Proc. Nati. Acad. Sci. USAVol. 89, pp. 11949-11953, December 1992Biochemistry

Detection of glutamine synthetase in the cerebrospinal fluid ofAlzheimer diseased patients: A potential diagnosticbiochemical marker

(photoaffnity/ATP binding protein)

DEBRA GUNNERSEN AND BOYD HALEY*Department of Biochemistry and Division of Medicinal Chemistry and Pharmaceutics, College of Pharmacy, University of Kentucky, 800 Rose Street,Lexington, KY 40536-0084

Communicated by Henry A. Lardy, September 21, 1992 (received for review June 7, 1992)

ABSTRACT In this report, 8- and 2-azidoadenosine 5'-[y-32Pltriphosphate were used to examine cerebrospinal fluid(CSF) samples for the presence of an ATP binding proteinunique to individuals with Alzheimer disease (AD). A 42-kDaATP binding protein was found in the CSF ofAD patients thatis not observed in CSF from normal patients or other neuro-logical controls. The photolabeling is saturated with 30 FM2-azidoadenosine 5'-[y-32P~triphosphate. Photoinsertion canbe totally prevented by the addition of 25 pM ATP. Photoin-sertion of 2-azidoadenosine 5'-triphosphate into the protein isonly weakly protected by other nucleotides such as ADP andGTP, indicating that this is a specific ATP binding protein. Atotal of 83 CSF samples were examined in a blind manner. The42-kDa protein was detected in 38 of 39 AD CSF samples andin only 1 of 44 control samples. This protein was identified asglutamine synthetase [GS; glutamate-ammonia ligase; L-glu-tamate:ammona ligase (ADP-forming), EC 6.3.1.2] based onsimilar nucleotide binding properties, comigration on two-dimensional gels, reaction with a polyclonal anti-GS antibody,and the presence of significant GS enzyme activity in AD CSF.In brain, GS plays a key role in elimination of free ammoniaand also converts the neurotransmitter and excitotoxic aminoacid glutamate to glutamine, which is not neurotoxic. Theinvolvement of GS, if any, in the onset of AD is unknown.However, the presence of GS in the CSF of terminal ADpatients suggests that this enzyme may be a useful diagnosticmarker and that further study is warranted to determine anypossible role for glutamate metabolism in the pathology of AD.

Alzheimer disease (AD) is a disorder of unknown etiologyand without effective treatment. It is estimated that 3-10%o ofthe population over 65 years old is affected by AD (1). AD ischaracterized by progressive dementia and brain atrophy (2).Clinically, patients are diagnosed with AD by excluding allother causes of dementia and by observation of a decline incognitive function. An unequivocal diagnosis of AD can beconfirmed only by pathological examination of brain tissue.Neurofibrillary tangles (NFTs) and senile plaques (SPs) arethe characteristic lesions found in AD brain tissue. However,these two lesions are not unique to AD because SPs andNFTs can be found, to a lesser extent, in the brains ofnormalelderly individuals and are also present in other neurodegen-erative diseases.There has been considerable effort toward development of

a reliable antemortem diagnostic test for AD. Several studieshave focused on the cerebrospinal fluid (CSF) for the pres-ence ofa biochemical change that may be a diagnostic markerforAD (reviewed in ref. 3). Amyloid, paired helical filament,and a1-antichymotrypsin immunoreactivity have been re-

ported to be elevated in the CSF of AD patients (4-6). TheCSF has also been examined for concentration changes inneuropeptides (7), biogenic amines and their metabolites (8),amino acid or amino acid derivatives (9), vitamins (10), andtrace elements (11). However, none of these compounds hasemerged as a reliable marker for AD.Although ,f-amyloid and the components of NFTs such as

ubiquitin and tau protein are important constituents of thelesions of AD, it cannot be concluded that their presence isa cause or is absolutely diagnostic of the disease. Anotheraberrant biochemical event could occur that results in pro-duction of the symptoms of the disease as well as theappearance of SPs and NFTs.

It has been hypothesized that neurotransmitter systemssuch as cholinergic (12), monoaminergic (13), and glutamater-gic (14) pathways may be involved since they are altered inAD. Specifically, impairment of the glutamatergic neuro-transmitter system has been proposed because of the obser-vation of loss in AD of specific pyramidal cells that useglutamate as a neurotransmitter and the observed decrease ofspecific glutamate receptors (15, 16). Related to this are theobservations that overstimulation of neurons by glutamategenerally results in neuronal death, that glutamate can induceNFT-like structures in cultured neurons (17, 18), and that,8-amyloid can leave cultured neurons more susceptible toexcitotoxic damage (19).

Brain glutamine synthetase [GS; glutamate-ammonia 1i-gase; L-glutamate:ammonia ligase (ADP-forming), EC6.3.1.2] is located primarily in the astrocytes that supportneurons that use glutamate as a neurotransmitter (20). GSutilizes ATP to convert glutamate and an ammonium ion toglutamine. Thus, GS is a key metabolic enzyme not only forammonia assimilation and detoxification but also for termi-nation of a neurotransmitter signal (21, 22). Changes in theactivity or expression of GS in AD could alter glutamatemetabolism and exposure of neurons to glutamate.GS is readily detected in human brain homogenates by

photolabeling with radiolabeled 8-azidoadenosine 5'-triphosphate (8N3ATP) or 2-azidoadenosine 5'-triphosphate(2N3ATP), separation of proteins by gel electrophoresis, andautoradiography. Using these techniques, we monitored GSlevels in a large number ofCSF samples from individuals withvarious medical problems. GS was detected only in the CSFof AD patients. Other intracellular ATP binding proteins,such as creatine kinase and actin, were not observed in theCSF. Our observations suggest that the amount of GS, thecompartmentalization of GS, or glutamate metabolism isaltered or dysfunctional in AD. In addition the selective

Abbreviations: AD, Alzheimer disease; 2N3ATP, 2-azidoadenosine5'-triphosphate; 8N3ATP, 8-azidoadenosine 5'-triphosphate; CSF,cerebrospinal fluid; GS, glutamine synthetase; NFT, neurofibrillarytangle; SP, senile plaque; ALS, amyotrophic lateral sclerosis.*To whom reprint requests should be addressed.

11949

The publication costs of this article were defrayed in part by page chargepayment. This article must therefore be hereby marked "advertisement"in accordance with 18 U.S.C. §1734 solely to indicate this fact.

11950 Biochemistry: Gunnersen and Haley

appearance of GS in the CSF appears to be a reliable andspecific diagnostic marker for advanced stages ofAD and hasthe possibility of being useful for early diagnosis.

MATERIALS AND METHODSMaterials. [y_32P]8N3ATP (specific activity, 10-28 mCi/

limol; 1 Ci = 37 GBq) was prepared and purified as reported(23). The purity of [-32P]8N3ATP was determined by HPLCand TLC analysis (24). 2N3AMP was prepared as describedby Kim and Haley (25). 2N3ATP was prepared from 2N3AMPand PPi according to the method of Michelson as modified bySalvucci et al. (26, 27). [y32P]2N3ATP (specific activity, 5-14mCi/,tmol) was then prepared and purified as described for['t-32P]8N3ATP. Protein molecular size standards were fromBio-Rad. Antibodies against GS were provided by RichardMiller (Washington University, St. Louis). All other reagentswere analytical grade and were from Sigma.Sample Collection and Storage. Samples were received

from three different sources. CSF from Athena Neuro-sciences (San Francisco) had been stored at -200C forseveral months and the methods of diagnosis were unknown.These samples were not collected specifically for this studyso no uniform protocol or handling conditions were specified.Fresh samples were received from the Sanders Brown AgingCenter and the Veteran's Administration Hospital at theUniversity of Kentucky (Lexington). These samples werealiquoted and stored at -700C until use. Samples from theUniversity of Kentucky were diagnosed clinically by Na-tional Institute of Neurological and Communicative Disor-ders and Stroke-Alzheimer's Disease and Related DisordersAssociation criteria (28). The ventricular samples were ob-tained upon autopsy and a diagnosis was confirmed bypathological examination of brain tissue. These CSF sampleswere stored at 40C for 4-10 hr and then transferred to -700Cuntil use. CSF samples with any sign of blood contaminationwere rejected for analysis.

Photolabeling of CSF Samples. Initial CSF analysis wasperformed by mixing a 10- to 15-,ul sample ofCSF with 10 p.M[y-32P]2N3ATP or [y32P]8N3ATP in 50 ul of 25 mM potas-sium phosphate buffer (pH 7.0) at room temperature for 30sec followed by a 45-sec irradiation with a hand-held 254-nmUV lamp (I = 5000 p.W/cm2). After photolysis, the sampleswere precipitated with 3.5% perchloric acid and resolubilizedin a buffer containing 10%o SDS, 3.6 M urea, 162 mMdithiothreitol, pyronin Y as a tracking dye, and 20 mMTris*HCl (pH 8.0). Additions to the basic reaction mixturedescribed above were made as indicated in the figure legends.For protection studies, a 10-pl aliquot of a CSF sample in 25mM phosphate and a competing nucleotide were mixed for 2min at room temperature before addition of 25 p.M[-32P]2N3ATP and photolysis as described above.

Results reported in this paper have been reproduced aminimum of three times using at least three or more differentsamples in which a diagnosis of AD had been confirmed. AllCSF samples were initially examined under blind conditions.SDS/PAGE. Photolabeled samples were solubilized and

subjected to electrophoresis on a 10%o separating gel with a4% stacking gel according to the method of Laemmli asdescribed (27, 29). Autoradiograms were scanned with animage acquisition densitometer (Biolmage, Ann Arbor, MI;MilliGen/Biosearch, Novato, CA) to determine the relative32p incorporation. Radiolabeled bands were also excisedfrom the gels and the level of 32p was determined by a PackardMinaxi scintillation counter (99% efficient). The 32p incor-porated into the 42-kDa band in these experiments wasroutinely in the 2000-17,000 cpm range.

Immunoblotting. SDS/PAGE was performed as describedabove. The gel was then transferred to Immobilon mem-branes (Millipore) in 25 mM Tris HCl/192 mM glycine, pH

8.0/20%o methanol for 8-12 hr at 100 mA. The blot was treatedin the following manner: it was blocked with 3% bovineserum albumin, incubated with polyclonal anti-GS antibodies(1:1000 dilution) for 2-4 hr, incubated with 125I-labeled pro-tein A (ICN), and exposed to film for 10-18 hr.

RESULTS AND DISCUSSIONThere has been much interest in an easily biopsied biologicalmarker that is indicative of AD and there have been severalreports of potential markers in the CSF. However, none ofthese has emerged as a reliable test for AD (reviewed in ref.3). There are reported changes in nucleotide-interactingproteins ofAD brain such as elevated phosphorylation statesoftau protein and decreased GTP binding sites of tubulin (30,31). In this report, CSF samples from AD, other neurologicaldiseases, and neurologically normal individuals have beenexamined for the presence of a nucleotide interactive proteinthat could serve as a biological marker for AD.

In Fig. 1, the photoprobe [(y32PJ8N3ATP was used toexamine 10 CSF samples under blind conditions. The resultsshowed a photolabeled protein with a mass of 42 kDa thatappeared only in the AD samples (lanes 4, 5, 8, and 9), Thisprotein was not present in the CSF from epileptic, Parkinsondiseased, or clinically normal individuals (lanes 1-3, 7, 10,and 11). There are two other proteins of 66 and 28 kDa thatwere photolabeled in all CSF samples tested. The presence ofthese proteins did not correlate with any disease state.Albumin, a protein of 66 kDa, constitutes '967% of the totalCSF protein (32). Albumin is known to have a high bindingcapacity for several nucleotides and the photoaffinity analogis nonspecifically bound to this protein. This probably ac-counts for photoincorporation into the 66-kDa protein. Thephotolabeled 28-kDa protein remains unidentified.Because limited quantities ofthese samples were obtained,

equal volumes of CSF were used for experiments. Thiscaused the amount of protein loaded on the gel to differ witheach sample. However, the differences in the total amount ofprotein examined do not account for the differences observedin photolabeling. That is, even though there was much lesstotal protein in lane 4, the 42-kDa protein is still observed.

Photolabeling of the 42-kDa protein was completely pro-tected by the addition of 400 p.M ATP, whereas the 66- and28-kDa proteins were only partially protected (lanes 13-16).This indicated that the 42-kDa binding protein is very specificfor ATP, as was confirmed later (see Fig. 2).

Table 1 shows the results of 83 age-matched CSF samplesthat have been examined, including 39 cases that werediagnosed as AD. One of these cases had symptoms ofamyotrophic lateral sclerosis (ALS) as well as AD andanother case had symptoms of Pick disease as well as AD.Both of these cases were positive for the 42-kDa protein.Other ALS samples did not contain the 42-kDa protein,indicating that it is unique to AD.A total of 44 control samples were examined. These

included CSF from patients with normal cognitive function,patients with epilepsy, Parkinson disease, and ALS. The42-kDa protein was detected in only one control case. In thiscase, the patient was clinically diagnosed as normal, butpathological examination of brain tissue revealed numerousNFTs and SPs that are usually indicative of AD. Thus, thissample is not necessarily a good control. Our laboratory hasalso presented information on the detection ofa 55- to 56-kDaisoprotein set that is found only in AD brain (33). Brain tissuefrom this particular control patient also contained the 55- to56-kDa constellation of isoproteins.To show that the 42-kDa photolabeled protein in the AD

CSF is a specific nucleotide binding protein, saturation of thephotolabeling (Fig. 2) and a decrease of photoincorporationby competition with ATP (Fig. 2 Inset) were demonstrated.

Proc. NatL Acad. Sci. USA 89 (1992)

Proc. Nati. Acad. Sci. USA 89 (1992) 11951

I 2) 3 4 S5 612 34 5

_ _m _-

97.4w

66.2w

42.7w

7 8 9 () 11 12 13 14 1 5 16

-- lW-

__E_.Ie

31.0,,m-

21.5v fif

Diagnosis: E C P AD AD

ATP(400INI):-8N3A1'P(10OINI): +- + + -a

P AD l) (:-C F AD AD ADO A:

FIG. 1. Photolabeling of a 42-kDa protein inAD CSF with [v.32P]8N3ATP. Autoradiogramfrom a SDS/10%o polyacrylamide gel is shown.Lanes 1-5 and 7-11, a 10-1l aliquot ofCSF wasmixed with 10 IAM (V-32PI8N3ATP in 25 mMphosphate buffer for 30 sec followed by irradi-ation for 45 sec. Samples were analyzed bySDS/PAGE as described. Lanes 13-16, an ali-quot of CSF was first mixed with 400 ,uM ATPand then photolabeled as for lanes 1-5 and 7-11.Sample in lane 13 corresponds to lane 4, lane 14

-) corresponds to lane 5, lane 15 corresponds tolane 8, and lane 16 corresponds to lane 9. Lanes6 and 12 contained molecular size markers(kDa). E, epileptic; C, control; P, Parkinsondisease.

The 42-kDa protein displays saturation ofphotoinsertion at30 uM [(t32P12N3ATP (Fig. 2). The apparent Kd ofthe proteinfor 2N3ATP is 5 ,uM. The 42-kDa protein could be photola-beled with either [y-32P]2N3ATP or [y-32P]8N3ATP. Bothphotoprobes had similar affinities but differing efficiencies ofphotoinsertion. That is, photoinsertion with 2N3ATP was 5-7times greater than that with 8N3ATP. Photoinsertion into the42-kDa protein was not detectable with 10AM [yt32P]8N3GTP,32PJ8N3cAMP, [32P]2N3NAD, [a-32P]2N3ADP, or [a-32P]-8N3GDP.

Table 1. Summary of CSF samples examinedn positive

Diagnosis CSF location n for 42 kDa Age, yrNormal Ventricular 9 1* 71 ± 20

Lumbar 20 0 58 ± 17Epileptic Ventricular 2 0 53 ± 2

Lumbar 2 0 60 ± 10Parkinson disease Ventricular 4 0 50 ± 19

Lumbar 2 0 60 ± 10ALS Ventricular 1 0 72

Lumbar 4 0 53 ± 7Total ventricular 16 1Total lumbar 28 0

AD Ventricular 30 29 77 ± 10Lumbar 7 7 70 ± 7

AD/ALS Ventricular 1 1 75AD/Pick disease Ventricular 1 1 84

Total ventricular 32 31Total lumbar 7 7

Twenty-nine ventricular and 25 lumbar CSF samples were ob-tained from the University of Kentucky. Nineteen ventricular and 10lumbar samples were obtained from Athena Neurosciences. All ofthe AD samples were clinically diagnosed according to NationalInstitute of Neurological and Communicative Disorders and Stroke-Alzheimer's Disease and Related Disorders Association criteria, anda diagnosis of AD has been confirmed by pathological examinationof brain tissue for the ventricular samples from the University ofKentucky. Patient age is reported as mean ± SD.*This CSF was obtained from a patient clinically diagnosed asnormal, but numerous SPs and NFTs were found upon pathologicalexamination of brain tissue.

Photolabeling of the 42-kDa protein was specifically pro-tected by addition ofATP (Fig. 2 Inset). Complete protectionof photoinsertion was obtained with 18 ILM ATP. A compar-ison ofthe protection ofphotoinsertion achieved with variouscompeting nucleotides is presented in Table 2. ADP, AMP,and adenosine did not protect nearly as well as ATP or8N3ATP. GTP, 8N3GTP, GDP, GMP, and guanosine pro-tected to a much lesser extent than the adenine nucleotides.Pyrimidine nucleotides did not protect as well as the purinenucleotides and cAMP and NAD protected very poorly.These data show that the 42-kDa protein is a specific ATPbinding protein.

.0 L 04 08

100~ ~~~~0

8-0

0 so

60O0 40

0 200400U'

0 20 40 60 800+.- ~~~~~~~~~AT?,j.&M

0 20 40 60 80[y-32P]2N3ATP, AM

FIG. 2. Saturation of [v-32PJ2N3ATP photoincorporation into the42-kDa AD CSF protein and protection of photoincorporation byATP. A 10-Ml aliquot of CSF was mixed with the indicated concen-tration of photoaffinity probe, photolyzed, and analyzed by SDS/PAGE. 32p incorporation was quantitated as described. Photoincor-poration of 100% corresponds to 0.9 x 104 cpm. (Inset) Protection of[r32P]2N3ATP photoincorporation was performed with the indi-cated ATP concentration as described. Photoincorporation obtainedwith 10 IAM [hA32P]2N3ATP was taken as 100o.

Am qw- pl? Of

A*- Alm -Am&.

Biochemistry: Gun'nersen and Haley

11952 Biochemistry: Gunnersen and Haley

Table 2. Effect of nucleotide competitors on photoinsertion intothe AD CSF 42-kDa protein

Competitor (40 AM)NoneATPADPAMPAdenosineGTPGDPGMPGuanosinecAMPNADPyrophosphateUTPTTP8N3ATP8N3GTP

32p incorporation, %

1002

7579827879929987986699951077

Photoinsertion obtained with 20 IAM [-32P]2N3ATP was taken as100%1. A 10-AI aliquot ofAD CSF was mixed with 40 /AM competingnucleotide for 1 min at room temperature followed by addition of 20,AM [(ty32P]2N3ATP and photolysis as described. Data representaverages of three experiments with SD not more than 10%o.

The 42-kDa ATP binding protein found only in AD CSFwas identified as GS by several different techniques. Initially,it was observed that if the CSF was (NH4)2SO4 precipitated,the 42-kDa protein was photolabeled more effectively thanthe original sample. Since NH4 is a cosubstrate with ATP forGS, purified sheep brain GS (Sigma) was tested for similar-ities with the CSF 42-kDa protein.The nucleotide interactive properties of the 42-kDa protein

and GS were found to be indistinguishable. The 42-kDa proteinand purified GS both photolabeled with [y-32P]2N3ATP andshowed saturation near 30 A&M probe with half-maximal pho-toinsertion at 5 ,uM. Both the 42-kDa protein and GS displayed5- to 7-fold enhanced photoinsertion of2N3ATP over8N3ATP.Photolabeling of GS can be protected by ATP with bindingproperties nearly identical to the 42-kDa protein ofAD CSF.Other nucleotides that interact weakly with the 42-kDa proteinhad a similar effect on photoinsertion into purifiedGS (data notshown).Nonnucleotide ligands affect the photolabeling of GS and

the 42-kDa AD CSF protein in a similar manner. The additionof NH' specifically enhances photoincorporation of[y-32P]2N3ATP into GS and the 42-kDa protein, whereas itdecreases photoinsertion into creatine kinase and other pro-tein kinases. Glutarate, a glutamate substrate analog, de-creases the photolabeling efficiency of both the 42-kDa CSFprotein and GS (data not shown).

In addition, the CSF 42-kDa protein and GS comigrate onboth SDS/polyacrylamide gel and two-dimensional poly-acrylamide gel (isoelectric focusing and SDS/PAGE). Theisoelectric points of GS and creatine kinase (43 kDa) aremarkedly different and allow separation and identification ofthese two proteins. Two-dimensional gels of photolabeledAD CSF showed no sign of the presence of creatine kinase.The 42-kDa AD CSF protein also reacts with polyclonal

antibodies raised against rat brain GS as determined byWestern blotting (Fig. 3). CSF from an AD patient is shownin lane 5. A 42-kDa immunoreactive band is found thatcomigrates with purified GS present in lane 1. No immuno-reactivity was observed at 42 kDa in control samples (lanes3 and 4). There is also some immunoreactivity with a 66-kDaprotein. This reactivity was eliminated if the antibody con-centration was diluted to 1:5000 (data not shown). Thecontrol samples (lanes 3 and 4) were loaded with much moreprotein than the AD lanes and the 42-kDa protein was still not

(S k

FIG. 3. CSF 42-kDa protein is recognized by anti-GS antibodies.Purified GS and creatine kinase (1.0 ixg per lane) aMO 40 1l of CSFwere subjected to SDS/PAGE and electrophoretic transfer as de-scribed. The blot was then blocked with 3% bovine serum albuminand probed with anti-GS antibodies followed by MI-labeled proteinA. Autoradiogram made from this blot is shown. Arrowhead pointsto purified GS. CK, creatine kinase; AD, AD CSF; C, control CSF.

detected. A total of four control and four AD samples weretested; GS immunoreactivity was detected in all four ADsamples and was absent in four control samples. The anti-body does not react with purified creatine kiiase (lane 2), andpreincubation of the antibody with purified GS eliminatedany reaction with the CSF 42-kDa protein (data not shown).This indicates that the 42-kDa protein is not present in controlCSF and eliminates enzyme inactivation as the cause of theabsence of photoinsertion. Under the same Western blottingconditions, it was possible to detect GS in both AD andcontrol brain homogenates (data not shown) and the GSlevels appeared 2- to 5-fold higher in the AD brain samples.On the same gel, it was observed that [y32P]2N3ATP and theanti-GS antibody interacted with protein of identical mass.

Significant GS enzyme activity was observed in the ADCSF over the background levels found in control CSF whenGS activity was measured by the method of Miller et al. (34).In samples in which the 42-kDa protein was detected byphotolabeling or immunoblotting, an enzymatic activity of0.025-1.95 units per mg of CSF protein was measured.However, in samples in which no 42-kDa protein was de-tected by photolabeling or immunological techniques, only0.005-0.017 unit per mg of CSF protein was detected (unitrefers to ,umol of L-y-glutamyl hydroxamate formed in 15 minat 37C). The background level found in the control isprobably due to enzymatic activity other than GS-for ex-ample, glutaminase. Alternatively, the background couldalso be due to formation of hydroxamates other than themeasured product. However, the amount of GS activity inAD CSF was proportional to the amount of the 42-kDaprotein observed by photolabeling. There was more GSactivity in ventricular CSF when compared to lumbar sam-ples. This corresponds to the observation that photoinsertioninto the 42-kDa protein was always 4- to 6-fold higher inventricular versus lumbar CSF isolated from the same pa-tient.The unique selectivity and sensitivity of the photolabeling

technique has allowed for detection of GS in the CSF ofADpatients and represents a potential antemortem diagnostic

Proc. Nad. Acad. Sci. USA 89 (1992)

Proc. Natl. Acad. Sci. USA 89 (1992) 11953

test based on an easily measured biochemical marker. Theappearance of GS in the AD CSF is not likely due to generalcell lysis or astrocytosis since other brain intracellular nu-cleotide binding proteins, such as creatine kinase, actin, andtubulin, are not detected.These observations suggest that elevated GS in the CSF of

AD patients may lead to a clinically useful diagnostic markerfor AD. However, it will first be necessary to demonstratethat GS is elevated in the CSF of patients with early onsetAD. Also, it must be determined whether GS in the CSF istotally specific for AD and not elevated in other dementingdiseases, which are difficult to distinguish clinically from AD.At present, the involvement of GS, if any, in the onset ofADis unknown and further research will be needed to determinethe medical importance of our observations. However, it iseasy to construct experimental models in which the loss ofGS compartmentalization affects glutamatergic activity,which could result in the neuronal degeneration seen in AD(35, 36).

Special thanks to Drs. Dale Schenk (Athena Neurosciences), JohnSlevin, and Edward Kasarskis (Veteran's Administration MedicalCenter, Lexington, KY) and to William Markesbery (Center onAging, University of Kentucky) for CSF samples. This work wassupported by National Institutes of Health Grant GM-35766 and EliLilly Research Laboratories, Inc.

1. Bennett, D. A. & Evans, D. A. (1992) Dis. Mon. 38, 1-64.2. Adams, R. D. & Victor, M., eds. (1985) Principles ofNeurol-

ogy (McGraw-Hill, New York), pp. 861-866.3. vanGool, W. A. & Bolhuis, P. A. (1991) J. Am. Geriatr. Soc.

39, 1025-1039.4. Prior, R., Monning, U., Schreiter-Gadder, U., Weidemann, A.,

Blennow, K., Gottfries, C. G., Masters, C. L. & Beyreuther,K. (1991) Neurosci. Lett. 124, 69-73.

5. Wang, G. P., Iqbal, K., Bucht, G., Winblad, B., Wisniewski,H. M. & Grundke-Iqbal, I. (1991) Acta Neuropathol. 82, 6-12.

6. Matsubara, E., Hirai, S., Amari, M., Shoji, M., Yamaguchi,H., Okamoto, K., Ishiguro, K., Harigaya, Y. & Wakabayashi,K. (1990) Ann. Neurol. 28, 561-567.

7. AMom, J., Galard, R., Catalan, R., Castellanos, J. M.,Schwartz, S. & Tolosa, E. (1990) Eur. Neurol. 30, 207-210.

8. Hartikainen, P., Soininen, H., Reinikainen, K. J., Sirvi6, J.,Soikkeli, R. & Riekkinen, P. J. (1991) J. Neural Transm. 84,103-117.

9. Perry, T. L., Yong, V. W., Bergeron, C., Hansen, S. & Jones,K. (1987) Ann. Neurol. 21, 331-336.

10. Ikeda, T., Furukawa, Y., Mashimoto, S., Takahashi, K. &Yamada, M. (1990) Acta Psychiatr. Scand. 82, 327-329.

11. Hershey, C. O., Hershey, L. A., Varnes, A., Vibhakar, S. D.,Lavin, P. & Strain, W. H. (1983) Neurology 33, 1350-1353.

12. Lehericy, S., Hirsch, E. C., Cervera, P., Hersh, L. B., Huaw,J. J., Ruberg, M. & Agid, Y. (1989) Proc. Nail. Acad. Sci. USA86, 8580-8584.

13. Reinikainen, K. J., Soininen, H. & Riekkinen, P. J. (1990) J.Neurosci. Res. 27, 576-586.

14. Maragos, W. F., Greenamyre, T., Penney, J. B., Jr., & Young,A. B. (1987) Trends NeuroSci. 10, 65-68.

15. Pearson, R. C. A., Esiri, M. M., Hiorns, R. W., Wilcock,G. K. & Powell, T. P. S. (1985) Proc. Nail. Acad. Sci. USA 82,4531-4534.

16. Penney, J. B., Maragos, W. F., Greenamyre, J. T., Debowey,D. L., Hollingsworth, Z. & Young, A. B. (1990) J. Neurol.Neurosurg. Psychiatr. 53, 314-320.

17. Mattson, M. P. (1990) Neuron 4, 105-117.18. Koh, J., Yang, L. L. & Cotman, C. W. (1990) Brain Res. 533,

315-320.19. DeBoni, U. & Crapper McLachlan, D. R. (1985) J. Neurol.

Sci. 68, 105-118.120. Norenberg, M. D. & Martinez-Hernandez, A. (1979) Brain

Res. 161, 303-310.21. Cooper, A. J. L. & Plum, F. (1987) Physiol. Rev. 67, 440-519.22. Derouiche, A. & Frotscher, M. (1991) Brain Res. 552, 346-350.23. Potter, R. & Haley, B. E. (1982) Methods Enzymol. 91, 613-

633.24. Czarnecki, J., Geahlen, R. T. & Haley, B. (1979) Methods

Enzymol. 56, 642-653.25. Kim, H. & Haley, B. (1990) J. Biol. Chem. 265, 3636-3641.26. Michelson, A. M. (1964) Biochim. Biophys. Acta 91, 1-13.27. Salvucci, M., Chavan, A. & Haley, B. (1992) Biochemistry 31,

4479 4487.28. McKhann, G., Drachman, D., Folstein, M., Katzmann, R.,

Price, D. & Stadlan, E. (1984) Neurology 34, 939-944.29. Laemmli, U. K. (1970) Nature (London) 227, 680-685.30. Grundke-Iqbal, I., Iqbal, K., Tung, Y. C., Quinlan, N.,

Wisniewski, H. M. & Binder, L. I. (1986) Proc. NatIl. Acad.Sci. USA 83, 4913-4917.

31. Khatoon, S., Campbell, S. R., Haley, B. E. & Slevin, J. T.(1989) Ann. Neurol. 26, 210-215.

32. Thompson, E. J. (1988) The CSF Proteins: A BiochemicalApproach (Elsivier, New York), p. 12.

33. Gunnersen, D. J., Slevin, J. T. & Haley, B. E. (1991) FASEBJ. 5, 456A (abstr.).

34. Miller, R. E., Hodenberg, R. & Gersham, H. (1978) Proc. Nail.Acad. Sci. USA 75, 1418-1422.

35. Greenamyre, J. T. (1988) Arch. Neurol. 43, 1058-1063.36. Procter, A. W., Palmer, A. M., Francis, P. T., Lowe, S. L.,

Neary, D., Murphy, E., Doshgi, R. & Bowen, D. M. (1988) J.Neurochem. 50, 790-802.

Biochemistry: Gunnersen and Haley