxel (ar- row) represents the area that is active during all four fMRI language tasks, the blue voxels represent the areas active during verb generation, the white voxels represent the areas active during picture naming, and the green voxel represents the area active during repetitive reciting of the days of the week.

ioESM prior to the Resection The stimulus duration was 5 seconds, and the average cur- Local anesthetics were applied, and propofol was adminis- rent intensity was 9.5 mA (range, 8–13 mA). Arrest of tered as a transient hypnotic drug. A left-sided frontal crani- speech or anomia that could be elicited repetitively was con- otomy was performed. The lesion had partially displaced the sidered to represent a significant language error. During cranial-posterior gyri (Fig 2). Electrocorticography revealed ioESM, no language errors were observed during 29 stimuli no clear epileptic activity. IoESM was performed with bipo- at 17 different cortical sites covering the area of the craniot- lar electrodes that were placed 5 mm apart on the cortical omy, including the tumor. This finding was in concordance surface while the patient performed a picture-naming task.12 with the fMRI results. At two cortical sites, blinking of the

Fig 2. Photographs of the craniotomy (A) before and (B) after resection. The head of the patient is facing toward the left and is aligned roughly along the canthomeatal line. The blue arrow on top of the sterile marker (white) points to the site where language errors were obtained with intraoperative electrocortical stimulation mapping (ioESM). The yellow arrows indicate the sites where ioESM did not induce language errors.

406 Annals of Neurology Vol 46 No 3 September 1999 right eye was elicited repetitively, indicating adequate current Pathology and Follow-Up intensity. At this point, the electrodes were removed, and Following histological and immunochemical analysis of the surgical removal commenced. removed tissue, the tumor was classified as a desmoplastic ganglioglioma. The postoperative clinical course was uncom- ioESM after the Resection plicated, and 6 months after her operation, the patient is still The tumor was radically removed (see Fig 2). As the wall of free of seizures. She has experienced no language deficits. the gyrus where fMRI had identified a critical language area was meticulously left intact and exposed, it was decided that Discussion this locus would be tested with ioESM. The patient was This study demonstrates an excellent agreement between again awakened and showed no signs of dysphasia. Electro- fMRI and ioESM of frontal language function in a 14-year- corticography revealed no epileptic activity. Speech arrest was old girl with a tumor in Broca’s area. The ioESM-derived then elicited repeatedly in the depth of the now exposed wall intrasulcal language area was located approximately 3 mm of a gyrus located cranial-posterior to the removed tumor at medial to the language area that was found with multiple- an estimated 10 mm from the cortical surface (see Fig 2). task fMRI. This difference may be explained by the fact that Intraoperative landmarks were compared with a surface ren- the brain surface shifts during surgery. Roberts and col- dering of the patient’s brain and with fMRI images (Fig 3). leagues13 reported an average brain shift of 10 mm as a result The site where speech arrest was elicited was located approx- of gravity alone, probably as a “consequence of displacement imately 3 mm medial to the locus where activity was found of cerebrospinal fluid and resulting loss of buoyancy.”13 with fMRI. ioESM at three other sites along this exposed To our knowledge, this is the first reported case in which gyrus (2 of which were located within 5 mm of the ioESM- fMRI of language areas has been validated with intrasulcal positive language site) induced no language errors, confirm- electrocortical stimulation. This case underscores the fact ing the fMRI results (see Fig 2). that routinely applied ioESM, as opposed to fMRI, cannot

Fig 3. Photographs indicating concordance of intraoperative electrocortical stimulation mapping (ioESM) and functional MRI (fMRI) localization of Broca’s area. (Top row) Images from the neuronavigational apparatus. The yellow pointer indicates the lo- cation of the intrasulcal language area found with ioESM. (Bottom row) fMRI images. Significant activation that was common in four language tasks is shown in red and is indicated by the yellow arrows. Left to right: transverse, coronal, and sagittal sections (radiological orientation).

Brief Communication: Rutten et al: 3D fMRI of Broca’s Area 407 identify language areas that are not in the immediate vicinity We are grateful to Philips Medical Systems, Best, The Netherlands, of the cortical surface. From earlier studies, it is known that for the use of the neuronavigational equipment. We also thank the frontal language sites found with ioESM are usually confined Dutch Collaborative Epilepsy Surgery Programme, Prof A. C. van to a small area in the most posterior portion of the inferior 2,3,14 Huffelen for electrophysiological recordings, P. W. A. Willems for frontal gyrus of the language-dominant hemisphere. In his assistance during surgery, and Prof W. P. Th. M. Mali for his an estimated 20% of patients, however, ioESM does not de- continued support of fMRI activities. tect language areas in this classic Broca’s area.2 Previous fMRI studies have shown that frontal language areas in nor- mal volunteers as well as epileptic patients are predominantly situated in gray matter that is located deep to the classic cor- References tical surface representation of Broca’s area.15,16 This may ex- 1. Geschwind N. The organization of language and the brain. Sci- plain the relative lack of sensitivity of ioESM in detecting ence 1970;170:940–944 inferior frontal language areas. The few cases in which 2. Ojemann GA, Sutherling WW, Lesser RP, et al. Cortical stim- ulation. In: Engel J, Jr, ed. Surgical treatment of the epilepsies. ioESM has included sulci have not shown independent in- New York: Raven Press, 1993:399–414 trasulcal language areas, although sometimes a surface site ex- 14 3. Penfield W, Roberts L. Speech and brain mechanisms. Prince- tended a short distance into the sulcus. In our patient, be- ton, NJ: Princeton University Press, 1959 cause of the mass effect of the lesion, the essential frontal 4. Binder JR. Neuroanatomy of language processing studied with language area has been displaced in cranial-posterior direc- functional MRI. Clin Neurosci 1997;4:87–94 tion. We did not detect any displacement of cortical struc- 5. Buchbinder BR, Cosgrove GR. Cortical activation MR studies tures into the depth of the cranial-posterior sulcus. There- in brain disorders. Magn Reson Imaging Clin North Am 1998; fore, we have demonstrated the existence of an essential 6:67–93 intrasulcal language area with no representation at the corti- 6. Belliveau JW, Kwong KK, Kennedy DN, et al. Magnetic reso- cal surface. nance imaging mapping of brain function. Invest Radiol 1992; Because of the intrinsically low signal-to-noise ratio of 27(Suppl):S59–S65 fMRI scanners, it is essential to repeat and average across 7. Ogawa S, Tank GW, Menon RS, et al. Intrinsic signal changes task and rest conditions to minimize false-negative results. accompanying sensory stimulation: functional brain mapping with magnetic resonance imaging. Proc Natl Acad Sci USA Our scan method (3D-PRESTO) scans a relatively large vol- 1992;89:5951–5955 ume (77-mm slab) in a short period of time (7.25 seconds). 8. Oldfield RC. The assessment and analysis of handedness: the These short scan times allow for a large number of repeti- Edinburgh inventory. Neuropsychologia 1971;9:97–113 tions and limit motion artifacts. 3D fMRI is less susceptible 9. Wada J, Rasmussen T. Intracarotid injection of sodium Amytal to inflow effects than single or multislice methods, thereby for the lateralization of cerebral speech dominance: experimen- further minimizing false-positive as well as false-negative re- tal and clinical observations. J Neurosurg 1960;17:266–282 sults.17 Functional and structural 3D volumes can be regis- 10. Ramsey NF, Van den Brink JS, Van Muiswinkel AMC, et al. tered directly to one another without correction for spatial Phase navigator correction in 3D fMRI improves detection of distortion, minimizing errors caused by misregistration. brain activation: quantitative assessment with a graded motor In general, frontal as well as temporal language areas activation procedure. Neuroimage 1998;8:240–248 found with ioESM are considerably smaller than those found 11. Van Gelderen P, Ramsey NF, Liu G, et al. Three-dimensional with fMRI.2–4,14–16,18 This is probably due to the fact that functional magnetic imaging of human brain on a clinical fMRI does not primarily distinguish between areas that are 1.5-T scanner. Proc Natl Acad Sci USA 1995;92:6906–6910 12. Snodgrass JG, Vanderwart MA. Standardized set of 260 essential for language function and areas that participate in pictures: norms of name, agreement, usage agreement, familiar- function but are not critical in the performance of language ity, and visual complexity. J Exp Psychol Learn Mem Cogn function or can be compensated for by other areas. This dis- 1980;6:174–215 tinction has to be made in order to validate fMRI as a useful 13. Roberts DW, Hartov AH, Kennedy FE, et al. Intraoperative neurosurgical tool. We found that a frontal brain area that brain shift and deformation: a quantitative analysis of cortical participated in four different fMRI tasks showed excellent displacement in 28 cases. Neurosurgery 1998;43:749–760 correlation with ioESM. Determining the overlapping re- 14. Ojemann GA. Some brain mechanisms for reading. In: Vol Eu- gions of activation in multiple-task fMRI may therefore be ler C, Lundberg I, Lennerstrand G, eds. Brain and reading. an accurate technique to identify essential language areas New York: MacMillan, 1989:47–59 with maximum specificity. 15. Desmond JE, Sum JM, Wagner AD, et al. Functional MRI This study underlines the fact that routinely applied measurement of language lateralization in Wada-tested patients. ioESM cannot identify essential language areas that are bur- Brain 1995;118:1411–1419 ied in a sulcus even at a depth of 5 to 10 mm. Taking this 16. Hinke RM, Hu X, Stillman AE, et al. Functional magnetic res- onance imaging of Broca’s area during internal speech. Neuro- into consideration, ioESM cannot be an absolutely accurate report 1993;4:675–678 predictor of the effects of cortical resection. In order for 17. Duyn JH, Mattay VS, Sexton RH, et al. 3-Dimensional func- fMRI to become a reliable presurgical tool, the relationship tional imaging of human brain using echo-shifting FLASH between involved and essential language cortex has to be MRI. Magn Reson Med 1994;32:150–155 further clarified and task designs (such as multiple-task 18. Berger MS, Kincaid J, Ojemann GA, Lettich E. Brain mapping and event-related fMRI) have to be further optimized and techniquesto maximize resection, safety, and seizure control in validated. children with brain tumours. Neurosurgery 1989;25:786–792

408 Annals of Neurology Vol 46 No 3 September 1999 microsomes and peroxisomes, but only peroxisomal Role of Very-Long-Chain VLCS activity is reduced in X-ALD cells.5,6 Acyl–Coenzyme A Initially, it was expected that X-ALD patients would have mutations in the gene encoding a peroxisomal Synthetase in X-Linked VLCS. The defective ALD gene was identified by using a positional cloning strategy,7 and mutations in this Adrenoleukodystrophy gene have been found in all X-ALD patients whose en- Steven J. Steinberg, PhD,* Stephan Kemp, PhD,† tire gene has been examined.8 Adrenoleukodystrophy Lelita T. Braiterman, PhD,† protein (ALDP), the product of the ALD gene, belongs and Paul A. Watkins, MD, PhD* to the ATP-binding cassette transporter protein super- family. ALDP does not have VLCS activity (Steinberg SJ, Watkins PA, unpublished data, 1998) and its pre- X-linked adrenoleukodystrophy (X-ALD) is characterized cise function remains unknown. Although X-ALD is biochemically by decreased ability of cells to activate (via very-long-chain acyl–coenzyme A synthetase [VLCS]) and not caused by mutations in VLCS gene(s), investigating subsequently degrade very-long-chain fatty acids in per- the properties of VLCS is essential to clarifying the oxisomes. It is noteworthy that the gene defective in pathogenetic mechanism of this disorder. To reconcile X-ALD encodes ALDP, a peroxisomal membrane protein the biochemical and genetic defects in X-ALD, it must unrelated to VLCS. We cloned human VLCS (hVLCS) be concluded that VLCS and ALDP interact function- and found that peroxisomes from X-ALD fibroblasts con- ally and perhaps directly. Investigating the role of per- tained immunoreactive hVLCS, refuting the earlier hy- oxisomal VLCS in peroxisomal ␤-oxidation also may pothesis that ALDP is required to anchor VLCS to the clarify the function of ALDP. We recently cloned and peroxisomal membrane. Furthermore, hVLCS was topo- characterized the human ortholog (hVLCS ) of the gene graphically oriented facing the peroxisomal matrix in encoding a VLCS purified from rat liver peroxisomes.9 both control and X-ALD fibroblasts, contradicting the al- ternative hypothesis that ALDP is required to translocate In this report, we examine the relationship between VLCS into peroxisomes. However, overexpression of both ALDP and hVLCS. hVLCS and ALDP in X-ALD fibroblasts synergistically increased very-long-chain fatty acid ␤-oxidation, indicat- ing that these proteins interact functionally. Subjects and Methods Steinberg SJ, Kemp S, Braiterman LT, Cell Lines and Immunofluorescence Analysis Watkins PA. Role of very-long-chain acyl– Primary and SV40-transformed human control and X-ALD coenzyme A synthetase in X-linked skin fibroblasts, obtained from the Mental Retardation Re- adrenoleukodystrophy. search Center of the Kennedy Krieger Institute, were origi- Ann Neurol 1999;46:409–412 nally obtained with informed consent under a protocol ap- proved by the Joint Committee for Clinical Investigation of the Johns Hopkins Medical Institutions. The four primary X-linked adrenoleukodystrophy (X-ALD; McKuisick X-ALD cell lines and the SV40-transformed X-ALD cell line 300100) is a neurodegenerative disorder with varied used were all immunonegative for ALDP (data not shown). phenotypic expression associated with the accumula- Two of the primary X-ALD cell lines had missense muta- tion of very-long-chain fatty acids (VLCFAs; chain tions (A123V and R418W) as did the transformed line (A626T); the mutations in the other two X-ALD cell lines length Ͼ22) in plasma, brain, adrenals, and cultured 1–3 have not been determined. Indirect immunofluorescence skin fibroblasts. Elevated VLCFA levels in X-ALD analysis was performed as described previously,10 using anti- cells are associated with reduced ability to degrade body specific for hVLCS,9 anti-human catalase (The Binding ␤ 4 these fatty acids by peroxisomal -oxidation. This, in Site) or anti-ALDP.10 turn, is caused by decreased activity of peroxisomal very-long-chain acyl–coenzyme A synthetase (VLCS), the enzyme that activates VLCFA to their coenzyme A Transfection of Fibroblasts and Measurement of Acyl- 5,6 (CoA) derivatives. VLCS activity is found in both CoA Synthetase and ␤-Oxidation Activity cDNA encoding full-length hVLCS and ALDP in pcDNA3 (Invitrogen) were prepared as previously described.9,11 For From the Kennedy Krieger Institute and Departments of *Neurol- each transfection, cells from one T-150 flask of SV40- ogy and †Pediatrics, Johns Hopkins University School of Medicine, transformed X-ALD fibroblasts was electroporated at 300 V Baltimore, MD. with 10 ␮g of full-length cDNA or vector as previously de- 9 Received Mar 22, 1999, and in revised form Apr 30. Accepted for scribed. Three days after transfection, cells were assayed for publication May 14, 1999. acyl-CoA synthetase9 or ␤-oxidation activity.12 Statistical Address correspondence to Dr Watkins, Kennedy Krieger Institute, comparison was done by using a one-tailed Student’s t test of 707 N. Broadway, Baltimore, MD 21205. paired samples.

Copyright © 1999 by the American Neurological Association 409 Results and Discussion punctate signal colocalized with the peroxisomal Peroxisomal hVLCS Protein in Control and marker catalase (see Fig, B). This result is similar to X-ALD Fibroblasts our earlier findings in human hepatocarcinoma– 9 In fibroblasts from X-ALD patients, defects in ALDP derived HepG2 cells. Although hVLCS was found in result in decreased peroxisomal VLCS enzyme activity both endoplasmic reticulum and peroxisomes by im- by an unknown mechanism. One hypothesis is that munoblot analysis of HepG2 subcellular fractions, the 9 ALDP is required to associate hVLCS with the perox- immunofluorescence pattern was primarily punctate. isomal membrane.7,13 To investigate this possibility, We next examined the hVLCS immunostaining pat- we cloned the gene encoding hVLCS, the human or- tern of X-ALD fibroblasts and found that the protein tholog of the VLCS purified from rat liver peroxi- was clearly present (see Fig, C) and colocalized with somes,14 and produced a polyclonal antibody specific catalase (see Fig, D). Although all four X-ALD fibro- for this protein.9 We examined fibroblasts from 4 nor- blast lines examined lacked detectable ALDP (not mal controls and 4 X-ALD patients by indirect immu- shown), their peroxisomal hVLCS immunofluorescence nofluorescence analysis, using this antibody. When was indistinguishable from controls. These results sug- normal cells were permeabilized with Triton X-100 be- gest that the decreased VLCS activity of X-ALD fore incubation with antibody, the immunofluores- peroxisomes is not caused by the absence of hVLCS cence pattern of hVLCS was primarily punctate (Fig, protein. A). Double-labeling experiments indicated that this Topography of hVLCS in Control and X-ALD Fibroblasts Fig. Indirect immunofluorescence of fibroblasts from normal A second hypothesis to explain the reduced VLCS ac- controls and X-linked adrenoleukodystrophy (X-ALD) patients. tivity in X-ALD peroxisomes is that functional ALDP Cells were plated on 18-mm circular coverslips and processed is required to translocate the VLCS protein into the 24 to 48 hours after plating. Fixation, permeabilization, and organelle.7 This hypothesis assumes that hVLCS is ori- incubation with antibodies was as previously described.10 (A–D) Cells permeabilized with Triton X-100. (A) Represen- ented facing the peroxisomal matrix and not the cyto- tative control cell line incubated with anti–human very-long- plasm. Although the topographic orientation of VLCS 15,16 chain acyl–coenzyme A synthetase (anti-hVLCS). (B) The in peroxisomes has been controversial, we recently same cell examined with anti-catalase. (C) Representative reported that the C-terminus of hVLCS is oriented fac- X-ALD fibroblast line with anti-hVLCS. (D) The same cell ing the peroxisomal matrix in HepG2 cells.9 To deter- examined with anti-catalase. (E–H) Cells permeabilized with mine whether hVLCS had a similar orientation in fi- digitonin (15 ␮g/ml). (E) Control cells with anti-hVLCS. broblast peroxisomes, cells were permeabilized with (F) The same cells with anti-catalase. (G) Control cells with digitonin (15 ␮g/ml) rather than Triton X-100 before anti-ALDP (anti-adrenoleukodystrophy protein). (H) X-ALD immunofluorescence analysis. This allows antibodies fibroblasts with anti-hVLCS. access to the cytoplasmic compartment but not the interior of peroxisomes and other organelles.17 As illustrated in part E of the Figure, no hVLCS im- munofluorescence was observed in four digitonin- permeabilized control cell lines. This result suggests that the C-terminus of fibroblast hVLCS faces the peroxisomal matrix, in agreement with our observa- tions in HepG2 cells.9 As expected, no immunofluo- rescence of the peroxisomal matrix protein catalase was observed under these conditions (see Fig, F), whereas immunofluorescence of ALDP, whose C-terminal hy- drophilic domain faces the cytoplasm,9,10 was readily detected (see Fig, G). If ALDP is required to translocate VLCS into peroxisomes, hVLCS associated with peroxisomes (see Fig, C) should be found only on the exterior of the organelle in X-ALD fibroblasts. However, we found no hVLCS immunofluorescence in digitonin-permeabilized fibroblasts from the 4 X-ALD patients (see Fig, H), indicating that the C-terminus of hVLCS had entered the peroxisome. Therefore, we conclude that hVLCS entered the peroxisome or peroxisomal membrane by an ALDP-independent mechanism.

410 Annals of Neurology Vol 46 No 3 September 1999 Functional Interaction of ALDP and hVLCS rescence) was equivalent in cells transfected with ALDP Results presented above have negated two hypotheses alone or both hVLCS and ALDP; thus, differential that have been proposed to explain the relationship be- ALDP expression did not explain this synergistic effect. tween ALDP and hVLCS. We therefore suggest that a These results indicate that ALDP and VLCS can in- functional, biochemical relationship must exist between teract functionally to regulate peroxisomal VLCFA these two proteins that affects peroxisomal ␤-oxidation ␤-oxidation. of VLCFAs. To test this, we transfected SV40- transformed X-ALD fibroblasts (TXFs) with full-length Conclusions cDNA encoding hVLCS, ALDP, or both. As a control, From the data presented here, we conclude that cells were transfected with the pcDNA3 vector alone. hVLCS can interact functionally with ALDP to affect TXFs transiently expressing hVLCS, ALDP, or both VLCFA ␤-oxidation, although previously postulated hVLCS and ALDP were assayed for their ability to ac- physical interactions of hVLCS with ALDP (eg, bind- tivate the VLCFA lignoceric acid (C24:0) and the ing or translocation) are unlikely. The nature of the long-chain fatty acid palmitic acid (C16:0) to their functional interaction is not yet clear. It should be CoA derivatives and to ␤-oxidize these fatty acids. noted that the protein family that includes hVLCS Compared with vector-transfected TXFs, there was a contains at least six human and five mouse members,18 twofold increase in total cellular VLCS activity (C24:0 and the distribution of these VLCS sequence homo- activation) in cells expressing either hVLCS alone or logues varies both in tissues and in species. For in- both hVLCS and ALDP (Table). Overexpression of stance, the mouse ortholog of hVLCS (mVLCS) was ALDP alone did not alter the rate of C24:0 activation. not detected in mouse fibroblasts when examined by In contrast, TXFs overexpressing exogenous ALDP reverse transcription–polymerase chain reaction (Kemp demonstrated a 3.2-fold increase in their ability to S, unpublished data, 1998). Despite the lack of ␤-oxidize C24:0 (see Table). This is in agreement with mVLCS, wild-type mouse fibroblasts degraded VLCFA our previous observation that overexpression of ALDP at rates comparable with human fibroblasts, suggesting in TXFs improved C24:0 ␤-oxidation.11 Overexpres- that in this species peroxisomal VLCFA activation is sion of hVLCS alone in these cells did not significantly catalyzed by a different enzyme. The mouse equivalent affect their ability to degrade C24:0. However, when of ALDP is also required for VLCFA ␤-oxidation, be- both ALDP and hVLCS were simultaneously overex- cause this process is deficient in a mouse model of hu- pressed in TXFs, a synergistic effect on VLCFA man X-ALD.19 Our studies do not preclude the possi- ␤-oxidation was observed. Improvement in C24:0 ox- bility that another human VLCS exhibits greater idation was 5.4-fold greater than pcDNA3-transfected dependency on ALDP than does hVLCS. Furthermore, cells. This was a statistically significant improvement overexpression of either ALD-related protein or compared with cells transfected with ALDP alone (p Ͻ PMP70 (peroxisomal ABC transporters closely related 0.05). ALDP expression (determined by immunofluo- to ALDP) partially restoredVLCFA ␤-oxidation defects

Table. ␤-Oxidation and Acyl-CoA Synthetase Activities in SV40-Transformed Human X-ALD Fibroblasts Overexpressing VLCS and/or ALDP

Ϫ ␤-Oxidation (nmol ⅐ hr 1 ⅐ mg Acyl-CoA Synthetase (nmol ⅐ 20 Ϫ Ϫ Ϫ of protein 1) min 1 ⅐ mg of protein 1)

C24:0 C16:0 C24:0 C16:0 pcDNA3 0.21 Ϯ 0.05 1.61 (1.44–1.77) 1.33 Ϯ 0.40 32.7 Ϯ 6.0 hVLCS 0.27 Ϯ 0.04 1.41 (1.38–1.44) 2.80 Ϯ 0.44 32.4 Ϯ 11.3 ALDP 0.68 Ϯ 0.13 1.69 (1.59–1.79) 1.23 Ϯ 0.38 26.6 Ϯ 5.8 hVLCS ϩ ALDP 1.13 Ϯ 0.26 1.85 (1.67–2.02) 2.77 Ϯ 0.59 37.9 Ϯ 2.6 n3233

SV40-transformed X-ALD fibroblasts (TXFs) were transiently transfected with full-length hVLCS, ALDP, hVLCS ϩ ALDP, or pcDNA3 vector alone. The fibroblasts were cultured for 3 days after transfection and then harvested for immediate analysis to determine ␤-oxidation of lignoceric (C24:0) and palmitic (C16:0) acids. For comparison, two SV40-transformed normal control fibroblast cell lines were transfected with Ϫ Ϫ pcDNA3 in duplicate and then assayed for ␤-oxidation; the mean oxidation of C24 and C16 was 0.55 and 1.29 nmol ⅐ hour 1 ⅐ mg protein 1, respectively. Cell suspensions remaining from ␤-oxidation assays were stored at Ϫ80°C and were later used to measure their ability to activate C24:0 and C16:0 to CoA derivatives. The number of cell pellets from independent transfections assayed (n) are indicated. The standard deviation for each parameter is shown, except in the case of C16:0 ␤-oxidation, where the range is reported. To estimate transfection efficiency, a coverslip was present in culture dishes of all cells expressing ALDP. Immunofluorescence analysis by using anti-ALDP antiserum revealed that 49 Ϯ 10% of cells transfected with ALDP alone and 53 Ϯ 4% of cells transfected with hVLCS ϩ ALDP expressed ALDP. CoA ϭ coenzyme A; X-ALD ϭ X-linked adrenoleukodystrophy; VLCS ϭ very-long-chain acyl-coenzyme A synthetase; ALDP ϭ adrenoleu- kodystrophy protein; hVLCS ϭ human VLCS.

Brief Communication: Steinberg et al: Human VLC Acyl-CoA Synthetase 411 in TXFs,8,11 implying that these proteins could also in- long-chain-fatty-acid-activating activity in peroxisomes from rat teract with hVLCS. Further studies should clarify the liver. Biochem J 1991;276:53–56 nature of the functional interactions between ALDP, 16. Lazo O, Contreras M, Singh I. Topographical localization of peroxisomal acyl-CoA ligases—differential localization of related ABC transporters, hVLCS and its homologues, palmitoyl-CoA and lignoceroyl-CoA ligases. Biochemistry 1990; and their roles in the pathogenesis of X-ALD. 29:3981–3986 17. Swinkels BW, Gould SJ, Bodnar AG, et al. A novel, cleavable peroxisomal targeting signal at the amino-terminus of the rat 3-ketoacyl-CoA thiolase. EMBO J 1991;10:3255–3262 Supported by NIH grants HD10981, NS10533, NS37355, and 18. Watkins PA, Pevsner J, Steinberg SJ. Human very long-chain HD24061. acyl-CoA synthetase and two human homologs: initial charac- We thank Ann Moser for providing cells from the Mental Retarda- terization and relationship to fatty acid transport protein. Pros- tion Research Center, Dr Hugo Moser for his continued support taglandins Leukot Essent Fatty Acids 1999;60:323–328 and advice, and Dr Kirby Smith for many thoughtful discussions. 19. Lu J-F, Lawler AM, Watkins PA, et al. A mouse model for X-linked adrenoleukodystrophy. Proc Natl Acad Sci USA 1997; 94:9366–9371 References 1. Moser HW, Smith KD, Moser AB. X-linked adrenoleukodys- Plasma Amyloid ␤-Peptide trophy. In: Scriver CR, Beaudet AL, Sly WS, Valle D, eds. The metabolic and molecular bases of inherited disease. New York: 1–42 and Incipient McGraw-Hill, 1995:2325–2349 2. Moser HW, Moser AB, Frayer KK, et al. Adrenoleukodys- Alzheimer’s Disease trophy: increased plasma content of saturated very long chain ࿣ fatty acids. Neurology 1981;31:1241–1249 Richard Mayeux, MD, MSc,*†‡§ 3. Moser HW, Moser AB, Kawamura N, et al. Adrenoleuko- Ming-Xin Tang, PhD,*†¶ Diane M. Jacobs, PhD,*†‡ dystrophy: elevated C-26 fatty acid in cultured skin fibroblasts. Jennifer Manly, PhD,*†‡ Karen Bell, MD,*†‡ Ann Neurol 1980;7:542–549 Carol Merchant, MD, MPH,*†‡ Scott A. Small, MD,*†‡ 4. Singh I, Moser AB, Goldfischer S, Moser HW. Lignoceric acid Yaakov Stern, PhD,*†‡§ is oxidized in the peroxisome: implications for the Zellweger Henry M. Wisniewski, MD, PhD,# cerebro-hepato-renal syndrome and adrenoleukodystrophy. Proc and Pankaj D. Mehta, PhD Natl Acad Sci USA 1984;81:4203–4207 5. Wanders RJA, van Roermund CWT, van Wijland MJA, et al. Direct demonstration that the deficient oxidation of very long Mutations in the amyloid precursor protein and preseni- chain fatty acids in X-linked adrenoleukodystrophy is due to an lin 1 and 2 genes result in elevated plasma levels of the impaired ability of peroxisomes to activate very long chain fatty amyloid ␤-peptide species terminating at amino acid res- acids. Biochem Biophys Res Commun 1988;153:618–624 ␤ 6. Lazo O, Contreras M, Hashmi M, et al. Peroxisomal idue 42 (A 1–42). In a longitudinal study of unrelated lignoceroyl-CoA ligase deficiency in childhood adrenoleukodys- elderly individuals, those who subsequently developed ␤ trophy and adrenomyeloneuropathy. Proc Natl Acad Sci USA Alzheimer’s disease had higher plasma levels of A 1–42 1988;85:7647–7651 at entry than did those who remained free of dementia. 7. Mosser J, Douar AM, Sarde CO, et al. Putative X-linked adre- The results indicate that elevated plasma levels of the re- noleukodystrophy gene shares unexpected homology with ABC leased A␤ peptide A␤1–42 may be detected several years transporters. Nature 1993;361:726–730 before the onset of symptoms, supporting that extracel- 8. Smith KD, Kemp S, Braiterman LT, et al. X-linked lular A␤1–42 plays an important role in the pathogenesis adrenoleukodystrophy: genes, mutations, and phenotypes. Neu- of late-onset Alzheimer’s disease. rochem Res 1999;24:521–535 9. Steinberg SJ, Wang SJ, Kim DG, et al. Human very long-chain Mayeux R, Tang M-X, Jacobs DM, Manly J, acyl-CoA synthetase: cloning, topography and relevance to Bell K, Merchant C, Small SA, Stern Y, branched-chain fatty acid metabolism. Biochem Biophys Res Wisniewski HM, Mehta PD. Plasma amyloid ␤- Commun 1999;257:615–621 peptide 1–42 and incipient Alzheimer’s disease. 10. Watkins PA, Gould SJ, Smith MA, et al. Altered expression of ALDP in X-linked adrenoleukodystrophy. Am J Hum Genet Ann Neurol 1999;46:412–416 1995;57:292–301 11. Braiterman LT, Zheng S, Watkins PA, et al. Suppression of per- oxisomal membrane protein defects by peroxisomal ATP binding cassette (ABC) proteins. Hum Mol Genet 1998;7:239–247 From the *Taub Alzheimer’s Disease Research Center, †Gertrude H. Sergievsky Center, Departments of ‡Neurology and §Psychiatry, 12. Watkins PA, Ferrell EV Jr, Pedersen JI, Hoefler G. Peroxisomal ࿣ fatty acid beta-oxidation in HepG2 cells. Arch Biochem Bio- Divisions of Epidemiology and ¶Biostatistics, School of Public phys 1991;289:329–336 Health, Columbia University, and #Department of Immunology, 13. Contreras M, Mosser J, Mandel JL, et al. The protein coded by Institute for Basic Research in Developmental Disabilities, Staten Island, NY. the X-adrenoleukodystrophy gene is a peroxisomal integral membrane protein. FEBS Lett 1994;344:211–215 Received Mar 25, 1999, and in revised form May 6. Accepted for 14. Uchida Y, Kondo N, Orii T, Hashimoto T. Purification and publication May 6, 1999. properties of rat liver peroxisomal very-long-chain acyl-CoA Address correspondence to Dr Mayeux, Gertrude H. Sergievsky synthetase. J Biochem (Tokyo) 1996;119:565–571 Center, 630 West 168th Street, Columbia University, New York, 15. Lageweg W, Tager JM, Wanders RJA. Topography of very- NY 10032.

412 Copyright © 1999 by the American Neurological Association The amyloid ␤-peptide (A␤) species terminating at for age, because levels of both A␤1–40 and A␤1–42 corre- amino acid residue 42 (A␤1–42) is deposited earlier lated with age, and for education, ethnic group, and apoli- in amyloid plaques in Alzheimer’s disease (AD) than the poprotein E (ApoE) genotype as each was associated with shorter form ending at amino acid 40 (A␤1–40).1,2 AD. Odds ratios (ORs) for levels of A␤1–40 and A␤1–42 A␤1–42 is fibrillogenic, facilitating accrual of A␤1–40.3,4 associated with AD were estimated as a continuous variable Amyloid precursor protein (APP) and presenilin 1 and and then by quartiles, using logistic regression with adjust- 2 (PS1/2) mutations in familial early-onset AD foster ments for age and other confounders. In 96 subjects with repeat A␤ measurements, A␤1–40 and A␤1–42 levels were A␤ deposition by increasing its extracellular concentra- 5 compared by computing the difference and by assessing con- tion. Carriers of APP and PS1/2 mutations have ele- sistency over time, using the intraclass correlation coefficient. vated plasma levels of A␤1–42 compared with noncar- riers.6,7 Scheuner and colleagues6 reported elevated ␤ plasma A 1–42 in 2 of 9 controls who developed AD, Results suggesting that higher concentrations might be detect- Measurements of plasma A␤1–40 and A␤1–42 were able before disease onset. We tested the hypothesis that performed in 169 healthy individuals (mean age, 74.9 plasma levels of A␤1–42 or the plasma ratio of A␤1– ␤ years; range, 65–92 years; 70% women, 36% African- 42/A 1–40 might be elevated before the onset of late- American, 24% white, and 40% Hispanic) free of de- onset, sporadic AD. mentia or cognitive impairment at the baseline visit. Each had an initial CDR10 score of 0 and all were Subjects and Methods followed annually, using the same clinical assessments,8 Subjects and Assessments for an average of 3.6 years. At the final examination, Data were used from individuals in a longitudinal study of 105 (62%) remained without dementia or cognitive aging and dementia who had received an annual standardized impairment, and 64 (38%) had developed AD. The medical and neurological history and examination and neu- total person-years of follow-up for the group was 603.1 ropsychological tests.8 For the diagnosis of probable and pos- years, yielding an overall annual incidence rate of sible AD, patients were required to meet the criteria of 10.6% (95% confidence interval, 5.9–15.2%). NINCDS-ADRDA (National Institute of Neurological and Baseline plasma A␤1–40 and A␤1–42 levels corre- Communicative Disorders and Stroke–Alzheimer’s Disease ϭ 9 lated significantly with age (p 0.001) and with each and Related Disorders Association), and severity was rated other (p ϭ 0.01), but did not vary by sex, ethnic by using the Clinical Dementia Rating Scale (CDR).10 Di- agnoses and CDR rating were made without knowledge of groups, ApoE genotype, or family history of AD. plasma A␤1–40 and A␤1–42 levels. Plasma levels of Those who developed AD were older and less well ed- A␤1–40 and A␤1–42 were repeated in 96 randomly selected ucated than those who remained dementia-free (Table subjects at the final assessment (3–4 years later). The fre- 1). Baseline plasma A␤1–40 levels were 20% higher quencies of dementia and other characteristics in these indi- (p ϭ 0.002) and the baseline plasma A␤1–42 levels viduals were similar to those with a single A␤ measurement. were 60% higher (p ϭ 0.001) in those who developed

Analysis of Plasma A␤1–40 and A␤1–42 Plasma obtained at baseline had been stored, within 2 hours Table 1. A␤1–40, A␤1–42, and A␤1–42/A␤1–40 Levels at after collection, at Ϫ70°C. A␤1–40 and A␤1–42 levels were Baseline in Subjects Followed over an Average of 3.6 Years measured by using a combination of monoclonal antibody 6EI0 (specific to an epitope present on 1–16-amino acid No Dementia Alzheimer’s Disease residues of A␤) and R162 (vs A␤1–40) and R165 (vs A␤1– (CDR 0) (CDR 0.5 and 1.0) 42) antisera, respectively, in a double-antibody sandwich enzyme-linked immunosorbent assay.11,12 The detection n 105 64 ␤ Age (yr) 73.4 Ϯ 5.3 77.4 Ϯ 5.9a limit for this assay was 5 pg/ml for A 1–40 and 10 pg/ml Ϯ Ϯ a ␤ Education 9.3 4.6 7.5 3.8 for A 1–42. The test–retest reliability of the measurement ␤ Ϯ Ϯ a ␤ A 1–40 pg/ml 111.8 4.3 134.7 5.8 of plasma A was excellent in repeated measurements of 106 ␤ Ϯ Ϯ a ␣ ϭ A 1–42 pg/ml 51.5 4.1 82.4 8.6 stored samples (Cronbach’s coefficient 0.91). A␤1–42/A␤1–40 0.51 Ϯ 0.04 0.67 Ϯ 0.07b

All subjects were without evidence of dementia (CDR 0)10 at base- Statistical Analysis line when plasma samples were obtained. Comparison of mean val- The incidence rate for AD was calculated by dividing the ues (ϮSEM) between those whose CDR score remained 0 com- number of individuals with AD by the total number of pared with those whose CDR scores increased to 0.5 or 1.0 at ␤ follow-up based on clinical, functional, and neuropsychological as- person-years of follow-up. Baseline plasma A 1–40 and a ϭ b Ͻ ␤ sessment indicated incipient Alzheimer’s disease ( p 0.001, p A 1–42 levels were compared in those who developed AD, 0.05). A␤1–42 levels, but not A␤1–40, remained significantly dif- defined as CDR 0.5 or 1.0, and those who remained free of ferent in multivariate analysis adjusting for age, education, apoli- dementia, CDR 0, using Student’s t test. Multivariate anal- poprotein E genotype, and ethnic group. ysis of variance of baseline A␤ levels was performed adjusting CDR ϭ Clinical Dementia Rating Scale.

Brief Communication: Mayeux et al: A␤1–42 in Alzheimer’s Disease 413 the analysis restricting the sample to subjects still dementia-free at 26 months after baseline. However, those who developed AD by the last follow-up (18 months later) still had elevated plasma A␤1–42 (data not shown; F ϭ 5.4, p ϭ 0.02 for A␤1–42; F ϭ 3.9, p ϭ 0.05 for ratio A␤1–42/A␤1–40). In repeated measurements of A␤ levels, the final plasma A␤1–42 was an average of 3% (1.47 pg/ml; SEM 2.0) lower than the first, but this difference was the same in those who developed AD and those who remained free of dementia (F ϭ 0.001, p ϭ 0.9). The final measurement of A␤1–40 was an average of 12% (14.7 pg/ml; SE 3.7) lower than the first, but also did not differ between groups (F ϭ 0.13, p ϭ 0.7). The consistency in the measures of plasma A␤ levels over time was high (intraclass correlation coefficient, A␤1– Fig. Plasma A␤1–40 and A␤1–42 levels from first assessment 42 ϭ 0.91 and A␤1–40 ϭ 0.81). when none of the subjects were found to have cognitive im- The Figure illustrates that 62.5% of individuals who pairment or dementia; 64 individuals subsequently developed developed incipient AD exceeded the group median ● Alzheimer’s disease (AD) ( ) over an average of 3.6 years of value (45.9 pg/ml) for plasma A␤1–42 at baseline follow-up, and 105 remained free of dementia (⅙). The compared with 41.9% of controls (␹2 ϭ 6.7, p ϭ dashed lines indicate the group median values for A␤1–42 and A␤1–40 levels; 62.5% of those who developed AD ex- 0.009). In contrast, 59.4% of those who developed AD ceeded the group median A␤1–42 level compared with exceeded the median A␤1–40 level compared with 41.9% of those who remained disease-free (␹2 ϭ 6.7, p ϭ 44.3% of those who remained disease-free (␹2 ϭ 3.6, 0.009); 59.4% of those who developed AD exceeded the group p ϭ 0.06). The OR for AD associated with each 1 ␤ median A 1–40 level compared with 44.3% of those who pg/ml of A␤1–42 was significantly increased (OR ϭ ␹2 ϭ ϭ remained disease-free ( 3.6, p 0.06). 1.0114; 95% confidence interval, 1.003–1.019, p ϭ 0.006). In contrast, A␤1–40 plasma levels were not as- sociated with AD. The logistic regression analysis was AD compared with those who remained dementia-free repeated, dividing plasma A␤1–42 into quartiles based (mean and SEM values in Table 1). The group differ- on the entire sample. By using the lowest quartile as ence in mean A␤1–42 level, but not mean A␤1–40 the reference, the OR for AD associated with A␤1–42 level, remained statistically significant, after adjusting significantly increased by three- to fourfold at the third for age and education, ethnic group, and ApoE geno- and fourth quartiles (Table 2). This analysis was re- type, in the multivariate analysis of variance (F ϭ 9.7, peated, adjusting for age, education, the presence or p ϭ 0.002). The ratio of A␤1–42/A␤1–40 was also absence of an ApoE-ε4 allele, and a family history of higher (p ϭ 0.035) in those with AD compared with AD, with no observable changes in the strength of the those without dementia (see Table 1). We repeated association.

Table 2. Odds Ratio for Incipient Alzheimer’s Disease Associated with Plasma A␤1–42 by Quartiles

Proportion Who Developed Alzheimer’s Disease Crude Odds Ratio Adjusted Odds Ratio A␤1–42 Quartile Total Group at Risk (CDR 0.5 and 1.0) and 95% CI and 95% CIa

Ͻ24.7 pg/ml 42 8 (19%) 1.0 reference 1.0 reference 24.7–45.9 pg/ml 43 16 (37%) 2.5; 0.9–6.8 2.9; 0.9–8.9 45.9–85.0 pg/ml 42 19 (45%) 3.5; 1.3–9.4 3.6; 1.2–10.8 Ͼ85.0 pg/ml 42 21 (50%) 4.2; 1.6–11.3 4.0; 1.3–12.2 ϭ Total 169 64 (38%) ptrend 0.004 aThis analysis adjusted for age, education, the presence or absence of an apolipoprotein E-ε4 allele, and a family history of Alzheimer’s disease. ␤ ptrend indicates a statistically significant linear trend for the odds ratio with each higher quartile of plasma A 1–42. CDR ϭ Clinical Dementia Rating Scale; CI ϭ confidence interval.

414 Annals of Neurology Vol 46 No 3 September 1999 Discussion A␤1–42 levels occur before the onset of symptoms We report that plasma levels of the released A␤ pep- in some individuals, further studies should explore tide, A␤1–42, may be elevated several years before the whether plasma A␤1–42 levels can be used as a bio- onset of AD. Whether plasma levels are increased for a logical risk factor for late-onset, sporadic AD. longer period preceding the onset of AD or only tran- siently remains uncertain. It is also unclear whether Support was provided by Federal grants AG07232, AG10963, plasma A␤1–40 and A␤1–42 levels decline or remain AG08702, and RR00645, the Taub Foundation, the Charles S. increased with disease progression, although a recent Robertson Memorial Gift for Alzheimer’s Disease Research from the study13 suggested that A␤1–42 levels in cerebrospinal Banbury Fund, the Blanchette Hooker Rockefeller Foundation, and fluid (CSF) may decrease with disease progression. Our the New York State, Office of Mental Retardation and Develop- mental Disabilities. data indicate that plasma A␤ levels are stable in the early stages of disease. We thank Sam Gandy, MD, PhD, and Nicole Schupf, PhD, for The major advantages of our investigation were the critical review of the manuscript and assistance in the development of this project. careful documentation of normal cognitive function when the initial plasma was obtained, the longitudinal follow-up of all subjects, and the repeated determina- References ␤ 1. Iwatsubo T, Odaka A, Suzuki N, et al. Visualization of A beta tion of plasma A levels in a representative subset. A 42(43) and A beta 40 in senile plaques with end-specific A beta limitation is the lack of postmortem confirmation of monoclonals: evidence that an initially deposited species is A AD, but all cases were identified within 1 to 2 years of beta 42(43). Neuron 1994;13:45–53 disease onset when death is unlikely. Nonetheless, our 2. Iwatsubo T, Mann DM, Odaka A, et al. Amyloid beta protein ␤ ␤ ␤ results lend strong support for the hypothesis proposed (A ) deposition: A 42(43) precedes A 40 in Down syndrome. Ann Neurol 1995;37:294–299 by Scheuner and associates,6 that “individuals who ␤ 3. Borchelt DR, Thinakaran G, Eckman CB, et al. Familial Alz- have elevated plasma A 1–42(43) are, in fact, at heimer’s disease-linked presenilin 1 variants elevated A␤1-42/ greater risk of developing AD.” 1-40 ration in vitro and in vivo. Neuron 1996;17:1005–1013 Mutations and polymorphisms in genes associated 4. Duff K, Eckman C, Zehr C, et al. Increased amyloid-␤42(43) with AD interact with A␤. APP and PS1 or PS2 genes in brains of mice expressing mutant presenilin 1. Nature 1996; ␤ 383:710–713 result in early-onset familial AD and alter A metabo- 5. Citron M, Westaway D, Xia W, et al. Mutant presenilins of 3,5–7 ε lism. The 4 polymorphism of the ApoE gene in- Alzheimer’s disease increase production of 42-residue amyloid creases susceptibility to familial and sporadic AD later ␤-protein in both transfected cells and transgenic mice. Nat in life, and recent work with transgenic mice also in- Med 1997;3:67–72 dicates a complex interaction with A␤.14 Transgenic 6. Scheuner D, Eckman C, Jensen M, et al. Secreted amyloid ␤-protein similar to that in the senile plaques of Alzheimer’s mice with either APP, PS1, or PS2 mutations produce ␤ ␤ disease is increased in vivo by the presenilin 1 and 2 and APP increased levels of A 1–42 compared with A 1– mutations linked to familial Alzheimer’s disease. Nat Med 40.3–5 PS1 and PS2 proteins enhance proteolysis of 1996;2:864–869 ␤ ␥ 7. Kosaka T, Imagawa M, Seki K, et al. The beta APP717 Alzhei- -amyloid precursor by the more amyloidogenic 42- secretase pathway, resulting in cleavage of the A␤ pep- mer mutation increases the percentage of plasma amyloid beta 15 16 protein ending at A beta42(43). Neurology 1997;48:741–745 tide at residue 42/43. Younkin and colleagues have 8. Stern Y, Andrews H, Pittman J, et al. Diagnosis of dementia in found mutations in the coding regions of PS1/2 in a heterogeneous population: development of a neuropsycholog- four families with late-onset AD who had elevated ical paradigm-based diagnosis of dementia and quantified cor- plasma A␤1–42. These studies indicate that elevated rection for the effects of education. Arch Neurol 1992;49: plasma levels of A␤1–42 in the elderly could have a 453–460 9. McKhann G, Drachman D, Folstein M, et al. Clinical diagno- genetic basis. ␤ sis of Alzheimer’s disease: report of NINCDS-ADRDA work Although cerebral deposition of A 1–42 is unlikely group under the auspices of the Department of Health and Hu- to result directly from increased plasma A␤1–42, ele- man Services task force on Alzheimer’s disease. Neurology vated plasma levels probably do reflect a generalized 1984;34:939–944 alteration in A␤ in both neural and nonneural cells.6 10. Hughes CP, Berg L, Danziger W, et al. A new clinical scale for ␤ the staging of dementia. Br J Psychiatry 1982;140:566–572 Presumably, A 1–42 levels gradually decrease in the 11. Mehta PD, Dalton AJ, Mehta SP, et al. Increased plasma amy- CSF in AD because of the preferential sequestration as loid ␤ protein 1–42 levels in Down syndrome. Neurosci Lett insoluble deposits in brain.13,17 Once the deposition 1998;241:13–16 process begins, CSF levels may decline below that of 12. Potempska A, Mack K, Mehta PD, et al. Quantification of sub- ␤ healthy older individuals. Changes in extracellular A␤ femtomole amounts of Alzheimer amyloid peptides. Amyloid 1999;6:14–21 levels in plasma and CSF appear to occur with onset 13. Jensen M, Schro¨der J, Blomberg M, et al. Cerebrospinal fluid or progression of AD, limiting their usefulness as diag- A␤42 is increased early in sporadic Alzheimer’s disease and de- nostic markers. However, because elevated plasma clines with disease progression. Ann Neurol 1999;45:504–511

Brief Communication: Mayeux et al: A␤1–42 in Alzheimer’s Disease 415 14. Bales KR, Verina T, Dodel RC, et al. Lack of apolipoprotein E mon double innervation of the posterior ampulla by two dramatically reduces amyloid beta-peptide deposition. Nat nerves running in two separate bony canals could offer Genet 1997;17:263–264 an alternative explanation for the regular sparing of pos- 15. Xia W, Zhang J, Kholodenko D, et al. Enhanced production and oligomerization of the 42-residue amyloid ␤-protein by terior canal function in vestibular neuritis. Chinese hamster ovary cells stably expressing mutant preseni- Arbusow V, Schulz P, Strupp M, lins. J Biol Chem 1997;272:7977–7982 16. Younkin SG, Younkin LH, Kawarabayashi T, et al. Role of Dieterich M, von Reinhardstoettner A, Rauch E, increased amyloid ␤ protein (A␤) in Alzheimer’s disease. Neu- Brandt T. Distribution of herpes simplex virus robiol Aging 1998;19(4S):S2 (Abstract) type 1 in human geniculate and vestibular 17. Motter R, Vigo-Pelfrey C, Kholodenko D, et al. Reduction of ganglia: implications for vestibular neuritis. ␤ -amyloid peptide42 in the cerebrospinal fluid of patients with Ann Neurol 1999;46:416–419 Alzheimer’s disease. Ann Neurol 1995;38:643–648

Vestibular neuritis (VN), also known as acute unilat- eral (idiopathic) vestibular paralysis, is the third most Distribution of Herpes common cause of peripheral vestibular vertigo.1 The chief symptom is the acute/subacute onset of pro- Simplex Virus Type 1 longed severe rotational vertigo, associated with spon- in Human Geniculate taneous horizontal rotatory nystagmus, postural im- balance, and nausea without concomitant auditory and Vestibular Ganglia: dysfunction.2 The epidemic occurrence of the condi- tion,3 the frequency of preceding upper respiratory Implications for tract infections,4 and a small number of postmortem Vestibular Neuritis studies5,6 indicate that a viral cause of VN is very likely. The demonstration of latent herpes simplex V. Arbusow, MD,* P. Schulz, PhD,* M. Strupp, MD,* M. Dieterich, MD,* A. von Reinhardstoettner,* virus type 1 (HSV-1) in human vestibular ganglia E. Rauch, MD,† and T. Brandt, MD, FRCP* (VG)7,8 and the induction of VN by inoculation of HSV-1 in the auricle of mice9 also suggest that, as in Bell’s palsy and sensorineural hearing loss,10,11 the Vestibular neuritis is a common cause of partial unilat- cause of the disease might be the reactivation of latent eral vestibular paralysis, which usually spares posterior HSV-1 in the VG. It is most likely that VN is a partial semicircular canal function. The cause is assumed to be a rather than a complete paresis, which predominantly viral reactivation of latent herpes simplex virus type 1 (HSV-1) in human vestibular ganglia. The existence of an involves the anterior and horizontal semicircular canals anastomosis between the intermediate nerve and the su- (SCCs) (usually sparing posterior SCC function) and 12,13 perior vestibular nerve suggests the question of whether the utricle. This view is also supported by the si- selective affliction of the superior vestibular nerve is the multaneous occurrence of benign peripheral positional result of migration of HSV-1 from the geniculate gan- vertigo12 and the three-dimensional properties of the glion along this faciovestibular anastomosis. We deter- vestibulo-ocular reflexes in patients suffering from VN.13 mined the distribution of HSV-1 among geniculate gan- Anatomical studies have shown an anastomosis be- glia, vestibular ganglia, and within Scarpa’s ganglion by examining 35 human temporal bones by polymerase tween the intermediate and the superior portion of the chain reaction. HSV-1 was found in 66% of geniculate VN formed by axons of bipolar VNs (faciovestibular 14–16 ganglia and 60% of vestibular ganglia; all examined parts anastomosis). After primary viral infection (stoma- of vestibular ganglia were almost equally HSV-1 infected. titis herpetica), HSV-1 can reach the ganglion geniculi Our data provided no support for viral migration along (GG) via the chorda tympani (eg, by retrograde axonal this anastomosis or for a preferential latency of HSV-1 in transport). From the GG it is possible that HSV-1 mi- the superior vestibular nerve. We suggest that the com- grates to the superior portion of the VG along these crossing axons (Fig 1). Reactivation of HSV-1 would then cause the selective inflammation of the superior From the *Department of Neurology, Klinikum Großhadern, and VN and the dysfunction of the anterior and horizontal †Institute of Forensic Medicine, Ludwig-Maximilians University of SCCs typical for VN. Munich, Munich, Germany. To determine HSV-1 distribution between the GG, Received Mar 2, 1999, and in revised form May 14. Accepted for VG, and within the different parts of the VG, 35 hu- publication May 16, 1999. man temporal bones were examined for HSV-1– Address correspondence to Dr Arbusow, Department of Neurology, University of Munich, Klinikum Grosshadern, Marchioninistrasse specific DNA by nested polymerase chain reaction 15, D-81377 Munich, Germany. (PCR).

416 Copyright © 1999 by the American Neurological Association Materials and Methods Table. Frequencies of Single and Combined Herpes Simplex The study was performed with the consent of the Ethics Virus Type 1 Infections of Human Geniculate and Different Committee of the Medical Faculty of the Ludwig-Maxi- Parts of Vestibular Ganglia milians University of Munich. Thirty-five temporal bones Stem of the Superior Inferior were obtained from the Institute of Forensic Medicine of the Geniculate Vestibular Vestibular Vestibular Frequency of Ludwig-Maximilians University of Munich at the time of au- Ganglion Ganglion Nerve Nerve Combination topsy. The cause of death was unrelated to cranial nerve dys- ϩϪ ϪϪ 8 function in all cases. Age distribution ranged from 3 months ϩϩ ϩϩ 1 to 74 years with a median of 44 years (13 males and 5 fe- ϩϩ ϩϪ 1 males). GG and VG were prepared free of bone and connec- ϩϩ Ϫϩ 3 ϩϪ ϩϪ 5 tive tissue, immediately embedded in OTC medium ϩϩ ϪϪ 3 (TissueTek, Elkhart, IN), and cryosectioned. Slides (30 ␮m) ϩϪ Ϫϩ 2 were stained with cresyl violet and separated with a micro- Ϫϩ ϪϪ 2 ϪϪ ϩϪ stilette into three parts under an inverted microscope—the 1 ϪϪ Ϫϩ 1 stem of the ganglion and the proximal most parts of superior Ϫϩ ϩϪ 1 and inferior portions of the vestibular nerve, containing neu- Ϫϩ Ϫϩ 1 rons of the superior and inferior portions of the VG. ϪϪ ϪϪ 6 To extract DNA, nerve tissue was immersed in a neutral phenol/water mixture with 10 ␮g salmon DNA, Vortex mixed vigorously, and centrifuged. The aqueous phase was tion and establishment of latency in GG and VG carefully recovered, and nucleic acids were precipitated with seems to occur early in life. ethanol. The pellet was washed twice with ethanol, air-dried for 10 minutes, and redissolved in 20 ␮l of pure water. Nested PCR for HSV-1 was performed according to the Discussion method described by Aurelius and colleagues.17 All reagents Viral reactivation is the most likely cause of sensorineu- 8–11 and equipment were from Biometra, Go¨ttingen, Germany. ral hearing loss, Bell’s palsy, and VN. Although The temperature profile was 30 seconds at 95°C, 30 seconds multiple neighboring cranial nerve ganglia may be in- at 60°C, and 1 minute at 72°C. PCR products were analyzed fected, experience shows that clinical manifestation of on 3% agarose gels. PCR fragments of the correct size were viral reactivation is restricted to single nerves. The VG HindIII cleaved; they yielded the expected 110-bp fragment. has three portions—the stem and the (ganglion cells in To exclude false positives, all positive reactions were repeated the) proximal superior and inferior vestibular nerves at least once. A suspension of boiled human HSV-1, kindly (see Fig 1). As in the innervation of the cochlea, ana- donated by Dr G. Ja¨ger, Institute of Medical Microbiology, tomical studies of Lorente de No´19 and Alexander20 Ludwig-Maximilians University of Munich, was used as a template to optimize the PCR protocol. Thereafter, positive showed a tonotopic organization of the VG. Bipolar controls were omitted to minimize the contamination risk. neurons of the superior portion form the superior ves- To confirm the presence of amplifiable DNA in the nu- tibular nerve innervating the anterior and horizontal cleic acid preparations, a nonpolymorphic mitochondrial SCC, the utricle, and parts of the saccule. Ganglion fragment of 530 bp was demonstrated in parallel PCR cells of the inferior portion of the VG form the inferior reactions.18 vestibular nerve, which receives signals from parts of the saccule and the posterior SCC. On the basis of this anatomy the following three different types of VN are Results theoretically possible: (1) inflammation of the entire In 29 of 35 human temporal bones, HSV-1 was found VG (stem), which causes complete loss of unilateral in at least one of the investigated ganglia (Table). vestibular dysfunction with ocular tilt reaction and ro- HSV-1–specific PCR8,17 regularly yielded the correct tatory spontaneous nystagmus toward the nonaffected HSV-1–specific product with a length of 138 bp or no ear; this has been described for herpes zoster oticus21; product at all. The presence of amplifiable DNA in the (2) inferior vestibular nerve inflammation, which nucleic acid preparations was proven by amplification causes complete loss of posterior canal and partial loss of a nonpolymorphic mitochondrial fragment of 530 of saccular function; this, to our knowledge, has not bp in parallel PCR.18 yet been described; and (3) superior vestibular nerve The GG was infected in 66% (23 of 35), VG in inflammation, which causes complete paresis of the an- 60% (21 of 35) of cases. Distribution of HSV-1 terior and horizontal canals, the utricle, and partial pa- among the VG was as follows: stem only, 5 of 21 resis of the saccule; this is the typical clinical pattern (24%); superior portion only, 6 of 21 (29%); inferior of VN.12 portion only, 3 of 21 (14%); combined infections of Because in 6 of 35 (17%) temporal bones HSV-1 two or three examined parts, 7 of 21 (33%) (see Ta- infection of the VG was not associated with HSV-1 ble). As HSV-1 was already latent in temporal ganglia latency in the GG and HSV-1 was not found to be of 3-month-old and 4-year-old children, HSV-1 infec- predominantly latent in the superior portion of the

Brief Communication: Arbusow et al: HSV-1 in Human Vestibular Ganglia 417 Fig 1. (Left) Schematic drawing of geniculate ganglion, vestibular ganglion, and the superior and inferior vestibular nerves. A fa- ciovestibular anastomosis connects the facial (intermediate) nerve with the superior vestibular nerve. Bipolar neurons of the vestibu- lar ganglion extend into the proximal superior and inferior vestibular nerves. An accessory posterior ampullary nerve branches off the posterior ampullary nerve. Sections of the vestibular ganglion showing the preparation for separation into stem (a), inferior portion (b), and superior portion of the vestibular ganglion (c). (Right) Longitudinal cryosection (30 ␮m) of a human vestibular ganglion stained with cresyl violet. The proximal-most parts of the superior (c) and the inferior (b) vestibular nerves containing neurons of the vestibular ganglion were separated from the ganglion stem (a) by transversal cuts and analyzed individually for herpes simplex virus type 1 latency. SCC ϭ semicircular canal.

VG, our hypothesis of viral migration along the facio- vestibular anastomosis to the ganglion cells of the su- perior vestibular nerve could not be confirmed (see Ta- ble). Therefore, the regular preservation of posterior canal function in VN requires an alternative explana- tion. This could be offered by an anatomical variant of the innervation of the posterior SCC. Anatomical studies in vertebrates with sections through the posterior ampulla revealed a small vestib- ular end organ in the floor of the posterior recess of the utricle, the so-called crista neglecta,22 which is inner- vated by a small accessory nerve that originates from the posterior ampullary nerve.14,22 This accessory cup- ula organ is rarely detected in humans,22 but neverthe- less, the posterior SCC is still innervated by two dis- tinct nerves (see Fig 1) that reach the posterior cupula in two separate bony canals. Montandon and co- workers22 have demonstrated this in more than 600

Š Fig 2. (A) Microscopic (ϫ15) preparation of a human vestib- ular ganglion and nerves. G ϭ vestibular ganglion; ✻ ϭ superior vestibular nerve;  ϭ saccular nerve; ● ϭ posterior ampullary nerve running in a separate bony canal (arrow). (B) Magnification of the posterior ampullary nerve (ϫ40) revealing the common accessory posterior ampullary nerve. ● ϭ posterior ampullary nerve; f ϭ accessory posterior ampullary nerve.

418 Annals of Neurology Vol 46 No 3 September 1999 human temporal bone specimens.18 In our temporal of herpes simplex encephalitis by nested polymerase chain reac- bone preparations, 27 of 35 (77%) exhibited double tion assay of cerebrospinal fluid. Lancet 1991;337:189–192 innervation of the posterior ampulla (Fig 2). This dou- 18. Mann VM, Cooper JM, Schapira AH. Quantitation of a mito- chondrial DNA deletion in Parkinson’s disease. FEBS Lett ble innervation by two ampullary nerves running in 1992;299:218–222 two separate bony canals provides two plausible expla- 19. Lorente de No´, R. E´tudes sur l’anatomie et la physiologie du nations for the sparing of posterior SCC function in labyrinthe de l’oreille et du VIIIe nerf. Deuxieme partie. VN: (1) if the neuronal pools of both branches were Quelques donne´es au sujet de l’anatomie des organes sensoriels located separately in the VG, one pool might be spared du labyrinthe. Trav Lab Rech Biol Univ Madrid 1926;24:53– 153 in viral inflammation, and (2) that both posterior am- 20. Alexander G. Zur Anatomie des Ganglion Vestibulare. Sitzungs- pullary nerves run through separate bony canals sug- berichte kais Akad Wiss Math Naturw Klasse 1899;108:449 gests that they are less affected by inflammatory swell- 21. Arbusow V, Dieterich M, Strupp M, et al. Herpes zoster neuritis ing of the perineural tissue of the vestibular nerve, involving superior and inferior parts of the vestibular nerve causes which is encapsulated within the intrameatal duct. ocular tilt reaction. Neuro-Ophthalmology 1998;19:17–22 22. Montandon P, Gacek RR, Kimura RS. Crista neglecta in the cat and human. Ann Otol Rhinol Laryngol 1970;79:105–112 This study was supported by the Alfried Krupp von Bohlen und Halbach Foundation, Essen, Germany. We thank K. Arbusow for preparing Figure 1 and J. Benson for copyediting the manuscript. Clinicopathological Findings References following Intraventricular 1. Brandt T. Vertigo: its multisensory syndromes, 2nd ed. London: Springer, 1999 Glial-Derived Neurotrophic 2. Strupp M, Brandt T. Vestibular neuritis. Advances in otorhi- nolaryngology, vol 55. Basel: Karger, 1999:111–136 Factor Treatment 3. Sekitani T, Imate Y, Noguchi T, Inokuma T. Vestibular neuronitis: epidemiological survey by questionnaire in Japan. in a Patient with Acta Otolaryngol Suppl (Stockh) 1993;503:9–12 4. Hara H, Sekitani T, Imate Y, et al. Vestibular neuronitis in aged patients: results from an epidemiological survey by ques- Parkinson’s Disease tionnaire in Japan. Acta Otolaryngol Suppl (Stockh) 1993;503: Jeffrey H. Kordower, PhD,* Stephane Palfi, MD, PhD,* 53–56 Er-Yun Chen, MD,* Shuang Y. Ma, MD, PhD,* 5. Schuknecht HF, Kitamura K. Vestibular neuritis. Ann Otol Timothy Sendera, PhD,* Elizabeth J. Cochran, MD,† Rhinol Laryngol Suppl 1981;90:1–19 Elliott J. Mufson, PhD,* Richard Penn, MD,‡ 6. Baloh RW, Ishyama A, Wackym PA, Honrubia V. Vestibular Christopher G. Goetz, MD,* neuritis: clinical-pathologic correlation. Otolaryngol Head Neck and Cynthia D. Comella, MD* Surg 1996;114:586–592 7. Furuta Y, Takasu T, Fukuda S, et al. Latent herpes simplex virus type 1 in human vestibular ganglia. Acta Otolaryngol As part of a safety and tolerability study, a 65-year-old Suppl (Stockh) 1993;503:85–89 man with Parkinson’s disease (PD) received monthly in- 8. Schulz P, Arbusow V, Strupp M, et al. Highly variable distri- bution of HSV-1-specific DNA in human geniculate, vestibu- tracerebroventricular injections of glial-derived neurotro- lar, and spiral ganglia. Neurosci Lett 1998;252:139–142 phic factor (GDNF). His parkinsonism continued to 9. Hirata Y, Gyo K, Yanagihara N. Herpetic vestibular neuritis: an worsen following intracerebroventricular GDNF treat- animal study. Acta Otolaryngol Suppl (Stockh) 1995;519:93–96 ment. Side effects included nausea, loss of appetite, tin- 10. Murakami S, Mizobuchi M, Nakashiro Y, et al. Bell palsy and gling, L’hermitte’s sign, intermittent hallucinations, de- herpes simplex virus: identification of viral DNA in endoneurial pression, and inappropriate sexual conduct. There was no fluid and muscle. Ann Intern Med 1996;124:27–30 evidence of significant regeneration of nigrostriatal neu- 11. Welleschik B, Rasinger GA, Brunner E. Does pure-tone audi- rons or intraparenchymal diffusion of the intracerebro- ometry provide inferences for a vascular cause of sudden deaf- ventricular GDNF to relevant brain regions. Alternative ness? HNO 1987;35:119–127 GDNF delivery systems should be explored. 12. Bu¨chele W, Brandt T. Vestibular neuritis—a horizontal semi- circular canal paresis? Adv Otorhinolaryngol 1988;42:157–161 13. Fetter M, Dichgans J. Vestibular neuritis spares the inferior di- vision of the vestibular nerve. Brain 1996;119:755–763 14. Bergstro¨m B. Morphology of the vestibular nerve. II. The num- From the Departments of *Neurological Sciences, †Pathology, and ber of myelinated vestibular nerve fibers in man at various ages. ‡Neurosurgery, and Research Center for Brain Repair, Rush- Presbyterian-St Luke’s Medical Center, Chicago, IL. Acta Otolaryngol (Stockh) 1973;76:173–179 15. Oort H. U¨ ber die Vera¨stelung des Nervus octavus bei Sa¨uget- Received Feb 16, 1999, and in revised form May 19. Accepted for ieren. Anat Anz 1918;51:272–280 publication May 19, 1999. 16. Retzius G. Zur Kenntnis des inneren Geho¨rorgans der Wirbel- Address correspondence to Dr Kordower, Department of Neurolog- tiere. Arch Anat Physiol Anat Abt Leipzig 1880;235–244 ical Sciences, Rush-Presbyterian-St Luke’s Medical Center, 2242 17. Aurelius E, Johansson B, Sko¨ldenberg B, et al. Rapid diagnosis West Harrison Street, Chicago IL 60612.

Copyright © 1999 by the American Neurological Association 419 Kordower JH, Palfi S, Chen E-Y, Ma SY, traventricular infusion of human recombinant GDNF for 3 Sendera T, Cochran EJ, Mufson EJ, Penn R, months at a dose of 7.5 ␮g/day. One of these animals dis- Goetz CG, Comella CD. Clinicopathological connected his catheter 2 or 3 days before perfusion. findings following intraventricular glial-derived neurotrophic factor treatment in a patient Preparation of Tissues with Parkinson’s disease. Frozen sections through the forebrain and/or midbrain were Ann Neurol 1999;46:419–424 processed for the visualization of tyrosine hydroxylase (TH), dopamine transporter (DAT), and GDNF immunoreactivity as well as for in situ hybridization for TH mRNA as previ- The concept that glial-derived neurotrophic factor ously reported.2–4 (GDNF) treatment may prevent or reverse structural and functional consequences of nigrostriatal degenera- Results tion in Parkinson’s disease (PD) emerged from exten- GDNF Case: Clinical/Pathological Findings sive in vitro and in vivo preclinical studies.1 A multi- The patient’s parkinsonism continued to worsen dur- center clinical trial was initiated to assess the safety and ing the course of GDNF treatment as illustrated by tolerability of this trophic factor in patients with idio- increased motor scores on the Unified Parkinson’s Dis- pathic PD. This report represents a clinicopathological ease Rating Scale during on (baseline ϭ 37, last visit ϭ analysis of the first patient from this clinical trial who 48) and off (baseline, 71; last visit, 76). Unified Par- has come to autopsy. kinson’s Disease Rating Scale Activity of Daily Living scores also increased during on (baseline, 22; last visit, Materials and Methods 38) and off (baseline ϭ 36; last visit ϭ 43). Prior to Case History entry into the GDNF trial, this patient was treated DN was a 65-year-old man with a 23-year history of PD. He daily with 950 mg of levodopa, 21.25 mg of bro- initially had excellent improvement with levodopa treatment mocriptine, and 200 mg of amantadine. Just prior to but began to experience severe motor fluctuations and pro- death, his daily medications were 900 mg of levodopa, gression of his PD symptoms. In November 1996, he en- 2 mg of ropinerole, and 37.5 mg of clozapine. His tered a double-blind placebo-controlled study of GDNF score on the Mini-Mental State Examination was therapy and was implanted with a catheter (Metronics, Min- 26/30 at the beginning of the open-labeled phase and neapolis, MN) into the right lateral ventricle. At this time, remained stable. During the controlled phase of the he was Hoehn and Yahr stage III in “on” and stage V in study, the patient developed visual hallucinations “off.” While in the double-blind phase of this trial, he re- ceived either vehicle or monthly injections of 75 ␮gof which improved with the elimination of amantadine GDNF. On completion of the double-blinded stage, he en- and addition of clozapine. Following each injection in tered the open-labeled phase, receiving his first dose of 25 the open-labeled study, the patient experienced tingling ␮g of GDNF in July 1997. Subsequent treatments were ad- radiating down his spine which resolved in about 2 ministered as follows: 75 ␮g in October 1997, 150 ␮ginNo- weeks. He also displayed a loss of appetite, nausea, and vember 1997, and 300 ␮g in December 1997, January 1998, L’hermitte’s sign, all of which were temporally related and February 1998. He died 3 weeks following his final to the injection. During treatment, he exhibited inap- GDNF injection. A general autopsy showed severe coronary propriate sexual conduct and experienced depression artery disease and acute bronchopneumonia. Autolysis time and, again, hallucinations, which improved with cloza- was 6 hours, and the brain was fixed in 4% paraformaldehyde. pine treatment. Exacerbation of these latter behaviors was not temporally associated with each GDNF injec- Control Cases tion. Neuropathologically, this patient displayed Lewy Five age-matched control cases were examined for neuro- bodies and a loss of neurons accompanied by free neu- pathological and neuroanatomical comparison with the romelanin within the substantia nigra. GDNF-treated patient. Two cases were normal controls without clinical or pathological signs of PD or other neuro- logical disorders. Three additional age-matched controls were Dopaminergic Markers in Control Cases clinically and pathologically diagnosed with PD but never Within the midbrain, normal aged controls displayed received GDNF. These PD cases were matched with the intense TH immunoreactivity (Fig 1A) and DAT im- GDNF case for disease severity and autolysis time. munoreactivity within neurons distributed throughout the substantia nigra, ventral tegmental area, and their Nonhuman Primates associated pathways. TH mRNA expression within the In the GDNF-treated human case, we attempted to detect nigra of normal cases was robust (see Fig 1D). In con- infused GDNF protein immunohistochemically. As a posi- trast, non-GDNF–treated PD cases displayed a reduc- tive control for this experiment, we assessed the expression of tion in the density of TH-immunoreactive neurons infused GDNF in 3 monkeys that had been rendered par- within the nigra (see Fig 1C) and a reduction in TH- kinsonian with MPTP and subsequently received chronic in- immunoreactive fibers in the nigrostriatal pathway,

420 Annals of Neurology Vol 46 No 3 September 1999 Fig 1. (A–C) Low-power photomicrographs of tyrosine hydroxylase (TH)–immunoreactive stained sections through the ventral mesen- cephalon from a normal case (A), glial-derived neurotrophic factor (GDNF)–treated Parkinson’s disease (PD) case (B), and un- treated PD control case (C). Both the GDNF-treated and control PD cases displayed a significant loss of TH-immunoreactive ni- gral neurons relative to the control case. At this level, there appear to be a greater number of TH-immunoreactive neurons in the GDNF-treated case relative to the PD control case. The residual number of TH-immunoreactive nigral neurons is still within the normal range for this disease (see Discussion section for further details). (D–F) High-power photomicrographs through the substantia nigra sections hybridized for TH mRNA from a normal case (D), GDNF-treated PD case (E), and nontreated PD control case (F). Note the intense level of hybridization seen in the normal case. Both PD cases displayed many moderate and lightly labeled cells (arrows). The scale bar in C represents 700 ␮m for panels A through C. The scale bar in F represents 40 ␮m for panels D through F. Rn ϭ red nucleus. caudate nucleus, and putamen. In these cases, however, preferential innervation in TH-immunoreactive fibers there was still significant TH-immunoreactive staining proximal to the site of delivery, however. Within the within the medial caudate nucleus (Fig 2B). Moderate striatum, dense TH immunoreactivity was observed to light TH mRNA expression was observed in residual within the medial aspect of the caudate nucleus (see nigral neurons in the non–GDNF-treated PD cases Fig 2A). The breadth and intensity of TH immunore- (see Fig 1F). activity in the medial caudate nucleus was clearly di- minished relative to that in the normal aged controls Dopamine Markers in GDNF-Treated PD Case and similar to that seen in non–GDNF-treated PD Staining for TH but not DAT appeared slightly more controls. Bilaterally, there was a dramatic diminution intense in the caudate nucleus on the infused side rel- of TH staining in the mid to lateral caudate nucleus ative to the contralateral hemisphere. There was no (see Fig 2A). The staining pattern in the caudate was

Brief Communications: Kordower et al: GDNF and PD 421 Fig 2. Low-power photomacrographs of a tyrosine hydroxylase(TH)–immunoreactive stained section through the striatum from the glial-derived neurotrophic factor (GDNF)–treated case (A) and an untreated control Parkinson’s disease (PD) case (B). Note the similarity of TH immunoreactivity within the caudate nucleus and putamen in the 2 cases; both display dense TH immunoreactiv- ity within the medial caudate nucleus. In the bed nucleus of the stria terminalis and the rostral nucleus basalis/substantia innomi- nata complex, an enhanced network of TH immunoreactivity was observed in the GDNF-treated case (A) but was not seen in the control PD case (B). (C) This enhanced staining pattern within the bed nucleus of the stria terminalis and nucleus basalis/substan- tia innominata region was not observed with dopamine transporter immunoreactivity. (D) High-power photomacrograph illustrating the density of TH-immunoreactive fibers in the putamen of a normal case. (E) In contrast, the putamen in the GDNF-treated case displayed only sparse TH-immunoreactive fibers. AC ϭ anterior commissure. The scale bar in D represents 5.95 mm for panels A through C. The scale bar in E represents 70 ␮m for panels D and E. similar to that seen in all non-GDNF–treated PD cases servedin the GDNF-treated case relative to the normal (see Fig 2B). There was a dramatic diminution of TH- or PD controls in a contiguous region beginning in the immunoreactive staining within the putamen (see Fig bed nucleus of the stria terminalis, curving ventrally 2A and D) relative to that in controls (see Fig 2E). around the internal capsule, and continuing through- TH-immunoreactive staining in the putamen was lim- out the nucleus basalis/substantia innominata region. ited rostrally to a thin collection of fibers immediately This enhanced staining pattern was not seen with DAT lateral to the internal capsule and caudally to a thin immunohistochemistry, was only seen on the side of collection of fibers immediately lateral to the external the infusion, and was most prominent at the level of medullary lamina of the globus pallidus (see Fig 2A). the decussation of the anterior commissure. The rest of this structure exhibited scattered individual TH-immunoreactive stained sections throughout the fibers (see Fig 2D). TH immunoreactivity within the nigra revealed an extensive loss of DA neurons in the putamen in the GDNF case (see Fig 2A) was similar to GDNF-treated case (see Fig 1B) relative to normal that observed in all PD control cases (see Fig 2B). aged controls (see Fig 1A). Qualitatively, this loss was In contrast to what was observed in the striatum, preferentially distributed along the dorsolateral aspect enhanced TH immunoreactivity (see Fig 2A) was ob- of the ventral midbrain and appeared similar in mag-

422 Annals of Neurology Vol 46 No 3 September 1999 Fig 3. Low-power photomacrographs through the striatum of glial-derived neurotrophic factor (GDNF)–immunoreactive stained sections from two monkeys (A and B) and the GDNF-treated Parkinson’s disease case (C). GDNF immunoreactivity was observed in the monkeys within the caudate nucleus ipsilateral to the infusion and in the septum bilaterally (A) as well as within the overly- ing cortex (A and B). (C) In contrast, GDNF immunoreactivity was not observed in the GDNF-treated case even at the level of the infusion (arrowhead).LVϭ lateral ventricle; Cd ϭ caudate nucleus; Pu ϭ putamen. The scale bars in B and C represent 4.78 and 5.58 mm, respectively. nitude to what was observed in PD control cases. In cannula 2 to 3 days prior to sacrifice. Expression was the GDNF-treated case, a moderate number of TH- extremely limited, however. In one monkey, GDNF- immunoreactive stained neurons could be seen within immunoreactive staining was only seen within the me- the ventral tegmental area and medial substantia nigra dial rim of the caudate nucleus and the septum bilat- (see Fig 1B). Stereological counts within the substantia erally (see Fig 3A). In the remaining two monkeys, nigra of the GDNF-treated case revealed 106,512 TH- GDNF immunoreactivity was limited to the cingulate immunoreactive neurons on the right side and 89,356 gyrus and neocortex overlying the cannula tract (see TH-immunoreactive neurons on the left side. Fig 3B). TH mRNA in situ hybridization was seen within the residual nigral neurons in all cases. In the GDNF case, Discussion some cells displayed a modest intensity of labeling, The present report indicates that monthly, low-dose, although others were more lightly labeled (see Fig 1E). intraventricular GDNF treatment neither improves In general, the intensity of labeled cells was similar clinical parkinsonism nor induces dopaminergic regen- to that seen in non-GDNF–treated PD cases (see Fig eration in a patient with PD. GDNF injections were 1F) and less than what was seen in normal controls associated with a number of side effects, some of which (see Fig 1D). were temporally related to the injections (tingling, nau- sea, L’hermitte’s sign), although others were not (hal- GDNF Immunoreactivity in GDNF-Treated PD Case lucinations, inappropriate sexual conduct). No GDNF immunoreactivity was observed in the PD A preponderance of the morphological and anatom- case receiving GDNF treatment (Fig 3C). GDNF im- ical data argues against the possibility that GDNF in- munoreactivity was seen in all three GDNF-treated duced a significant trophic effect on nigrostriatal neu- monkeys, including the monkey that disconnected his rons in this patient. Dense TH-immunoreactive

Brief Communications: Kordower et al: GDNF and PD 423 staining was observed along the medial portion of the son’s disease: Pathophysiologic and clinical implications. caudate nucleus in this case, and the intensity of the N Engl J Med 1988;318:876–880 TH-immunoreactive staining appeared greater ipsilat- 7. Kojima H, Abiru Y, Sakajiri K, et al. Adenovirus-mediated trans- duction with human glial cell line–derived neurotrophic factor eral to the infusion compared with the contralateral gene prevents 1-methyl-4-phenyl-1,2,3,6-tetrahydropyridine– side. However, the intensity and breadth of TH immu- induced dopamine depletion in striatum of mouse brain. Biochem noreactivity within the caudate nucleus was similar to Biophys Res Commun 1997;238:569–573 what can be seen within the caudate nucleus of PD 8. Bjo¨rklund A, Rosenblad C, Winkler C, et al. Studies on neu- patients who have not received GDNF (eg, present roprotective and regenerative effects of GDNF in a partial 2,5,6 lesion model of Parkinson’s disease. Neurobiol Dis 1997;4: study). The postcommissural putamen appears to 186–200 be the most critical region for dopaminergic innerva- 9. Choi-Lundberg DL, Lin Q, Chang YN, et al. Dopaminergic tion in PD, because this region receives innervation neurons protected from degeneration by GDNF gene therapy. from the motor cortex, is involved in motor circuitry, Science 1997;275:838–841 and is the subregion of the striatum that displays the 10. Mandel RJ, Spratt SK, Snyder RO, et al. Midbrain injection of 5,6 recombinant adeno-associated virus encoding rat glial cell line– most comprehensive loss of striatal dopamine in PD. derived neurotrophic factor protects nigral neurons in a pro- There was negligible TH immunoreactivity within the gressive 6-hydroxydopamine–induced degeneration model of putamen in the GDNF-treated case. This indicates that Parkinson’s disease in rats. Proc Natl Acad Sci USA 1997;94: intracerebroventricular GDNF, at least in the dose and 14083–14088 manner presently administered, would not be predicted 11. Lindner MD, Winn SR, Baetge EE, et al. Implantation of en- capsulated catecholamine and GDNF-producing cells in rats to enhance motor function in PD. with unilateral dopamine depletions and parkinsonian symp- The absence of GDNF immunoreactivity in the toms. Exp Neurol 1995;132:62–76 GDNF-treated human case and extremely limited 12. Emerich DF, Plone M, Francis MJ, et al. Alleviation of behav- GDNF immunoreactivity in the primate cases indicate ioral deficits in aged rodents following implantation of en- that the intracerebroventricular route of delivery may capsulated GDNF-producing fibroblasts. Brain Res 1996;736: 99–110 suboptimal in primates. Successful cellular delivery of 13. Choi-Lundberg DL, Lin Q, Schallert T, et al. Behavioral and GDNF has been achieved in animal models of PD us- cellular protection of rat dopaminergic neurons by an adenovi- ing ex vivo and in vivo gene therapy approaches.7–15 ral vector encoding glial cell line–derived neurotrophic factor. With these site-specific novel delivery techniques, there Exp Neurol 1998;154:261–276 is still great potential for this trophic factor to power- 14. Bilang-Bleuel AF, Revah P, Colin I, et al. Intrastriatal injection of an adenoviral vector expressing glial-cell-line–derived neuro- fully modify nigrostriatal activity and clinical function trophic factor prevents dopaminergic neuron degeneration and in patients with PD. behavioral impairment in a rat model of Parkinson disease. Proc Natl Acad Sci USA 1997;94:8818–8823 15. Sautter J, Tseng JL, Braguglia D, et al. Implants of polymer- encapsulated genetically modified cells releasing glial cell We thank Theodora Kladis for expert technical assistance. We also line–derived neurotrophic factor improve survival, growth, and thank Drs Don Gash, Zhiming Zhang, and Richard Grondin for function of fetal dopaminergic grafts. Exp Neurol 1998;149: providing the primate tissue and Dr Allan Levey for providing the 230–236 DAT antibody. We thank Drs Gash and Levey for helpful com- ments as well.

References 1. Lapchak PA, Gash DM, Collins F, et al. Pharmacological ac- tivities of glial cell line–derived neurotrophic factor (GDNF): Preclinical development and application to the treatment of Parkinson’s disease. Exp Neurol 1997;145:309–321 2. Kordower JH, Rosenstein JM, Collier TJ, et al. Functional fetal nigral grafts in a patient with Parkinson’s disease: Chemoana- tomic, ultrastructural, and metabolic studies. J Comp Neurol 1995;370:203–230 3. Ciliax BJ, Heilman C, Demchyshyn LL, et al. The dopamine transporter: Immunochemical characterization and localization in brain. J Neurosci 1995;15:1714–1723 4. Kordower JH, Freeman TB, Snow BJ, et al. Neuropathological evidence of graft survival and striatal reinnervation after the transplantation of fetal mesencephalic tissue in a patient with Parkinson’s disease. N Engl J Med 1995;332:1118–1124 5. Fearnly J, Lees A. Pathology of Parkinson’s disease. In: Calne D, ed. Neurodegenerative diseases. Philadelphia: WB Saunders, 1994:545–554 6. Kish SJ, Shannak K, Hornykiewicz O. Uneven pattern of do- pamine loss in the striatum of patients with idiopathic Parkin-

424 Annals of Neurology Vol 46 No 3 September 1999 addition to the evolution of an individual MS lesion Analysis of Gene Expression from an acute inflammatory process to a chronic inac- in Multiple Sclerosis Lesions tive process, suggest a complex pattern of gene expres- sion within MS lesions. Using cDNA Microarrays Various methods such as differential display,1 in situ hybridization,2 and reverse transcriptase–polymerase Laurie Ward Whitney, PhD,* Kevin G. Becker, PhD,†¶ 3,4 Nancy J. Tresser, MD,* Carla I. Caballero-Ramos, BS,* chain reaction have been used to monitor differences Peter J. Munson, PhD,‡ Vinayakumar V. Prabhu, PhD,‡ in gene expression in MS; however, to date, only a Jeffrey M. Trent, PhD,† Henry F. McFarland, MD,* small number of genes that potentially contribute to and William E. Biddison, PhD* lesion pathology have been identified. Recently, cDNA microarrays have been developed containing thousands of cDNAs of known sequenced In multiple sclerosis (MS) patients, a coordinated attack human genes robotically printed on glass microscope of the immune system against the primary constituents of slides at high density.5–8 Microarrays allow a quantita- oligodendrocytes and/or the myelin sheath of oligoden- drocytes results in the formation of lesions in the brain tive parallel analysis of expression of thousands of genes and spinal cord. Thus far, however, a limited number of within a sample at the same time and under identical genes that potentially contribute to lesion pathology have conditions. In this study, cDNA microarrays were used been identified. Using cDNA microarray technology, we to examine large-scale changes in gene expression asso- have performed experiments on MS tissue monitoring the ciated with lesion pathology in MS. Gene expression in expression pattern of over 5,000 genes and compared the each MS lesion studied was compared with that in nor- gene expression profile of normal white matter with that mal white matter from the same human MS patient found in acute lesions from the brain of a single MS pa- brain. Results demonstrated that even when using very tient. Sixty-two differentially expressed genes were iden- small quantities of pathological human brain tissue, tified, including the Duffy chemokine receptor, inter- cDNA microarrays report large-scale gene expression feron regulatory factor-2, and tumor necrosis factor alpha changes that are highly reproducible and can be vali- receptor-2 among others. Thus, cDNA microarray tech- nology represents a powerful new tool for the identifica- dated by immunohistochemistry. tion of genes not previously associated with the MS dis- ease process. Materials and Methods The general methodology of arraying is based on the proce- Whitney LW, Becker KG, Tresser NJ, 5 Caballero-Ramos CI, Munson PJ, Prabhu VV, dures of DeRisi and co-workers. Briefly, glass microscope Trent JM, McFarland HF, Biddison WE. slides were treated for 1 hour with a 0.01% poly-L-lysine Analysis of gene expression in multiple sclerosis solution (Sigma, St Louis, MO) and subsequently washed in lesions using cDNA microarrays. water to remove excess poly-L-lysine. Purified polymerase 9 ϫ Ann Neurol 1999;46:425–428 chain reaction products resuspended in 3 saline sodium citrate (SSC) were robotically printed in small volumes (ap- proximately 5 nl) onto glass slides. Slides were then washed in 0.2% sodium dodecyl sulfate (SDS). Bound DNA was Multiple sclerosis (MS) is the primary demyelinating subsequently denatured in 95°C distilled water for 2 minutes disease of the central nervous system and is character- and washed with 95% ethanol. The DNA was then ultravi- ized by areas of inflammation and myelin damage, olet cross-linked to the slides using a Stratagene Stratalinker called lesions, in the brain and spinal cord. Genetic set at 60 mJ (Stratagene, La Jolla, CA). Slides were blocked influences documented through population studies, in by rinsing in a solution of 70 mM of succinic anhydride dissolved in 0.1 M of boric acid, pH 8.0, containing 35% 1-methyl-2-pyrrolidinone (Aldrich, Milwaukee, WI). The or- igin of cDNA clones for arraying has been described else- From the *Molecular Immunology Section, Neuroimmunology where.9 Arrays for this study contained either 1,344 or 5,000 Branch, National Institute of Neurological Disorders and Stroke, elements that were primarily derived from I.M.A.G.E. con- ‡Analytical Biostatistics Section, Mathematical and Statistical Com- sortium cDNA libraries.10,11 Each array contains a set of 88 puting Laboratory, Center for Information Technology, and †Can- cer Genetics Branch, National Human Genome Research Institute, housekeeping genes used for normalization of data when National Institutes of Health, Bethesda, MD. comparisons between arrays are made. ¶Present address: DNA Array Unit, National Institute on Aging, Brain tissue was obtained at autopsy (8 hours postmor- National Institutes of Health, Baltimore, MD. tem) from a 46-year-old male patient with primary progres- sive MS. A detailed analysis of the patient’s history has been Received Feb 26, 1999, and in revised form May 19. Accepted for 12 publication May 19, 1999. described previously. Research described in this study was performed in accordance with protocol 76-N-0021 approved Address correspondence to Dr Whitney, Neuroimmunology Branch, National Institute of Neurological Disorders and Stroke, National by the Institutional Review Board of the National Institute Institutes of Health, Building 10, Room 5B-16, 10 Center Drive, of Neurological Disorders and Stroke. Apparent plaques were Bethesda, MD 20892-1400. dissected; part of each plaque was fixed in 4% buffered for-

Brief Communication: Whitney et al: Gene Expression in MS Lesions Using cDNA Microarrays 425 malin and embedded in paraffin, and the remaining tissue mixtures at 0.5 and 0.6 ␮g/␮l, respectively. Hybridizations was quick-frozen on dry ice and stored at Ϫ70°C. Paraffin- were carried out at 42°C overnight in 150-␮l volumes. embedded sections were cut at 7 ␮m. Antibodies used for Washes for hybridizations were carried out three times at 55°C staging of lesions and assessment of inflammatory cells were in 0.01% SDS and 0.5ϫ SSC for 5 minutes each, followed KP1 (macrophage), CD8 (T cells), L26 (B cells), glial fibril- by a 5-minute room temperature wash in 0.06% SSC. lary acidic protein (astrocytes), and neurofilament triplet pro- Radioactive hybridizations were exposed on a phospho- tein (NFTP) (neurons) (all purchased from DAKO, Carpin- rimager (Molecular Dynamics Storm; Molecular Dynamics, teria, CA) as well as CD4 (T cells) (Novocastro, Newcastle- Sunnyvale, CA) at a resolution of 50 ␮m. DNA targets on upon-Tyne, UK) and S100 (oligodendrocytes and astrocytes) the arrays were located using grid overlays, and spot intensi- (Biogenex, San Ramon, CA). Staining with these antibodies ties were subsequently measured (ImageQuant; Molecular was performed on a Ventana 320 automated immunohisto- Dynamics). Spot intensities between arrays were normalized chemistry stainer (Ventana, Tuscon, AZ) using an indirect to the average intensities of a set of 88 housekeeping genes avidin-biotin method with 3,3Ј–diaminobenzidine tetrahy- present on each array. Relative levels of gene expression be- drochloride as a chromogen. Antigen retrieval consisted of tween arrays were calculated by dividing normalized intensi- microwaving specimens in the presence of 10 mM of citrate ties of spots on one array by normalized intensities of spots buffer, pH 6, with 0.1% Tween-20 (Fischer Scientific, Pitts- on a second array. Significance levels of gene expression burgh, PA). Antibodies were used at concentrations of 1:40 changes were based on a 99% CI.13 to 1:16,000. Total RNA was generated from samples using RNAzol B (Tel-Test, Friendswood, TX) as directed by the manufac- Results turer. cDNA probes were prepared from 5 to 10 ␮g of total Immunostaining and immunohistochemistry were per- RNA by oligo dT–primed polymerization using SuperScript formed to determine the general stage of each lesion. II reverse transcriptase (Life Technologies, Gaithersburg, Lesions were evaluated for myelin loss, myelin debris, 33 MD). P-dCTP was incorporated into the reverse transcrip- and the presence of macrophages, T cells, B cells, neu- tase reaction in the presence of 0.5 mM of the appropriate rons, and astrocytes; they were also evaluated indirectly additional cold dNTPs. Probes were purified by gel chroma- for oligodendrocytes (Table 1). The lesions studied tography (Biospin 6; BioRad, Hercules, CA) followed by 14 ethanol precipitation. were classified as acute (see Lassmann and colleagues ) Labeled probes were resuspended in a hybridization buffer because they contained macrophages with myelin de- containing 50% deionized formamide, 6ϫ SSC, 5ϫ Den- bris (“demyelinating”) and T cells (“inflammatory”) hardt’s solution, 1% SDS, and 0.25 ␮g/␮l of tRNA (Sigma). throughout. There was a complete absence of B and T Oligo dA (Pharmacia, Piscataway, NJ) and Cot1 DNA (Life cells and no activated microglial cells in the normal Technologies) were added as blocking agents to the reaction white matter used in these experiments.

Table 1. Pathological Characterization of Lesionsa

Sample NMW wx11 Lesion wx28 Lesion wx41 Stage Normal Acute Acute Pathology None Inflammation/Demyelinationb Inflammation/Demyelinationb

Staging markers H & E NWM Lesion present Lesion present Ϫ ϩ Ϫ ϩ MDP NWM LFB /PAS LFB /PAS GFAP NWM ŒŒ ŒŒ S100 NWM NWM NWM LFB NWM   NFTP NWM  Inflammatory cell markers KP1 None Many Many L26 None Few None CD4 None Few Few CD8 None Many Many aAll lesions staged according to Lassmann and co-workers.14 bInflammation and/or demyelination throughout lesion indicates an acute lesion. H&Eϭ hemotoxylin and eosin; MDP ϭ macrophage degradation products; LFB/PAS ϭ luxol fast blue/periodic acid Schiff (early myelin breakdown/late myelin breakdown products in acute lesion macrophage); GFAP ϭ glial fibrillary acidic protein (activated astrocytes); S100 (oligodendrocytes and astrocytes); LFB (myelin); NFTP ϭ neurofilament triplet protein (neurons); KP1 (macrophage); L26 (B cells); CD4 (CD4ϩ T cells); CD8 (CD8ϩ T cells); NWM ϭ no change from normal white matter; Œ ϭ barely visible increase in staining intensity; ŒŒ ϭ easily visible increase in staining intensity;  ϭ barely visible loss in staining intensity;  ϭ easily visible loss in staining intensity;  ϭ nearly total loss of staining intensity; none ϭ none present or no change compared with NWM; few ϭ sparse increase compared with NWM; many ϭ moderate increase compared with NWM.

426 Annals of Neurology Vol 46 No 3 September 1999 Each radioactive probe sample was hybridized under MS lesions with those found in normal white matter. standard conditions (see Materials and Methods sec- Results identified many genes potentially involved in tion) to separate but identical cDNA microarrays ar- plaque pathology that have not previously been associ- rayed at high density on glass slides. Hybridizations ated with the MS disease state. For instance, the Duffy were performed using arrays containing both 1,400 chemokine receptor was identified and has recently and 5,000 genes. Table 2 lists genes that were overex- been shown to be a promiscuous chemokine receptor pressed in more than one experiment in lesions relative expressed in brain as well as on endothelial cells.15,16 to normal white matter. Significance levels of gene ex- Further tests are required to determine exactly what pression changes listed, generally found to be a ratio of role, if any, these genes play in lesion pathology. twofold or greater, were based on a 99% CI.13 Interar- The results of these microarray experiments, the ex- ray intensities in expression of a given cDNA for a given istence of multiple genetic susceptibility loci for MS,17 plaque type were highly reproducible (p Ͻ 0.005, Stu- and the heterogenous pathology of plaques18 empha- dent’s t test), suggesting that differences in gene expres- size the complex genetic interactions occurring in MS sion between experiments were not a result of differences plaque formation. In these experiments, the expression inherent in the microarrays. Identified changes in gene changes recognized in plaques relative to normal white expression were reproducible and were validated using matter may reflect a loss resulting from the disease pro- immunohistochemistry for several of the genes listed cesses or influx, respectively, of cells in the plaque (data not shown). Overall, changes in gene expression rather than a direct decrease or increase in overall gene were consistent with altered cell metabolism/structure expression in a consistent cell population. Clearly, sec- and a shift in cytokine and cell adhesion molecule ex- ondary analysis of gene expression changes via immu- pression indicative of inflammatory processes. nohistochemistry or in situ hybridization is required to identify the cell types in which expression changes are Discussion occurring. Comparison of results presented here with The study presented here uses cDNA microarrays to similar data derived from other MS patient lesions is compare large-scale gene expression patterns of acute necessary and will provide a novel mechanism to iden-

Table 2. Changes in Gene Expression Associated with Plaque Formation: High Expression in Acute Plaques Relative to Normal White Matter

Gene Name (Unigene Identification Number) Gene Name (Unigene Identification Number)

Adhesion/Structure/Transport Growth Factor–Related

ICAM-2 (131854) Activin type II receptor (52338) GAP-43 (79000) ␣-2–chimaerin (75092)

Myelin Formation Signaling

EST-myelin transcription factor-1 (79245)a Retinoic acid receptor ␣-1 (155389)

Cell Cycle/Homeostasis/Unknown Immune-Related

EF-1 ␦ elongation factor (90319) IRF-2 (83795) Neuron-specific enolase (75675) 56-K autoantigen annexin XI (75510) Proton ATPase (7476) High endothelial venule (75445) Glutaminyl-tRNA synthetase (79322) Leukotriene A-4 hydrolase (81118) Na,K-ATPase ␤ subunit (78629) Transcription factor E2A (101047) Neuroendocrine-specific protein A (99947) TNF␣ receptor-2 (75275) Alpha-N-acetylgalactosaminidase (75372) Duffy chemokine receptor (183) Liver type alkaline phosphatase (117854) Male-enhanced antigen (155638) Mitochondrial enoyl-CoA hydratase (75860) Import precursor-ATP synthase ␣ (155101) Serine dehydratase (76751) Adenine nucleotide translocator-2 (79172) Ornithine decarboxylase antienzyme (71303) Protein kinase C type ␤-1 (6749) Cell surface glycoprotein A15 (82749) aThe inclusion of EST in front of a gene name indicates that the gene is highly similar but not identical in sequence to the named gene.

Brief Communication: Whitney et al: Gene Expression in MS Lesions Using cDNA Microarrays 427 tify gene changes occurring in individual plaque types, 17. Becker KG, Simon RM, Bailey-Wilson JE, et al. Clustering of indicating how variable MS pathology is over the non-MHC susceptibility candidate loci in human autoimmune course of the disease and across individuals. diseases. Proc Natl Acad Sci USA 1998;95:9979–9984 18. Lucchinetti CF, Bruck W, Rodriquez M, Lassmann H. Distinct patterns of multiple sclerosis pathology indicates heterogeneity in pathogenesis. Brain Pathol 1996;6:259–274 This work was supported in part by a postdoctoral fellowship from the National Multiple Sclerosis Society (L.W.W.). We thank Drs Michael Bittner and Yidong Chen (NIH, Bethesda, MD) for help with microarray printing and data analysis. We also The Neurological thank Gerald Gooden (NIH) for technical assistance and Drs James Powers (University of Rochester, Rochester, NY) and David Matt- Manifestations of Nipah son (Indiana University, Indianapolis, IN) for MS tissue and pathol- ogy support. Virus Encephalitis,

References a Novel Paramyxovirus 1. Grekova MC, Robinson ED, Faerber MA, et al. Deficient ex- Kim-En Lee, MRCP(UK),* T. Umapathi, MRCP(UK),* pression in multiple sclerosis of the inhibitory transcription factor Chai-Beng Tan, MMed (Int Med),* Sp3 in mononuclear blood cells. Ann Neurol 1996;40:108–112 Helen Tjoei-Lian Tjia, MMed (Int Med),* 2. Simpson JE, Newcombe J, Cuzner ML, Woodroofe MN. Ex- Tju-Siang Chua, MRCP(UK),† pression of monocyte chemoattractant protein-1 and other beta- Helen May-Lin Oh, MRCP(UK),† chemokines by resident glia and inflammatory cells in multiple Kwong-Ming Fock, FRCP(Edin),† sclerosis lesions. J Neuroimmunol 1998;84:238–249 Ashok Kurup, MRCP(UK),‡ Asha Das, MD,* 3. Monteyne P, Van Laere V, Marichal R, Sindic CJ. Cytokine Adrian Keng-Yew Tan, MRCP(UK),* mRNA expression in CSF and peripheral blood mononuclear and Wei-Ling Lee, FRCP(Edin)* cells in multiple sclerosis: Detection by RT-PCR without in vitro stimulation. J Neuroimmunol 1997;80:137–142 4. Talbot PJ, Ekande S, Cashman NR, et al. Neurotropism of A novel Hendra-like paramyxovirus named Nipah virus human coronavirus 229E. Adv Exp Med Biol 1993;342:339– (NiV) was the cause of an outbreak among workers from 346 5. DeRisi J, Penland L, Brown PO, et al. Use of a cDNA mi- one abattoir who had contact with pigs. Two patients croarray to analyze gene expression patterns in human cancer. had only respiratory symptoms, while 9 patients had en- Nat Genet 1996;14:457–460 cephalitis, 7 of whom are described in this report. Neu- 6. Heller RA, Schena M, Chai A, et al. Discovery and analysis of rological involvement was diverse and multifocal, includ- inflammatory disease–related genes using cDNA microarrays. ing aseptic meningitis, diffuse encephalitis, and focal Proc Natl Acad Sci USA 1997;94:2150–2155 brainstem involvement. Cerebellar signs were relatively 7. Shalon D, Smith SJ, Brown PO. A DNA microarray system for common. Magnetic resonance imaging scans of the brain analyzing complex DNA samples using two-color fluorescent showed scattered lesions. IgM antibodies against Hendra probe hybridization. Genome Res 1996;6:639–645 virus (HeV) were present in the serum of all patients. 8. Duggan DJ, Bittner M, Chen Y, et al. Expression profiling us- Two patients recovered completely. Five had residual def- ing cDNA microarrays. Nat Genet 1999;21(suppl):10–14 9. Ermolaeva O, Rastogi M, Pruitt KD, et al. Data management icits 8 weeks later. and analysis for gene expression arrays. Nat Genet 1998;20: Lee K-E, Umapathi T, Tan C-B, Tjia HT-L, 19–23 Chua T-S, Oh HM-L, Fock K-M, Kurup A, 10. Lennon G, Auffray C, Polymeropoulos M, Soares MB. The I.M.A.G.E. Consortium: An integrated molecular analysis of Das A, Tan AK-Y, Lee W-L. The neurological genomes and their expression. Genomics 1996;33:151–152 manifestations of Nipah virus encephalitis, a novel 11. Schuler GD, Boquski MS, Stewart EA. A gene map of the hu- paramyxovirus. Ann Neurol 1999;46:428–432 man genome. Science 1996;274:540–546 12. Becker KG, Mattson DH, Powers JM, et al. Analysis of a se- quenced cDNA library from multiple sclerosis lesions. J Neu- Japanese encephalitis (JE) is endemic in most parts of roimmunol 1997;77:27–38 Asia, including Malaysia, affecting 50,000 people per 13. Chen Y, Dougherty ER, Bittner ML. Ratio-based decisions and the quantitative analysis of cDNA microarray images. Biomed year. Singapore, a city-state with a population of al- Optics 1997;2:364–374 14. Lassmann H, Raine CS, Antel J, Prineas JW. Immunopathol- ogy of multiple sclerosis: Report on an international meeting held at the Institute of Neurology of the University of Vienna. From the *Department of Neurology, National Neuroscience Insti- J Neuroimmunol 1998;86:213–217 tute, †Department of Medicine, Changi General Hospital, and 15. Peiper SC, Wang ZX, Neote K, et al. The Duffy antigen/ ‡Department of Infectious Diseases, Communicable Disease Centre, Tan Tock Seng Hospital, Singapore. receptor for chemokines (DARC) is expressed in endothelial cells of Duffy negative individuals who lack the erythrocyte re- Received Apr 15, 1999, and in revised form May 20. Accepted for ceptor. J Exp Med 1995;181:1311–1317 publication May 20, 1999. 16. Horuk R, Martin AW, Wang Z, et al. Expression of chemokine Address correspondence to Prof W.-L. Lee, National Neuroscience receptors by subsets of neurons in the central nervous system. Institute, Level 3, Irrawaddy Block, 11 Jalan Tan Tock Seng, J Immunol 1997;158:2882–2890 Singapore 308433.

428 Copyright © 1999 by the American Neurological Association most 4 million people, was considered to be a nonen- Three patients had central hypoventilation with hy- demic area, as the last swine farm had been phased out percapnia and hypoxemia requiring mechanical venti- in 1990, with the last local case reported in 1992. Pigs lation. Admitting chest radiographs were normal in 2 are important amplifying hosts in the cycle of JE vi- patients and showed mild diffuse granular shadows in 1 rus. Five months after an apparent outbreak of JE the third. Two other patients had infiltrates on admit- among swine farm workers in Malaysia, an outbreak of ting chest radiographs. One was asymptomatic, and the 9 cases of encephalitis and 2 cases of pneumonitis oc- other had fever and cough. Follow-up chest radio- curred in Singapore among the workers in one abattoir, graphs were normal within a week in all 3 patients. where pigs imported from one farm in Malaysia were Table 2 summarizes the investigations. A complete slaughtered. A second virus was suspected when the se- blood cell count showed either a normal or low total rological tests for JE virus were negative. It has since been identified as a new strain of paramyxovirus, an white blood cell count in all of the patients; monocy- RNA virus similar to Hendra virus (HeV),2 and has tosis was seen in 4 patients, and thrombocytopenia was been named Nipah virus (NiV) after Kampung Sungei present in 3 patients. Electrolytes were within a normal Nipah, the village in Malaysia where the virus was first range, except for 1 patient with mild hyponatremia discovered. The predominant presenting features in (128 mmol/L) from a syndrome of inappropriate anti- this infection were localized to the central nervous sys- diuretic hormone secretion (Patient 7). The cell count tem. We report the neurological manifestations in 7 in cerebrospinal fluid (CSF) ranged from 0 to 920/mm3; cases of NiV encephalitis. total protein ranged from 40 to 468 mg/dl; and the CS- F:blood glucose ratio ranged from 0.41 to 0.87. Materials and Methods A computed tomography scan of the brain was nor- The case histories of 7 patients with encephalitis who were mal in all patients. Magnetic resonance imaging was admitted to Tan Tock Seng and Changi General Hospitals, performed in 4 patients. It showed subpial T2- two regional hospitals in Singapore, were reviewed. Capture- weighted hyperintense and T1-weighted hypo/isoin- IgM enzyme-linked immunosorbent assay for IgM antibodies tense lesions in the right frontal and left posterior pa- against prototype HeV antigens was positive in the serum of rietal lobes in 1 patient, which enhanced with contrast. all cases. The other patients had multiple foci of hyperintense lesions on T2-weighted and diffusion-weighted images Results in the corpus callosum, pons, middle cerebellar pedun- All of the patients were men ranging in age from 24 to 65 years. These patients worked in the abattoir and cle, corona radiata (Fig), and subcortical white matter were involved either in the slaughtering of pigs or in of the frontal and parietal lobes. Electroencephalo- the cleaning and processing of the carcasses. Each had grams in 3 patients showed generalized intermittent received two doses of the JE vaccine (lyophilized slow activity. “Biken”) ranging from 2 days to 2 weeks prior to At presentation, IgM antibodies against HeV were presentation. detected in the serum of all patients, although IgG an- Table 1 summarizes the neurological manifestations. tibodies were present only in Patients 2 and 4. IgM Six patients had prodromal symptoms of fever, nausea, antibodies against HeV were not detected in the CSF anorexia, vomiting, giddiness, malaise, and myalgia. of these 7 patients. An untyped virus was detected in Four had fever and signs of meningism, including stiff the CSF of Patient 6. Syncytial cells suggesting neck. Six had disturbances of mentation ranging from paramyxovirus were seen on tissue cultures. The CSF acute confusion to deep coma. Marked cerebellar dys- cultures of all 7 patients were subsequently sent to the function, including gait ataxia, bilateral limb dysmetria, Centers for Disease Control and Prevention (Atlanta, and dysdiadochokinesia, was seen in 5 patients. Three GA). Results are pending. Complement-fixing antibod- patients had either absent or diminished deep tendon ies against herpes simplex virus, measles, mumps, and reflexes. One patient had monoparesis of his left upper Flavivirus (JE/Dengue) were not detected in the CSF limb, and another had flaccid quadriparesis. One pa- of 5 patients. The titer against Flavivirus (JE/Dengue) tient had an abrupt onset of dysphonia, myoclonus of the right leg, and truncal ataxia, followed by left Hor- was 1:32 in 2 patients, which is consistent with their ner’s syndrome, bilateral dysmetria, and right abducens history of recent immunization. palsy. Another had vivid visual and auditory hallucina- All 7 patients received empirical treatment with in- tions consisting of pigs running around the bed. One travenous acyclovir and ceftriaxone. Two weeks after patient presented with severe blurring of vision in both the onset of the illness, 2 patients demonstrated com- eyes lasting less than 24 hours. An ophthalmological plete recovery. Although the remaining 5 patients assessment, including the visual field at the time of vi- showed significant clinical improvement, the neurolog- sual impairment, was normal. ical deficits persisted at 8 weeks.

Brief Communication: Lee et al: NiV Encephalitis 429 Table 1. Summary of Biodata and Neurological Manifestations

Case Age (yr)/Sex/Race Prodrome Headache Nuchal Rigidity Mental Status GCS

1 24/M/C 3 days Yes No Slow 15 2 55/M/C 3 days Yes No Hallucinations 15 3 45/M/C 3 days Yes Yes Withdrawn 15

4 42/M/I 3 days No Yes Coma 3

5 65/M/C Nil No No Alert 15

6 37/M/C 3 days No Yes Coma 8 7 55/M/C 4 days No Yes Coma 8

GCS ϭ Glasgow coma scale score; M ϭ male; C ϭ Chinese; I ϭ Indian; N ϭ normal; L ϭ left; R ϭ right; Ͼϭmore affected.

Table 2. Summary of Biochemical, Radiological, and Serological Investigations

Cerebrospinal Fluid Complete Blood Cell Count Magnetic Resonance Imaging Cells Protein Glucose TWC Plt (hyperintense on T2-weighted Case (per mm3) (mg/dl) (mmol/L)a (ϫ 109/L) (ϫ 109/L) imaging)

1 9 49 2.7 (41%) 7.4 (5%M) 200 — 2 58 46 3.4 (74%) 7.5 (15%M) 166 — 3 16 64 2.7 (55%) 3.4 (16%M) 124 Corpus callosum, left pons, middle cerebellar peduncle 4 577 (87%L) 432 7.4 (87%) 6.7 (9%M) 82 Corpus callosum, frontal and parietal lobes 5 0 40 6.5 (66%) 2.4 (7%M) 110 Deep white matter of frontal and parietal lobes 6 920 (70%P) 393 2.9 (57%) 3.1 (10%M) 96 — 7 770 (90%L) 468 2.8 (50%) 4.5 (11%M) 169 Right frontal, left posterior parietal lobes aFigure in parentheses indicates percentage of cerebrospinal fluid glucose to blood glucose. TWC ϭ total white blood cell count; Plt ϭ platelet count; IRS ϭ intermittent rhythmic slow activity; L ϭ lymphocytes; P ϭ polymorphs; M ϭ monocytes.

Fig. Axial (A) T2-weighted and (B) diffusion-weighted magnetic resonance imaging scans of Case 3 showing a focal hyperintense lesion at the rostral aspect of the right corona radiata.

430 Annals of Neurology Vol 46 No 3 September 1999 Motor Sensory Cerebellar Others Outcome (2 wk)

N N N Transient visual loss Recovered N N N R basal pneumonia Recovered N N Bilateral — Improved L Ͼ R Monoparesis N Bilateral Respiratory failure (hypercapnia, hypoxia) Improved L Ͼ R Diminished reflexes N Bilateral Focal myoclonus, dysphonia, Horner’s Improved syndrome, abducens palsy Tetraparesis, areflexia Yes Bilateral Respiratory failure (hypercapnia, hypoxia) Improved Areflexia in lower limbs N Bilateral Respiratory failure (hypercapnia, hypoxia) Improved R Ͼ L

Serology

Cerebrospinal Electroencephalogram Chest Radiograph on Admission Serum Fluid

— Left upper zone infiltrates IgM — — Right basal pneumonia IgM, IgG Negative IRS Normal IgM Negative

IRS Diffuse granular shadows IgM, IgG Negative

IRS Normal IgM —

— Normal IgM Negative — Normal IgM Negative

Discussion culopathy with endothelial damage and syncytial cell Members of the Paramyxoviridae family that can infect formation in the brain, lung, and kidney (N. I. Paton the human nervous system include measles, mumps, and co-workers, unpublished data). and parainfluenza viruses.3 In September 1994, a new The clinical manifestations of NiV infection appear member of the Paramyxoviridae family was identified as to be mainly neurological. In those cases with pulmo- the cause of an outbreak of respiratory diseases in nary infiltrates at presentation, respiratory symptoms horses and people and was named Hendra virus.4,5 In were mild. The respiratory failure in Patients 4, 6, and 1995, a single fatal case of HeV encephalitis was re- 7 was thought to be from central hypoventilation and ported in Australia.6 In this current outbreak of en- not as a result of lung involvement. Two other patients cephalitis, isolation material was obtained from the au- were admitted to Tan Tock Seng Hospital for mild topsies of several cases in Malaysia and Singapore, respiratory symptoms only and had infiltrates on chest including that from a patient admitted to another hos- radiographs. In contrast, respiratory involvement pre- pital in Singapore who worked at the same abattoir dominated the clinical picture in the HeV outbreak.2 dealing with pigs imported from the same farm. He The diversity and multifocality of neurological man- also had IgM antibodies against HeV. Virus-like struc- ifestations were particularly striking. In the extreme tures consistent with a paramyxovirus on electron mi- form, they appeared to involve the neural axis exten- croscopy studies were demonstrated, and immunofluo- sively, including the cerebrum, brainstem, cerebellum, rescence tests of cells infected with this virus suggested and meninges. Milder forms of the disease presented a virus related to HeV.7 Preliminary nucleotide se- with aseptic meningitis, mild diffuse encephalitis, and quencing indicated that this virus, named Nipah virus, focal brainstem involvement. JE infection has been re- had a 90% homology with HeV. These specimens ported to cause dystonia, rigidity, coarse tremors, car- were negative for JE antigens. Histology revealed vas- diorespiratory failure, flaccid paralysis, and death.8–10

Brief Communication: Lee et al: NiV Encephalitis 431 The propensity to affect the cerebellum may therefore We thank the Department of Diagnostic Imaging, Tan Tock Seng help to distinguish NiV from JE infection. Hospital; the Department of Pathology, Singapore General Hospi- The scattered lesions seen on magnetic resonance tal; and Dr Chew Suok Kai, Director, Epidemiology and Disease imaging could represent areas of focal cerebritis or vas- Control, Ministry of Health, for their invaluable assistance. culopathy; however, these lesions did not correlate with the clinical findings. The typical neuroimaging features of JE infection, namely, the involvement of the thala- References mus, basal ganglia, hippocampus, and brainstem, were 1. Johnson RT. Meningitis, encephalitis and poliomyelitis. In: Vi- not seen.11 In endemic areas of JE, this difference ral infections of the nervous system, 2nd ed. Philadelphia: could be an additional tool in distinguishing NiV from Lippincott-Raven, 1998:122–123 2. Selvey LA, Wells RM, McCormack JG, et al. Infection of hu- JE infection. mans and horses by a newly described Morbillivirus. Med J The CSF profile was remarkable for the markedly Aust 1995;162:642–645 raised protein and cell count in those severely affected 3. Murphy FA, Fauques CM, Bishop DHL, et al. Virus patients. Antibodies against HeV were notably absent taxonomy: Classification and nomenclature of virus. New York: in the CSF, although the viral antigens were identified Spring-Verlag, 1995:268–270 7 4. Murray K, Selleck P, Hooper P, et al. A Morbillivirus that in the brain. The HeV immunoassay may not be sen- caused fatal disease in horses and humans. Science 1995;268: sitive enough to detect NiV antibodies in CSF. 94–97 The portal of entry is likely through the respiratory 5. Yu M, Hansson E, Shiell B, et al. Sequence analysis of the system, as autopsy of infected pigs showed marked re- Hendra virus nucleoprotein gene: Comparison with other spiratory involvement with little involvement of the members of the subfamily Paramyxovirinae. J Gen Virol 1998; 7 79:1775–1780 nervous system. The origin of the virus is obscure, al- 6. O’Sullivan JD, Allworth AM, Paterson DL, et al. Fatal enceph- though like HeV, the fruit bat (which is found in rural alitis due to novel paramyxovirus transmitted from horses. Lan- swine breeding farms) may be the natural host. Studies cet 1997;349:93–95 by the public health authorities of Singapore and Ma- 7. Outbreak of Hendra-like virus—Malaysia and Singapore, 1998– laysia are underway to determine the epidemiology of 1999. MMWR Morb Mortal Wkly Rep 1999;48:265–269 8. Johnson RT, Burke DS, Elwell M, et al. Japanese encephalitis: this novel viral outbreak, including the prevalence of Immunocytochemical studies of viral antigens and inflamma- antibodies in asymptomatic abattoir workers, their tory cells in fatal cases. Ann Neurol 1985;18:567–573 families, and health care workers. 9. Burke DS, Lorsomrudee W, Leake CJ, et al. Fatal outcome in The fact that all infected patients worked in the Japanese encephalitis. Am J Trop Med Hyg 1985;34:1203–1210 same abattoir with unaffected family members and the 10. Hoke CH, Jr, Vaughn DW, Nisalak A, et al. Effect of high- dose dexamethasone on the outcome of acute encephalitis due absence of new cases following a total ban on the im- to Japanese encephalitis virus. J Infect Dis 1992;165:631–637 port of pigs suggests that this is a zoonotic disease with 11. Abe T, Kojima K, Shoji H, et al. Japanese encephalitis. J Magn transmission via direct contact with pigs. Reson Imaging 1998;8:755–761

432 Annals of Neurology Vol 46 No 3 September 1999