Vol. 181, No. 2, 1991 December
BIOCHEMICAL
AND BIOPHYSICAL RESEARCH COMMUNICATIONS
16, 1991
Pages
ALZHEIMER
DISEASE PAIRED HELICAL FILAMENT STRUCTURES CONTAIN GLYCOLIPID
771-779
CORE
Dennis R. S arkman*, Warren J. Goux#, Christine M. Jones, 8 harles L. White, III and Sandra J. Hill The University of Texas Southwestern Medical Center at Dallas Department of Pathology, 5323 Harry Hines Boulevard Dallas, Texas 752359072 #The Universiy of Texas at Dallas Department of Chemistry P.O. Box 830688 Richardson, Texas 75083 Received
,
October
23,
1991
SUMMARX The core structures of sodium dodecyl sulfate extracted, pronase digested paired helical filaments of Alzheimer disease were solubilized by beating in dimethyl sulfoxide. Electron microscopy revealed that a&x heating in dimethyl sulfoxide, intact paired helical filaments were no longer present in the dimethyl sulfaxide soluble fractions or in the insoluble lipofuscin-containing fractions. Enzyme-linked immunosorbant assays of the various fractions with the monospecific antibody A128 to paired helical filaments demonstrated 96% of the immunoreactivity to be in the dimethyl salfoxide soluble fraction, and only 4% in the dimethyl sulfoxide insoluble fhWions. LyophiIization of the dimethyl sulfoxide soluble supernatant and resuspension in water failed to reassociate the paired helical filaments, but did result in an insoluble precipitate. Analysis of the dimethyl sulfoxide solubilized paired helical filament &action by nuclear magnetic resonance revealed it to be composed of glycolipid in a form that was distinct brn similar fractions isolated &om normal aged control brains. The aggregation of an altered glycolipid to form paired helical fSlaments in Alzheimer disease could explain their insolubility. e,1991Academic PESS, inc.
Alzheimer disease (AD) is a progressive degenerative dementia which is characterized neuropathologically by the accumulation of neurofibrillary tangles (NFT), senile plaques (SP) and cerebrovascular amyloid (CVA) in the brain. The p/A4 amyloid fibrils which are present in SP and CVA have been solubilized and their amino acid sequence determined (1,2). However, the composition of paired helical filaments (PHF), the predominant structures in NFT and degenerating neurites of SP, has remained elusive. The difficulty in analyzing PHF stems from their high degree of insolubility in detergents and chaotropic salts (3,4), but this insolubility has allowed the extraction and purification of PHF from NFT from AD brains ($6).
II
*To whom correspondence should be addressed. 0006-291x/91
771
$1.50
Copyright 0 1991 by Academic Press, Inc. All rights of reproduction in any form reserved.
Vol.
181, No. 2, 1991
BIOCHEMICAL
AND BIOPHYSICAL RESEARCH COMMUNICATIONS
PHF have been assumed to be proteins which are resistant to solubilization, proteolysis and chemical cleavage (3,4). Although a number of studies have identified several proteins to be associated with the NFT (7-9), quantitative solubilization of the PHF core structures has not been demonstrated. Recent studies of PHF have used pronase digestion to remove their outer “fuzzy” coat, which was determined by amino acid sequence analysis to consist of the microtubule-associated protein, tau (10). This outer tau coat was shown to compose only 10% of the PHF, the remaining 90% being resistant to solubilization and thus precluding further analysis. PHF-containing NFT are one of the most consistent lesions in AD, and their density at autopsy has been correlated with the degree of clinical dementia (II); thus determining their structural composition would be a major step toward identifying the underlying cause of AD.
MATERIALS
AND MGFHODS
PHF Iso&tions. PHF were isolated from AD brain as previously described (6) with the following modifications. Routine preparations employed 200 g of NFT-rich gray matter and employed a Beckman 45Ti pre arative ultracentrifuge rotor. Prior to sucrose density adient centrifu ation, the crude N K- were resuspended in 100 mM Tris-HCI, pH 810 mM E DTA, 2 mM Ai Cl,, 0.5% SDS containing lOC&/ml pronase. The NFT were incubated at 42“ C for 18 hours, at which time the suspension was adjusted to 1% SDS and 0.1 M amercaptoethanol and boiled for 5 minutes. After recovery of the 1.4-2.0 M fraction, the final suspension of disaggregated PHF was dialyzed 18 hours against 0.1% dimethyl sulfoxide DMSO), followed by extensive dialysis against distilled water for 2 days at 4” C.TheP I-A were dried down in a Speed-Vat and weighed, To monitor the complete removal of sucrose, [14-C] sucrose was included in some of the gradients. Control brain fractions were obtained using aged control brains in parallel isolations. &&so Acti Analysis.The sample was hydrolyzed for 21 hours at 104“ C in 6N HCl in VUCUO. The hydrolysate was analyzed on a Waters Model 441 amino acid analyzer using the picotag program. Nitrogen Analysis. Elemental microanalysis was performed by Galbraith Laboratories, Knoxville, Tennessee. Electron Microsco . The PHF were negatively stained for electron microscopy as previously described (67 with 1% aqueous uranyl acetate. Enzyme-Linked fmmunosarbant Assay (ELZSA). The DMSO solubilized PHF core structures were centrifuged at 10,000 x g for 10 minutes at 2.5” C to remove insoluble lipofuscin particles. The lipofuscin was resuspended in distilled water and both it and the DMSO supernatants were lyophilized in a Speed-Vat. The samples were diluted in 0.1 M sodium bicarbonate buffer, H 9.5 and 100~1 dried in each well at 37” C for 16 hours. The wells were blocked with 1%o BSA in 10 mM Tris-HCl, pH 82,200 mM NaCl TBS-200) with 0.01% Tween 20 for one hour at 37” C, followed by three washings with TBS-2 06 /Tween. Al28 antiserum (6) was serially diluted in blockin buffer and incubated for one hour at 37” C. The wells were washed four times with TB iii-2OOKween. Horseradish eroxidase-labeled swine anti-rabbit immunoglobulins were diluted to 1:lOOO in blockin bu f!fer and incubated for one hour at 37” C. The wells were washed three times with TB 8 -2OO/rween and twice with TBS-200. The color was developed with 100~1 per well of 0.8 mg/ml o-phenylene diamine in 0.1 M sodium citrate phosphate buffer, pH 5, containing 0.03% hydrogen peroxide for 30 minutes at ambient temperature. The reactions were stopped by the addition of 10~1 of 10 N sulfuric acid and absorbance read at OD, . Nuclear Magnetic Resonance (Nh4R) Anafysis. PHF were soq)ubilized by heating at 65” C for 3 hours in d -DMSO/D,O (9:l) while stirring. 500 MHz ‘H NMR s ectra were run overmg5 t on a General Electric GN-500 NMR system. The residual I! ,O resonance of the solvent occuring at about 3.30 ppm was saturated rior to the observe pulse and acquisition. Spectra are reported using the residual R ,d,-DMSO solvent resonance at 2.49 ppm as an internal chemical shift reference. 772
Vol.
181, No. 2, 1991
BIOCHEMICAL
AND BIOPHYSICAL
RESEARCH COMMUNICATIONS
RESULTSANDDISCUSSION Total amino acid analysis of the PHF core preparations showed the presence of amino acids (Table l), but these only accounted for about 4.76% of the total mass. Amino acid analysis of material isolated from aged normal control brains showed similar amounts of protein (4.36%), although the individual values for each amino acid differed. These small amounts of amino acids are most likely due to residual lipofuscin granules, indicating that the PHF core structures are most likely composed of nonproteinaceous material. An analysis of total elemental nitrogen showed that the PHF core preparations contained only 2.25% N, again much lower than the expected value of 15.26% N for a protein composed of equal amounts of all 20 amino acids. To exclude the possibility of residual sucrose contaminating the PHF preparations, [14-C]sucrose was included in the sucrose gradients in several of the preparations. Dilution of the PHF fraction from the 1.4-2.0 M interface in distilled water followed by centrifugation removed 99.6% of the radioactivity, and the dialysis completely removed the remaining radioactivity. Electron microscopy (EM) of the PHF fractions revealed intact twisted filaments (Fig 1A) which were used as starting material for solubilization in DMSO. After heating in DMSO for 3 hours, the brown lipofuscin fraction was separated from the solubilized PHF by centrifugation. The supernatant contained no visible intact PHF core structures (Fig 1B) and appeared identical to the control supernatant (Fig 1C). The DMSO insoluble
TABLE
1
Total Amino Acid Analysisof the AkzheimerBrain Samplesand Similar Fractionsfrom Aged Control Brain Amino Acid
Y,i%%T
Ala Pro Tr dl Met
10.45kO.64 11.95k1.48 7.40-e1.56 6.85~~0.78 4.5020.28 4.1OkO.85 3.8521.06 4.8520.21 2.4020.14 4.45-cO.78 8.9O-cO.42 2.55rt0.35 2.3523.32 5.5521.20 6.7020.42 7.15k2.05 6.OO-cO.42 ND
CYS Ile Leu
Percent protein of total mass
4.76%
Aged C;:;o&Sample 5.60~0.71 8.6020.28 6.4020.14 5.90r0.14 2.6520.07 390~0.28 7.3020.42 7.2540.50 3.00*0.14 5.0520.21 6.95k0.49 1.75kO.21 3.6522.90 4.90k0.28 12.05kO.50 9.75kO.07 5.40-0.42 ND 4.36%
ND = Not Determined.Data are the averageof two experiments(mean-c standarddeviation). 773
Vol.
181, No. 2, 1991
BIOCHEMICAL
AND BIOPHYSICAL RESEARCH COMMUNICATIONS
Electron microscopy of A) the PHF fraction of AD brain, (B) the 0 supernatantfrom the PHb fraction, (C) the DMSO/D 0 su ernatant from the norm& a ed brain, (D) the DMSOQO insolublepellet from tie Pl-fF fraction, (E) the DMSO h z0 insoluble pellet from the normal aged control brain, (F) the DMSO/D 0 soluble PHF material after lyophilization and resuspensionin distilled Hz, andr (G) the DMSO/D,O soluble normal aged brarn material after lyophilization and resuspension1x1distilled HzO. Both sampleswere isolated using identicalprocedures.Line scale= 2.50nm. 774
Vol. 181, No. 2, 1991
BIOCHEMICAL
AND BIOPHYSICAL
RESEARCH COMMUNICATIONS
TABLE 2
Percent
PHF Immunoreactivity
in the DMSO Soluble and Insoluble Fractions of Alzheimer Brain and Similar Fractions from Aged Control Brain . 2 AD Brain Fractions’ C ntrol Brain Fractrnns Insoluble Soluble Insoluble Soluble 95.7%
4.3%
5.5%
0.7%
’ Total immunoreactivity of the AD brain fraction was measured by ELISA and calculated for the total volume of both the soluble and insoluble fractions, divided by the sum of both. 2 Percent immunoreactivity for the control brain fractions, relative to the PHF fractions, was measured by ELISA and calculated for the total volume of both the soluble and insoluble fractions, divided by the sum of the PHF immunoreactivity.
lipofuscin fractions from the AD and control brains contained a coarse brown granular material (Figs 1D and lE), but no PHF remained in the AD fraction. The DMSO soluble fractions from both the AD and control brains were lyophilized to dryness and resuspended in distilled water. The PHF soluble fraction yielded an insoluble material that appeared lattice-like by EM (Fig lF), but showed no reformed PHF structures. The control also contained insoluble material, but it was not well formed (Fig 1G). Lyophilization of non-solubilized PHF, followed by resuspension in distilled water had no effect on the morphology of the PHF (data not shown). Analysis of the DMSO soluble and insoluble fractions from the AD brain by ELISA revealed that 95.7% of the PHF immunoreactivity was present in the DMSO soluble fraction, with only 4.3% remaining in the insoluble fraction (Table 2). There was only 5.5% and 0.7% PHF immunoreactivity in the DMSO soluble and insolubIe fractions, respectively, from the aged control brain relative to the PHF fraction from the AD brain. Although most of the PHF immunoreactivity appears to be in the soluble fraction, it must be noted that it is only the relative percentage compared to the insoluble values. It is not known if DMSO treatment altered antigenic sites and reduced the antibody’s ability to recognize the solubilized PHF structures. Fig. 2A shows the 500 MHz proton NMR spectrum of a typical DMSO/D,O solubilized PHF fraction isolated from an AD brain. For comparison, the spectrum of a similarly prepared sample from normal aged brain is shown in Fig. 2B. Both spectra resemble high-field NMR spectra of human brain gangliosides (12-19). Downfield resonances in these spectra were previously assigned to the olefinic protons (5.2-5.5 ppm), the anomeric protons of the carbohydrate moieties (4.0-5.0), sphingosine protons Rl-R4 and the remaining carbohydrate pyranosyl ring protons (3.0-4.0 ppm). Resonances occuring in the upfield region were similarly assigned to the acetamido methyl protons of the carbohydrate moieties (1.7-1.9 ppm), the allylic and the ar-carbonyl methylene protons (1.9-2.2 ppm), and the ceramide methylene (1.2 ppm) and methyl protons (0.85 ppm). Based upon these previous assignments and upon a comparison of the resonances in the Fig. 2 spectra with those of a commercially obtained ceramide, the triplet at 0.85 ppm in both spectra can be assigned to the ceramide methyl protons while the singlet at 2.05 ppm and 775
Vol.
181, No. 2, 1991
BIOCHEMICAL
AND BIOPHYSICAL
RESEARCH COMMUNICATIONS
B
I ““l”“l”“l”“l”“l”‘~1 6 5
A. 4 GM
2
1
0
Figure 2. 500MHz NMR of (A) the PHF fraction of AD brain, and (B) normal apedbrain. Both sampleswere isolatedusingidentical procedures.Sampleswere dissolvedm a.%10 mixture of DMSOQO. Spectra were acquired at 333” K. The structure shown m the top figure is that of glucoceramideas determinedin ref.12 and is shownas a guide for resonanceassignments.
the triplet at 2.20 ppm in the spectra of the PHF sample can be assigned to the allylic and the cu-carbonyl methylene protons of the ceramide moiety. While the 2.05 ppm allylic proton singlet is much less prominent in the spectrum of the normal aged brain sample, there appear to be two broad based groups of resonance at about 5.25 and 5.5 ppm which, by comparison to the ceramide standard, can be assigned to the olefinic protons. In spectra of other PHF samples prepared in an identical manner, there is substantial variation in the intensity ratio of the allylic proton to methylene proton ceramide resonances, indicating a variation in the degree of unsaturation in the alkyl chains. 776
Vol.
181, No. 2, 1991
BIOCHEMICAL
AND BIOPHYSICAL RESEARCH COMMUNICATIONS
Neither of the spectra of Fig. 2 show resonances in the region immediately upfield of the ceramide aliylic protons at 2.05 ppm. This suggests the absence of any acetamido carbohydrate residues typically found in other brain gangliosides and glycoproteins (i.e., N-ace@ glucosamine, N-ace@ galactosamine or sialic acid) and is consistent with lectin studies on isolated PHF (20). In both spectra, however, there is a well resolved doublet present in the spectral region characteristic of anomeric proton resonances of attached carbohydrates (4.68 ppm). The magnitude of the coupling constant for this resonance (3.1 Hz) suggests the sugar has assumed an cu-anomeric ring configuration. The shift and coupling constant for this resonance, when taken together, are similar to those observed for cu-D-galactose residues in glycolipid samples previously reported (21,22). Doublets also occur in both spectra which fall just outside the region typical of anomeric protons of other gangliosides in similar solvent systems. These include those in both spectra at 5.10 ppm (J,, = 5.0 Hz) and in the spectrum of the normal aged brain sample at 5.20 and 5.23 ppm (J,, = 3.7 and 5.1 Hz, respectively). We also acquired NMR spectra in DMSO of the AD lipofuscin fraction, obtained from the 1.0-1.2 M interface of the sucrose gradients, which is often a contaminant of the PHF fraction (data not shown). The spectra is absent of resonance which could be assigned to anomeric protons of monosaccharides residues and closely resembles the ‘H NMR spectrum previously reported for purified lipofuscin from normal human brains (23). The most striking difference between the PHF spectrum and that of the normal aged brain sample is the appearance of the sphingosine Rl-R4 and pyranosyl ring proton resonances between about 3.2 and 3.8 ppm. The general overall broadened nature of the envelope in the PHF spectrum is consistent with broadened linewidths and shorter Tz relaxation times characteristic of polymerized or aggregated samples having slower rotational relaxation times. Although PHF have been assumed for a number of years to be composed of a highly insoluble protein(s), certain properties tended to argue against their being proteinaceous at all. For example, we found the core structures to be resistant to numerous proteases, chemical cleavages and partial acid hydrolysis. This, along with the low percentage composition of amino acids and elemental nitrogen, fails to support their being composed of significant amounts of protein. The present results indicate that they may, instead, be composed of a glycolipid, possibly a glycoceramide, capable of assembling into PHF. Ceramides are insoluble in non-polar solvents and self-aggregation or polymerization would result in a large, highly insoluble molecule. There have been reports of altered membrane fluidity in AD which correlates with the severity of neuropathology (24). Brain lipids reported to be altered in AD include gangliosides (25), phosphoinositides (26), phosphomonoesters (27), and phosphatidylethanolamines (28). Antibodies to gangliosides have also been shown to immunolabel both SP (29, 30) and NFT (30,31). A monoclonal antibody which reacts with sulfatides and gangliosides showed high levels of immunoreactivity in cell bodies and degenerating neurites in AD brain (32). Electron microscopic analysis of brain biopsies 777
Vol.
181, No. 2, 1991
BIOCHEMICAL
AND BIOPHYSICAL
RESEARCH COMMUNICATIONS
from Alzheimer patients showed that PHF may arise at or from the cell membrane, a lipid-rich structure (33). It has been shown that lipids can aggregate to form filaments in vitro (34), as well as in aged erythrocytes (35). These filaments can exist as double helical filaments (36). Should PHF cores be confirmed to be composed of aggregated or polymerized glycolipid(s), it would explain their relative insolubility and could offer explanations of other changes observed in AD. For example, alterations in glycolipid metabolism could destabilize neuronal membranes, making the amyloid precursor protein more susceptible to proteolytic cleavage, resulting in the deposition of amyloid, which is also characteristic of AD pathology. Finally, the search for a defect in glycolipid metabolism represents a new avenue of research into the etiology of AD. ACKNOWLEDGMENTS
Supported by an Advanced Technology Education Coordinating Board, a National Research Center Grant (AG08013), a French American Health Assistance Foundation grant grant (AT-1162) (WJG).
Grant (003660-108) from the Texas Higher Institutes of Health Alzheimer’s Disease Foundation for Alzheimer Research grant, a (900066) and a Robert A. Welch Foundation
REFERENCES
Glenner, G.G. and Won , C.W. (1984) Biochem. Biophy. Res. Comm. 120,885-890. :: Masters, CL, Simm, G., &v einman, G., Multhaup, G., McDonald, B.L. and Beyreuther, K. 1985) Proc. Natl. Acad. Sci. USA 82,4245-4249. &elkoe, D.J., Ihara, Y. and Salazar, F.J. (1982) Science215,1243-1245. :: Branbury Selkoe, D.J., Ihara, Y., Abraham, C., Rasool, C.G. and McCluskey, A.H. (1983) Report 15: Biological Aspects of Alzheimer’s Disease, Cold Spring Harbor Sympostum, 125-135.
Y., Abraham, C. and Selkoe, D.J. (1983) Nafure 304,727-730. 2 S3Ihara, arkman, D.R., Hammon, K.M. and White, C.L. III (1990) J. Histochem. Cytochem. ii ,703-715. 7. &$kig8S., w
Joachim,
CL. and Selkoe, D.J. (1986) Proc. Natl. Acad. Sci. USA 83,
8. Grundke-I bal I I bal K Quinlan M., Tung, Y.C., Zaidi, MS. and Wisniewski, H.M. 1988 J. &di. a&. 28i, 6084-6089. Mori, i-I ., Kondo, J. and Ihara, Y. (1987) Science235, 1641-1644. zi. Wischik, C.M., Novak, M., Thogersen, H.C., Edwards, PC., Runswick, M.J., Jakes, R., Walker, J.E., Milstein, C., Roth, M. and Klug, A. (1988) Proc. Natl. Acad. Sci. USA, 85,4506-4510. 11. Crystal, H., Dickson, D., Fuld, P., Masur, D., Scott, R., Mehler, M., Masde, J., Kawas, C., Aronson, M. and Wolfson, L. 1988) Neurology 38,1682-1687. 12. Koerner,, T.A.W.,Jr., Prestegard, J.i-I ., Demou, P.C. and Yu, R.K. (1983) Biochemtstry 22: 2676-2687.
13. Koerner, B&hem&try
T.A.W.,Jr.,
Prestegard,
J.H.,
Demou,
PC.
and
Yu,
R.K.
(1983)
222687-2690.
14. Li., Y.-T., Hirabayashi, Y., DeGas eri, R., Yu, R.K., Ariga, T., Koerner, T.A.W. and IA, S.-C. 1984)J. Biol. Chem. 259:8f 80-8985. 15. Dabrows L i, J., Hanfland, P. and Egge, H. (1982) Methods in Enzymology, Vol. 83 Ed. V. Ginsbur ), Academic Press, New York, 69-86. 16. 6 abrowski, J., I-f anfland, P. and Egge, H. 1980) Biochemistry 19:5652-5658. 17. Hanfland, P., E e, H., Dabrowski, U., L uhn, S., Roelcke, D. and Dabroski, J. (1981) Biochemtstry 20:8 10-5319. 778
Vol. 181, No. 2, 1991
BIOCHEMICAL
AND BIOPHYSICAL
RESEARCH COMMUNICATIONS
18. Scarsdale, J,N., Prestegard, J.H., and Yu, R.K. (1990) Biochemists 299843-9855. 19. Acquotti, D., Poppe, L., Dabrowski, J., von der Lieth, C.-W., Sonnino, S. and Tettamanti, G. 1990) J. Am. Chem. Sot. 112:7772-7778. 20. S arkman, D.k ., Hill, S.J. and White, CL., III (1990) Acta Neuropathol. 78 ~640-646. Scarsdale, J.N. and Prestegard, J.H. (1986) Carbohydr. Res. 155:45-56. % Inagaka, F., Kohda, D., Kodama, C. and Suzuki, A. 1987) FEBS Lett. 212:91-97. Taubold, R.D., Siakotos, A.N. and Perkins, E.G. (19(75) Lip& l&383-390. E Zubenko, G. (1986) Bruin Res. 385,115-121. 25: Crino, P.B., Ullman, D., Vogt, B.A., Bird, ED. and Volicer, L. (1989) Arch. Neural. 46,398-401. Stokes, C.E. and Hawthorne, J.N. (1987) J. Neurochem. 48,1018-1021. 2 Pettegrew, J.W., Moossy, J., Withers, G., McKeag, D. and Panchalingam, K. (1988) J. Neuropathol. EIxp. Neural. 47,235-248. 28. Sdderber , M., Edlund, C., Kristensson, K. and Dallner, G. (1991) Lipids 26~421-42 s . 29. Iwamoto, N., Suzuki, Y., Makino, Y., Haga, C., Kosaka, K. and Iizuka, R. (1990) Brain Rex 522,152-156. 30. Takahashi, H., Hirokawa, K., Ando, S. and Obata, K. (1991) Actu Neuropathol. 81:626-631. Emory, C.R., Ala, T.A. and Frey, W.H. II (1987) Neurology 37,768-772. 2 Majocha, R.E., Jungalwala, F.B., Rodenrys, A. and Marotta, C.A. (1989) J. Neurochem. 53,953-961. 33. 73ra i l”;:., Paula-Barbosa, M and Roher, A. (1987) Neuropathol. Appl. Neurobiol. d 34. Rddoiph, AS., Ratna, B.R. and Kahn, B. (1991) Nature 352:52-55. 35. Bessis, M. (1973) Livirlg Blood Cells and their Ultrastructure, Springer-Verlag, New York, 60-61. 36. Lin, K.-C., Weis, R.M. and McConnell, H.M. (1982) Nature 296164-165.
779
BIOCHEMICAL
AND BIOPHYSICAL RESEARCH COMMUNICATIONS
16, 1991
Pages
ALZHEIMER
DISEASE PAIRED HELICAL FILAMENT STRUCTURES CONTAIN GLYCOLIPID
771-779
CORE
Dennis R. S arkman*, Warren J. Goux#, Christine M. Jones, 8 harles L. White, III and Sandra J. Hill The University of Texas Southwestern Medical Center at Dallas Department of Pathology, 5323 Harry Hines Boulevard Dallas, Texas 752359072 #The Universiy of Texas at Dallas Department of Chemistry P.O. Box 830688 Richardson, Texas 75083 Received
,
October
23,
1991
SUMMARX The core structures of sodium dodecyl sulfate extracted, pronase digested paired helical filaments of Alzheimer disease were solubilized by beating in dimethyl sulfoxide. Electron microscopy revealed that a&x heating in dimethyl sulfoxide, intact paired helical filaments were no longer present in the dimethyl sulfaxide soluble fractions or in the insoluble lipofuscin-containing fractions. Enzyme-linked immunosorbant assays of the various fractions with the monospecific antibody A128 to paired helical filaments demonstrated 96% of the immunoreactivity to be in the dimethyl salfoxide soluble fraction, and only 4% in the dimethyl sulfoxide insoluble fhWions. LyophiIization of the dimethyl sulfoxide soluble supernatant and resuspension in water failed to reassociate the paired helical filaments, but did result in an insoluble precipitate. Analysis of the dimethyl sulfoxide solubilized paired helical filament &action by nuclear magnetic resonance revealed it to be composed of glycolipid in a form that was distinct brn similar fractions isolated &om normal aged control brains. The aggregation of an altered glycolipid to form paired helical fSlaments in Alzheimer disease could explain their insolubility. e,1991Academic PESS, inc.
Alzheimer disease (AD) is a progressive degenerative dementia which is characterized neuropathologically by the accumulation of neurofibrillary tangles (NFT), senile plaques (SP) and cerebrovascular amyloid (CVA) in the brain. The p/A4 amyloid fibrils which are present in SP and CVA have been solubilized and their amino acid sequence determined (1,2). However, the composition of paired helical filaments (PHF), the predominant structures in NFT and degenerating neurites of SP, has remained elusive. The difficulty in analyzing PHF stems from their high degree of insolubility in detergents and chaotropic salts (3,4), but this insolubility has allowed the extraction and purification of PHF from NFT from AD brains ($6).
II
*To whom correspondence should be addressed. 0006-291x/91
771
$1.50
Copyright 0 1991 by Academic Press, Inc. All rights of reproduction in any form reserved.
Vol.
181, No. 2, 1991
BIOCHEMICAL
AND BIOPHYSICAL RESEARCH COMMUNICATIONS
PHF have been assumed to be proteins which are resistant to solubilization, proteolysis and chemical cleavage (3,4). Although a number of studies have identified several proteins to be associated with the NFT (7-9), quantitative solubilization of the PHF core structures has not been demonstrated. Recent studies of PHF have used pronase digestion to remove their outer “fuzzy” coat, which was determined by amino acid sequence analysis to consist of the microtubule-associated protein, tau (10). This outer tau coat was shown to compose only 10% of the PHF, the remaining 90% being resistant to solubilization and thus precluding further analysis. PHF-containing NFT are one of the most consistent lesions in AD, and their density at autopsy has been correlated with the degree of clinical dementia (II); thus determining their structural composition would be a major step toward identifying the underlying cause of AD.
MATERIALS
AND MGFHODS
PHF Iso&tions. PHF were isolated from AD brain as previously described (6) with the following modifications. Routine preparations employed 200 g of NFT-rich gray matter and employed a Beckman 45Ti pre arative ultracentrifuge rotor. Prior to sucrose density adient centrifu ation, the crude N K- were resuspended in 100 mM Tris-HCI, pH 810 mM E DTA, 2 mM Ai Cl,, 0.5% SDS containing lOC&/ml pronase. The NFT were incubated at 42“ C for 18 hours, at which time the suspension was adjusted to 1% SDS and 0.1 M amercaptoethanol and boiled for 5 minutes. After recovery of the 1.4-2.0 M fraction, the final suspension of disaggregated PHF was dialyzed 18 hours against 0.1% dimethyl sulfoxide DMSO), followed by extensive dialysis against distilled water for 2 days at 4” C.TheP I-A were dried down in a Speed-Vat and weighed, To monitor the complete removal of sucrose, [14-C] sucrose was included in some of the gradients. Control brain fractions were obtained using aged control brains in parallel isolations. &&so Acti Analysis.The sample was hydrolyzed for 21 hours at 104“ C in 6N HCl in VUCUO. The hydrolysate was analyzed on a Waters Model 441 amino acid analyzer using the picotag program. Nitrogen Analysis. Elemental microanalysis was performed by Galbraith Laboratories, Knoxville, Tennessee. Electron Microsco . The PHF were negatively stained for electron microscopy as previously described (67 with 1% aqueous uranyl acetate. Enzyme-Linked fmmunosarbant Assay (ELZSA). The DMSO solubilized PHF core structures were centrifuged at 10,000 x g for 10 minutes at 2.5” C to remove insoluble lipofuscin particles. The lipofuscin was resuspended in distilled water and both it and the DMSO supernatants were lyophilized in a Speed-Vat. The samples were diluted in 0.1 M sodium bicarbonate buffer, H 9.5 and 100~1 dried in each well at 37” C for 16 hours. The wells were blocked with 1%o BSA in 10 mM Tris-HCl, pH 82,200 mM NaCl TBS-200) with 0.01% Tween 20 for one hour at 37” C, followed by three washings with TBS-2 06 /Tween. Al28 antiserum (6) was serially diluted in blockin buffer and incubated for one hour at 37” C. The wells were washed four times with TB iii-2OOKween. Horseradish eroxidase-labeled swine anti-rabbit immunoglobulins were diluted to 1:lOOO in blockin bu f!fer and incubated for one hour at 37” C. The wells were washed three times with TB 8 -2OO/rween and twice with TBS-200. The color was developed with 100~1 per well of 0.8 mg/ml o-phenylene diamine in 0.1 M sodium citrate phosphate buffer, pH 5, containing 0.03% hydrogen peroxide for 30 minutes at ambient temperature. The reactions were stopped by the addition of 10~1 of 10 N sulfuric acid and absorbance read at OD, . Nuclear Magnetic Resonance (Nh4R) Anafysis. PHF were soq)ubilized by heating at 65” C for 3 hours in d -DMSO/D,O (9:l) while stirring. 500 MHz ‘H NMR s ectra were run overmg5 t on a General Electric GN-500 NMR system. The residual I! ,O resonance of the solvent occuring at about 3.30 ppm was saturated rior to the observe pulse and acquisition. Spectra are reported using the residual R ,d,-DMSO solvent resonance at 2.49 ppm as an internal chemical shift reference. 772
Vol.
181, No. 2, 1991
BIOCHEMICAL
AND BIOPHYSICAL
RESEARCH COMMUNICATIONS
RESULTSANDDISCUSSION Total amino acid analysis of the PHF core preparations showed the presence of amino acids (Table l), but these only accounted for about 4.76% of the total mass. Amino acid analysis of material isolated from aged normal control brains showed similar amounts of protein (4.36%), although the individual values for each amino acid differed. These small amounts of amino acids are most likely due to residual lipofuscin granules, indicating that the PHF core structures are most likely composed of nonproteinaceous material. An analysis of total elemental nitrogen showed that the PHF core preparations contained only 2.25% N, again much lower than the expected value of 15.26% N for a protein composed of equal amounts of all 20 amino acids. To exclude the possibility of residual sucrose contaminating the PHF preparations, [14-C]sucrose was included in the sucrose gradients in several of the preparations. Dilution of the PHF fraction from the 1.4-2.0 M interface in distilled water followed by centrifugation removed 99.6% of the radioactivity, and the dialysis completely removed the remaining radioactivity. Electron microscopy (EM) of the PHF fractions revealed intact twisted filaments (Fig 1A) which were used as starting material for solubilization in DMSO. After heating in DMSO for 3 hours, the brown lipofuscin fraction was separated from the solubilized PHF by centrifugation. The supernatant contained no visible intact PHF core structures (Fig 1B) and appeared identical to the control supernatant (Fig 1C). The DMSO insoluble
TABLE
1
Total Amino Acid Analysisof the AkzheimerBrain Samplesand Similar Fractionsfrom Aged Control Brain Amino Acid
Y,i%%T
Ala Pro Tr dl Met
10.45kO.64 11.95k1.48 7.40-e1.56 6.85~~0.78 4.5020.28 4.1OkO.85 3.8521.06 4.8520.21 2.4020.14 4.45-cO.78 8.9O-cO.42 2.55rt0.35 2.3523.32 5.5521.20 6.7020.42 7.15k2.05 6.OO-cO.42 ND
CYS Ile Leu
Percent protein of total mass
4.76%
Aged C;:;o&Sample 5.60~0.71 8.6020.28 6.4020.14 5.90r0.14 2.6520.07 390~0.28 7.3020.42 7.2540.50 3.00*0.14 5.0520.21 6.95k0.49 1.75kO.21 3.6522.90 4.90k0.28 12.05kO.50 9.75kO.07 5.40-0.42 ND 4.36%
ND = Not Determined.Data are the averageof two experiments(mean-c standarddeviation). 773
Vol.
181, No. 2, 1991
BIOCHEMICAL
AND BIOPHYSICAL RESEARCH COMMUNICATIONS
Electron microscopy of A) the PHF fraction of AD brain, (B) the 0 supernatantfrom the PHb fraction, (C) the DMSO/D 0 su ernatant from the norm& a ed brain, (D) the DMSOQO insolublepellet from tie Pl-fF fraction, (E) the DMSO h z0 insoluble pellet from the normal aged control brain, (F) the DMSO/D 0 soluble PHF material after lyophilization and resuspensionin distilled Hz, andr (G) the DMSO/D,O soluble normal aged brarn material after lyophilization and resuspension1x1distilled HzO. Both sampleswere isolated using identicalprocedures.Line scale= 2.50nm. 774
Vol. 181, No. 2, 1991
BIOCHEMICAL
AND BIOPHYSICAL
RESEARCH COMMUNICATIONS
TABLE 2
Percent
PHF Immunoreactivity
in the DMSO Soluble and Insoluble Fractions of Alzheimer Brain and Similar Fractions from Aged Control Brain . 2 AD Brain Fractions’ C ntrol Brain Fractrnns Insoluble Soluble Insoluble Soluble 95.7%
4.3%
5.5%
0.7%
’ Total immunoreactivity of the AD brain fraction was measured by ELISA and calculated for the total volume of both the soluble and insoluble fractions, divided by the sum of both. 2 Percent immunoreactivity for the control brain fractions, relative to the PHF fractions, was measured by ELISA and calculated for the total volume of both the soluble and insoluble fractions, divided by the sum of the PHF immunoreactivity.
lipofuscin fractions from the AD and control brains contained a coarse brown granular material (Figs 1D and lE), but no PHF remained in the AD fraction. The DMSO soluble fractions from both the AD and control brains were lyophilized to dryness and resuspended in distilled water. The PHF soluble fraction yielded an insoluble material that appeared lattice-like by EM (Fig lF), but showed no reformed PHF structures. The control also contained insoluble material, but it was not well formed (Fig 1G). Lyophilization of non-solubilized PHF, followed by resuspension in distilled water had no effect on the morphology of the PHF (data not shown). Analysis of the DMSO soluble and insoluble fractions from the AD brain by ELISA revealed that 95.7% of the PHF immunoreactivity was present in the DMSO soluble fraction, with only 4.3% remaining in the insoluble fraction (Table 2). There was only 5.5% and 0.7% PHF immunoreactivity in the DMSO soluble and insolubIe fractions, respectively, from the aged control brain relative to the PHF fraction from the AD brain. Although most of the PHF immunoreactivity appears to be in the soluble fraction, it must be noted that it is only the relative percentage compared to the insoluble values. It is not known if DMSO treatment altered antigenic sites and reduced the antibody’s ability to recognize the solubilized PHF structures. Fig. 2A shows the 500 MHz proton NMR spectrum of a typical DMSO/D,O solubilized PHF fraction isolated from an AD brain. For comparison, the spectrum of a similarly prepared sample from normal aged brain is shown in Fig. 2B. Both spectra resemble high-field NMR spectra of human brain gangliosides (12-19). Downfield resonances in these spectra were previously assigned to the olefinic protons (5.2-5.5 ppm), the anomeric protons of the carbohydrate moieties (4.0-5.0), sphingosine protons Rl-R4 and the remaining carbohydrate pyranosyl ring protons (3.0-4.0 ppm). Resonances occuring in the upfield region were similarly assigned to the acetamido methyl protons of the carbohydrate moieties (1.7-1.9 ppm), the allylic and the ar-carbonyl methylene protons (1.9-2.2 ppm), and the ceramide methylene (1.2 ppm) and methyl protons (0.85 ppm). Based upon these previous assignments and upon a comparison of the resonances in the Fig. 2 spectra with those of a commercially obtained ceramide, the triplet at 0.85 ppm in both spectra can be assigned to the ceramide methyl protons while the singlet at 2.05 ppm and 775
Vol.
181, No. 2, 1991
BIOCHEMICAL
AND BIOPHYSICAL
RESEARCH COMMUNICATIONS
B
I ““l”“l”“l”“l”“l”‘~1 6 5
A. 4 GM
2
1
0
Figure 2. 500MHz NMR of (A) the PHF fraction of AD brain, and (B) normal apedbrain. Both sampleswere isolatedusingidentical procedures.Sampleswere dissolvedm a.%10 mixture of DMSOQO. Spectra were acquired at 333” K. The structure shown m the top figure is that of glucoceramideas determinedin ref.12 and is shownas a guide for resonanceassignments.
the triplet at 2.20 ppm in the spectra of the PHF sample can be assigned to the allylic and the cu-carbonyl methylene protons of the ceramide moiety. While the 2.05 ppm allylic proton singlet is much less prominent in the spectrum of the normal aged brain sample, there appear to be two broad based groups of resonance at about 5.25 and 5.5 ppm which, by comparison to the ceramide standard, can be assigned to the olefinic protons. In spectra of other PHF samples prepared in an identical manner, there is substantial variation in the intensity ratio of the allylic proton to methylene proton ceramide resonances, indicating a variation in the degree of unsaturation in the alkyl chains. 776
Vol.
181, No. 2, 1991
BIOCHEMICAL
AND BIOPHYSICAL RESEARCH COMMUNICATIONS
Neither of the spectra of Fig. 2 show resonances in the region immediately upfield of the ceramide aliylic protons at 2.05 ppm. This suggests the absence of any acetamido carbohydrate residues typically found in other brain gangliosides and glycoproteins (i.e., N-ace@ glucosamine, N-ace@ galactosamine or sialic acid) and is consistent with lectin studies on isolated PHF (20). In both spectra, however, there is a well resolved doublet present in the spectral region characteristic of anomeric proton resonances of attached carbohydrates (4.68 ppm). The magnitude of the coupling constant for this resonance (3.1 Hz) suggests the sugar has assumed an cu-anomeric ring configuration. The shift and coupling constant for this resonance, when taken together, are similar to those observed for cu-D-galactose residues in glycolipid samples previously reported (21,22). Doublets also occur in both spectra which fall just outside the region typical of anomeric protons of other gangliosides in similar solvent systems. These include those in both spectra at 5.10 ppm (J,, = 5.0 Hz) and in the spectrum of the normal aged brain sample at 5.20 and 5.23 ppm (J,, = 3.7 and 5.1 Hz, respectively). We also acquired NMR spectra in DMSO of the AD lipofuscin fraction, obtained from the 1.0-1.2 M interface of the sucrose gradients, which is often a contaminant of the PHF fraction (data not shown). The spectra is absent of resonance which could be assigned to anomeric protons of monosaccharides residues and closely resembles the ‘H NMR spectrum previously reported for purified lipofuscin from normal human brains (23). The most striking difference between the PHF spectrum and that of the normal aged brain sample is the appearance of the sphingosine Rl-R4 and pyranosyl ring proton resonances between about 3.2 and 3.8 ppm. The general overall broadened nature of the envelope in the PHF spectrum is consistent with broadened linewidths and shorter Tz relaxation times characteristic of polymerized or aggregated samples having slower rotational relaxation times. Although PHF have been assumed for a number of years to be composed of a highly insoluble protein(s), certain properties tended to argue against their being proteinaceous at all. For example, we found the core structures to be resistant to numerous proteases, chemical cleavages and partial acid hydrolysis. This, along with the low percentage composition of amino acids and elemental nitrogen, fails to support their being composed of significant amounts of protein. The present results indicate that they may, instead, be composed of a glycolipid, possibly a glycoceramide, capable of assembling into PHF. Ceramides are insoluble in non-polar solvents and self-aggregation or polymerization would result in a large, highly insoluble molecule. There have been reports of altered membrane fluidity in AD which correlates with the severity of neuropathology (24). Brain lipids reported to be altered in AD include gangliosides (25), phosphoinositides (26), phosphomonoesters (27), and phosphatidylethanolamines (28). Antibodies to gangliosides have also been shown to immunolabel both SP (29, 30) and NFT (30,31). A monoclonal antibody which reacts with sulfatides and gangliosides showed high levels of immunoreactivity in cell bodies and degenerating neurites in AD brain (32). Electron microscopic analysis of brain biopsies 777
Vol.
181, No. 2, 1991
BIOCHEMICAL
AND BIOPHYSICAL
RESEARCH COMMUNICATIONS
from Alzheimer patients showed that PHF may arise at or from the cell membrane, a lipid-rich structure (33). It has been shown that lipids can aggregate to form filaments in vitro (34), as well as in aged erythrocytes (35). These filaments can exist as double helical filaments (36). Should PHF cores be confirmed to be composed of aggregated or polymerized glycolipid(s), it would explain their relative insolubility and could offer explanations of other changes observed in AD. For example, alterations in glycolipid metabolism could destabilize neuronal membranes, making the amyloid precursor protein more susceptible to proteolytic cleavage, resulting in the deposition of amyloid, which is also characteristic of AD pathology. Finally, the search for a defect in glycolipid metabolism represents a new avenue of research into the etiology of AD. ACKNOWLEDGMENTS
Supported by an Advanced Technology Education Coordinating Board, a National Research Center Grant (AG08013), a French American Health Assistance Foundation grant grant (AT-1162) (WJG).
Grant (003660-108) from the Texas Higher Institutes of Health Alzheimer’s Disease Foundation for Alzheimer Research grant, a (900066) and a Robert A. Welch Foundation
REFERENCES
Glenner, G.G. and Won , C.W. (1984) Biochem. Biophy. Res. Comm. 120,885-890. :: Masters, CL, Simm, G., &v einman, G., Multhaup, G., McDonald, B.L. and Beyreuther, K. 1985) Proc. Natl. Acad. Sci. USA 82,4245-4249. &elkoe, D.J., Ihara, Y. and Salazar, F.J. (1982) Science215,1243-1245. :: Branbury Selkoe, D.J., Ihara, Y., Abraham, C., Rasool, C.G. and McCluskey, A.H. (1983) Report 15: Biological Aspects of Alzheimer’s Disease, Cold Spring Harbor Sympostum, 125-135.
Y., Abraham, C. and Selkoe, D.J. (1983) Nafure 304,727-730. 2 S3Ihara, arkman, D.R., Hammon, K.M. and White, C.L. III (1990) J. Histochem. Cytochem. ii ,703-715. 7. &$kig8S., w
Joachim,
CL. and Selkoe, D.J. (1986) Proc. Natl. Acad. Sci. USA 83,
8. Grundke-I bal I I bal K Quinlan M., Tung, Y.C., Zaidi, MS. and Wisniewski, H.M. 1988 J. &di. a&. 28i, 6084-6089. Mori, i-I ., Kondo, J. and Ihara, Y. (1987) Science235, 1641-1644. zi. Wischik, C.M., Novak, M., Thogersen, H.C., Edwards, PC., Runswick, M.J., Jakes, R., Walker, J.E., Milstein, C., Roth, M. and Klug, A. (1988) Proc. Natl. Acad. Sci. USA, 85,4506-4510. 11. Crystal, H., Dickson, D., Fuld, P., Masur, D., Scott, R., Mehler, M., Masde, J., Kawas, C., Aronson, M. and Wolfson, L. 1988) Neurology 38,1682-1687. 12. Koerner,, T.A.W.,Jr., Prestegard, J.i-I ., Demou, P.C. and Yu, R.K. (1983) Biochemtstry 22: 2676-2687.
13. Koerner, B&hem&try
T.A.W.,Jr.,
Prestegard,
J.H.,
Demou,
PC.
and
Yu,
R.K.
(1983)
222687-2690.
14. Li., Y.-T., Hirabayashi, Y., DeGas eri, R., Yu, R.K., Ariga, T., Koerner, T.A.W. and IA, S.-C. 1984)J. Biol. Chem. 259:8f 80-8985. 15. Dabrows L i, J., Hanfland, P. and Egge, H. (1982) Methods in Enzymology, Vol. 83 Ed. V. Ginsbur ), Academic Press, New York, 69-86. 16. 6 abrowski, J., I-f anfland, P. and Egge, H. 1980) Biochemistry 19:5652-5658. 17. Hanfland, P., E e, H., Dabrowski, U., L uhn, S., Roelcke, D. and Dabroski, J. (1981) Biochemtstry 20:8 10-5319. 778
Vol. 181, No. 2, 1991
BIOCHEMICAL
AND BIOPHYSICAL
RESEARCH COMMUNICATIONS
18. Scarsdale, J,N., Prestegard, J.H., and Yu, R.K. (1990) Biochemists 299843-9855. 19. Acquotti, D., Poppe, L., Dabrowski, J., von der Lieth, C.-W., Sonnino, S. and Tettamanti, G. 1990) J. Am. Chem. Sot. 112:7772-7778. 20. S arkman, D.k ., Hill, S.J. and White, CL., III (1990) Acta Neuropathol. 78 ~640-646. Scarsdale, J.N. and Prestegard, J.H. (1986) Carbohydr. Res. 155:45-56. % Inagaka, F., Kohda, D., Kodama, C. and Suzuki, A. 1987) FEBS Lett. 212:91-97. Taubold, R.D., Siakotos, A.N. and Perkins, E.G. (19(75) Lip& l&383-390. E Zubenko, G. (1986) Bruin Res. 385,115-121. 25: Crino, P.B., Ullman, D., Vogt, B.A., Bird, ED. and Volicer, L. (1989) Arch. Neural. 46,398-401. Stokes, C.E. and Hawthorne, J.N. (1987) J. Neurochem. 48,1018-1021. 2 Pettegrew, J.W., Moossy, J., Withers, G., McKeag, D. and Panchalingam, K. (1988) J. Neuropathol. EIxp. Neural. 47,235-248. 28. Sdderber , M., Edlund, C., Kristensson, K. and Dallner, G. (1991) Lipids 26~421-42 s . 29. Iwamoto, N., Suzuki, Y., Makino, Y., Haga, C., Kosaka, K. and Iizuka, R. (1990) Brain Rex 522,152-156. 30. Takahashi, H., Hirokawa, K., Ando, S. and Obata, K. (1991) Actu Neuropathol. 81:626-631. Emory, C.R., Ala, T.A. and Frey, W.H. II (1987) Neurology 37,768-772. 2 Majocha, R.E., Jungalwala, F.B., Rodenrys, A. and Marotta, C.A. (1989) J. Neurochem. 53,953-961. 33. 73ra i l”;:., Paula-Barbosa, M and Roher, A. (1987) Neuropathol. Appl. Neurobiol. d 34. Rddoiph, AS., Ratna, B.R. and Kahn, B. (1991) Nature 352:52-55. 35. Bessis, M. (1973) Livirlg Blood Cells and their Ultrastructure, Springer-Verlag, New York, 60-61. 36. Lin, K.-C., Weis, R.M. and McConnell, H.M. (1982) Nature 296164-165.
779
