Abnormal Erythrocyte Anion Exchange in Alzheimer Disease.

Alzheimer disease (AD) is defined pathologically by the presence of argyrophilic plaques and tangles in the brain.[1,2] Clinically, the onset of AD is insidious, leading to the widespread notion that the pathologic processes begin years before the dementia.[3] Some genes cause AD by promoting the accumulation of [BETA]-amyloid in the plaques.[4] Typically, these genes occur in families with a strong history of AD and lead to an early onset. It has been estimated that such genes account for 2% to 10% of all AD cases.[4-6] The cause of AD in the remaining cases is unknown. There is no pattern of gross pathologic condition in organs other than the brain, and most approaches follow the trail of some unique aspect of brain physiology, either biochemical pathways[7-10] or the blood-brain barrier.[11] Nonetheless, there are manifestations of AD outside the brain, and these have been studied particularly in blood and skin.[5,12]

In 1991, Bosman et al[13] showed that there was an abnormality of [MATHEMATICAL EXPRESSION NOT REPRODUCIBLE IN ASCII] transport in red blood cells from patients with AD but not from patients with multi-infarct dementia or from age-matched, healthy controls. In the red blood cells, [MATHEMATICAL EXPRESSION NOT REPRODUCIBLE IN ASCII] transport is mediated by band 3.[14,15] Band 3 is a membrane glycoprotein and derives its name from its relative position after electrophoresis. It is the prototype of the anion exchanger gene family, hence its other name, AE1. There are 3 well-characterized genes in this family, each located on a different chromosome and each transcribing multiple forms of messenger RNA.[16,17] In humans, erythrocyte band 3 serves 3 functions. The cytoplasmic domain binds to the cytoskeleton, which in the red blood cells consists of a netlike mesh subjacent to the lipid bilayer.[18] The membrane domain mediates [Cl.sup.-]/ [MATHEMATICAL EXPRESSION NOT REPRODUCIBLE IN ASCII] exchange physiologically; other anions, such as [MATHEMATICAL EXPRESSION NOT REPRODUCIBLE IN ASCII], may cross the membrane through band 3 and have been used to probe its function.[14,15] Changes in antigenic determinants on band 3 signal red blood cell senescence and lead to the cell's removal.[19] Although there is polymorphism of the band 3 gene, the common genotypes do not manifest themselves by different rates of anion exchange.[15,17] Those mutations that affect band 3 function, when not fatal, result in anion exchange that is at least 80% of normal.[15] In contrast to other tissues, there is no known pathway in vivo for regulating anion exchange in the red blood cells.[15] Erythrocyte anion exchange, therefore, remains rather constant with respect to both inter-individual and intraindividual variation. The effect of AD on red blood cell anion exchange is remarkable, and its mechanism is of considerable interest.

We have developed a technique for studying [Cl.sup.-]/ [MATHEMATICAL EXPRESSION NOT REPRODUCIBLE IN ASCII] exchange in resealed erythrocyte ghosts, that is, red blood cells from which most hemoglobin has been removed.[20] Our method uses a fluorescent probe of [[Cl.sup.-]] to monitor [Cl.sup.-]/[MATHEMATICAL EXPRESSION NOT REPRODUCIBLE IN ASCII] exchange in a stopped-flow apparatus at 4 [degrees] C. When ghosts that have been resealed with the probe and NaCl are rapidly mixed with [Cl.sup.-]/[MATHEMATICAL EXPRESSION NOT REPRODUCIBLE IN ASCII] solutions, the time course of fluorescence reports the intracellular [[Cl.sup.-]] and its approach to a new equilibrium value. This time course is not a single exponential function at 4 [degrees] C, and we have developed a model to account for this result.[20] Band 3 mediates anion exchange through a sequence of steps known as the ping-pong mechanism.[14,15] A transfer site binds either [Cl.sup.-] or [MATHEMATICAL EXPRESSION NOT REPRODUCIBLE IN ASCII], transports it across the membrane through a conformational change in band 3, and releases it. Another anion binds to the site, and a second conformational change carries it back to the initial face of the membrane. An unloaded transfer site can also cross the membrane, a pathway known as slippage. Our model extends the usual ping-pong model and predicts that the [[Cl.sup.-]] inside the ghosts, [C.sup.i](t), is given by

[MATHEMATICAL EXPRESSION NOT REPRODUCIBLE IN ASCII] (1)

where t is the time after mixing and C is the equilibrium [[Cl.sup.-]]. Empirically, the parameters [a.sub.1], [k.sub.1], [a.sub.2], and [k.sub.2] are determined by fitting the time course of fluorescence. The number of transfer sites on either face of the membrane depends on composition of the internal and external media, because each species of anion has its own affinity for binding and rate of transport. The balance between binding and transport determines the relative number of transfer sites on either face. When the external medium changes suddenly, as in our stopped-flow experiments, the number of transfer sites re-equilibrates, and this process determines the empirical rate constant [k.sub.1]; [k.sub.2] reflects the discharge of anion gradients. Theoretically, the model predicts certain functional relationships that should exist between each parameter of equation 1 and the equilibrium [[Cl.sup.-]] or [MATHEMATICAL EXPRESSION NOT REPRODUCIBLE IN ASCII]. According to our model, the empirical rate constant [k.sub.1] is the sum of the 2 theoretical slippage rate constants, inside to outside and outside to inside. In other words, the model predicts that

[k.sub.1] = constant (2)

and is independent of [[Cl.sup.-]] or [MATHEMATICAL EXPRESSION NOT REPRODUCIBLE IN ASCII]. For the other empirical rate constant, the model predicts that [k.sub.2] = f[MATHEMATICAL EXPRESSION NOT REPRODUCIBLE IN ASCII], [[Cl.sup.-]]. Because of our experimental design, the model predicts more specifically that

1/[k.sub.2] = [MATHEMATICAL EXPRESSION NOT REPRODUCIBLE IN ASCII] + b (3)

where m and b consist of several theoretical rate constants of the model. We will refer to the constant of equation 2 and the slope and intercept of equation 3 as model parameters. Having confirmed the predictions of our model in erythrocyte ghosts from healthy young adults,[20] we now use it to examine subjects with AD and age-matched controls. The present study focuses on the following questions: Does the model fit both groups? Do model parameters distinguish the 2 groups? Can model parameters usefully classify the individual subjects?

SUBJECTS AND METHODS

Subjects

Subjects with dementia met the NINCDS-ADRDA (National Institute of Neurological and Communicative Disorders and Stroke-Alzheimer's Disease and Related Disorders Association) criteria for probable AD[21]; one of the authors (A.S.) assessed all patients with AD at McLean Hospital. Age-matched controls were either spouses of patients with AD (assessed by A.S.) or volunteers (assessed by F.A.G.) at an ambulatory clinic. No subject in the study was being treated with antipsychotics or antidepressants. All medical problems that required treatment excluded a subject, with the exception of mild and well-controlled hypertension, glaucoma, or hypothyroidism. Smoking history was not considered in selection of subjects. All subjects with dementia scored less than 4 on the Hachinski Ischemic Scale.[22] Subjects with AD consisted of 2 men and 9 women (mean age, 76 years; range, 63-83 years). Age-matched controls were 4 men and 7 women (mean age, 72 years; range, 62-81 years). Some measurements were also performed on young adult volunteers from the ambulatory clinic (2 men and 4 women; mean age, 32 years; range, 23-45 years). The protocol for conducting this study has been approved by the institutional review boards of both McLean Hospital and Harvard Medical School. Subjects gave informed consent for phlebotomy and standards conformed to the Declaration of Helsinki. Blood samples used in this study were drawn at the time of diagnosis of AD. Subjects with AD scored an average of 17 [+ or -] 6 (SD) on the Mini-Mental State Examination.[23]

Averages of Individuals' Parameters(*)

Group                 [k.sub.1], [s.sup.-1]

Alzheimer disease     1.73 [+ or -] 0.11
Age-matched control   1 .1 7 [+ or -] 0.1 5

P (t test)                     <.01

Group                 1/[k.sup.2](0), s

Alzheimer disease     2.1 8 [+ or -] 0.1 9
Age-matched control   3.02 [+ or -] 0.28

P (t test)                      <.05

Group                       [Lambda]

Alzheimer disease     0.040 [+ or -] 0.01 5
Age-matched control   -0.038 [+ or -] 0.01 3

P (t test)                     <.001

(*) Values are mean [+ or -] SE.

Measurement of Anion Exchange

To follow the protocol of Bosman et al[13] as closely as possible, initial measurements were made on blood drawn into EDTA. However, the results exhibited a high degree of variability, and citrate was chosen as the anticoagulant.[24] All data shown have been obtained in citrate anticoagulant. The details of the method for preparing hemoglobin-free ghosts are given in our previous article.[20] Briefly, blood was kept on ice until use, which was always within 24 hours. Following 3 washes in isotonic sodium chloride solution, the cells were lysed in 5 mmol/L [Na.sub.2][HPO.sub.4], pH 8, and washed 3 more times in lysing buffer to remove hemoglobin. The resulting pellet of white ghosts was incubated in 10 mmol/L 6-methoxy-N-(3-sulfopropyl)quinolinium (SPQ; Molecular Probes, Eugene, Ore) for 5 min, and sufficient NaCl was added to restore isotonicity. The ghosts were then resealed by incubation at 37 [degrees] C for 1 hour and washed 3 more times in phosphate-buffered saline (PBS) to remove extracellular SPQ. The final pellet was resuspended to a 1% hematocrit in phosphate-buffered saline (pH 7.4). The chloride-loaded ghosts were mixed with equal volumes of [Cl.sup.-]/[MATHEMATICAL EXPRESSION NOT REPRODUCIBLE IN ASCII] solutions in a stopped-flow apparatus (model SF.17MV; Applied Photophysics, Leatherhead, UK). [Cl.sup.-]/[MATHEMATICAL EXPRESSION NOT REPRODUCIBLE IN ASCII] exchange was followed by the fluorescence of intracellular SPQ (excitation 350 nm; emission measured through a 500-nm Corion cutoff filter). All measurements were made at 4 [degrees] C. The fluorescence of SPQ was converted to [[Cl.sup.-]] as previously described and, under the conditions of these measurements, was a nearly linear function of [[Cl.sup.-]].[20] The time course was fit to equation 1 by the method of least squares. The [Cl.sup.-]/[MATHEMATICAL EXPRESSION NOT REPRODUCIBLE IN ASCII] mixtures were such that [[Cl.sup.-]] + [MATHEMATICAL EXPRESSION NOT REPRODUCIBLE IN ASCII] = 150 mmol/L, and the measurements are labeled according to the equilibrium [MATHEMATICAL EXPRESSION NOT REPRODUCIBLE IN ASCII] in millimoles per liter, for example, as [k.sub.2](11.25). Seven to 10 replicate measurements were made at each [MATHEMATICAL EXPRESSION NOT REPRODUCIBLE IN ASCII].

Statistics

Model parameter [k.sub.1] was calculated for each subject as the average of measurements at all bicarbonate concentrations; model parameter 1/[k.sub.2](0) was obtained by a weighted least-squares fit of equation 3. Group estimates were obtained by pooling all measurements for AD and age-matched control. The Fisher linear discriminant,[25] designated as [Lambda], was calculated for the 2 parameters and adjusted to a zero cutoff.[26] A receiver operating characteristic curve[27] was then constructed using the Fisher linear discriminant. Statistical significance was determined by Student's t test for the comparison of means and by the method of Pearson for the correlation of variates.

RESULTS

Figure 1, obtained by pooling all measurements under the conditions specified, shows the results for the AD and age-matched control groups. The predictions of our model, equations 2 and 3, fit the data from both groups. The average value of [k.sub.1] for the AD group is 1.83 [+ or -] 0.09 [s.sup.-1], whereas [k.sub.1] for the controls is 1.15 [+ or -] 0.07 [s.sup.-1] (P [is less than] .001). The regression line for the plot of 1/[k.sub.2] against [MATHEMATICAL EXPRESSION NOT REPRODUCIBLE IN ASCII] (equation 3) is 1.43 + 0.13 [MATHEMATICAL EXPRESSION NOT REPRODUCIBLE IN ASCII] for the AD group and 2.24 + 0.12 [MATHEMATICAL EXPRESSION NOT REPRODUCIBLE IN ASCII] for the controls; although the slopes of these lines are the same, the difference between the intercepts is significant (P [is less than] .001; SE for each intercept is [+ or -] 0.06 s). Therefore, of the 3 model parameters defined by equations 2 and 3, 2 distinguishes AD from the control.

To assess the potential for classifying individuals, we analyzed the values of the model parameters as determined for each subject. In this analysis, [k.sub.1] denotes the parameter of equation 2 assigned to each subject and 1/[k.sub.2](0) will be used for the intercept b of equation 3. The results are displayed in Figure 2; averages obtained by this analysis are summarized in the Table. There is no significant correlation of [k.sub.1] and 1/[k.sub.2](0) (r = -0.3; P [is greater than] .1), and the 2 parameters are, therefore, independent predictors of AD. Calculating the Fisher linear discriminant for the AD and age-matched control groups and adjusting for zero cutoff gives

[Lambda] = 0.089 [1/[k.sub.2](0)] + [0.089k.sub.1] (4)

The results for [Lambda] are also summarized in the Table.

The hypothesis that patients with AD had normal anion exchange kinetics ([Lambda] [is less than] 0) before the onset of dementia requires that measurements on young adults should also differ from those on patients. The difference in [Lambda] between patients with AD and young adults reached statistical significance at N = 6 (P [is less than] .002). The mean value of x for young adults is -0.055 [+ or -] 0.011; this does not differ significantly from the value for age-matched controls (Table), and points from young adults are included in the plot of Figure 2 for comparison. There is no significant correlation between [Lambda] and age within either the AD group (r = -0.5; P [is greater than] .1) or the control group (r = 0.2; P [is greater than] .1) (Figure 3). Furthermore, there is no significant correlation between h and severity of dementia as measured by the Mini-Mental State Examination (r = 0.1; P [is greater than] .1; data not shown).

[Figure 3 ILLUSTRATION OMITTED]

To compare our findings with previous work,[13] we also analyzed our data according to Michaelis-Menten kinetics; the Lineweaver-Burk plot of the pooled data is shown in Figure 4, which leads to the following estimates: [K.sub.1/2] - AD 16 [+ or -] 9 mmol/L, control 69 [+ or -] 10 mmol/L; [v.sub.max] (maximal velocity) - AD 19 [+ or -] 4 mmol/L [Cl.sup.-]/s, control 24 [+ or -] 10 mmol/L [Cl.sup.-]/s. Assigning Michaelis-Menten parameters to each subject gives comparable estimates. The [K.sub.1/2] of the patients with AD is 59% of control (AD: 32 [+ or -] 4 mmol/L; control: 54 [+ or -] 8 mmol/L; P [is less than] .05); [v.sub.max] of the patients with AD is 121% of control but this difference does not reach statistical significance (AD: 17 [+ or -] 1 mmol/L [Cl.sup.-]/s; control: 14 [+ or -] 1 mmol/L [Cl.sup.-]/s).

COMMENT

Several studies have examined red blood cell physiology in AD. There is no difference in hemoglobin level[13,28] mean corpuscular volume,[28] or reticulocyte count[23] between patients with AD and age-matched controls. Studies of abnormalities of red blood cell choline levels[29,30] and cholinesterase activity[31] in AD have lead to reports of various and conflicting effects. Acanthocytes are found more frequently in peripheral smears from patients with AD; in some individuals, this increase was noted a decade before the onset of dementia.[28] Walter and Widen[24] have demonstrated that electrophoretic mobility of red blood cells from patients with AD differs from healthy controls and that this difference depends on the anticoagulant used, appearing in citrate and oxalate but not in EDTA or heparin. In 1991, Bosman et al[13] found altered erythrocyte [MATHEMATICAL EXPRESSION NOT REPRODUCIBLE IN ASCII] exchange at 37 [degrees] C in patients with advanced AD; red blood cells from patients with multi-infarct dementia behaved in the same manner as controls. Using Michaelis-Menten kinetics to analyze their data, they determined that the [K.sub.1/2] for patients with AD was 62% of that for controls (P [is less than] .05) and the [v.sub.max] was 117% of that for controls (nonsignificant). In the erythrocyte, band 3 mediates [MATHEMATICAL EXPRESSION NOT REPRODUCIBLE IN ASCII] transport; the relationship between [MATHEMATICAL EXPRESSION NOT REPRODUCIBLE IN ASCII] transport and [MATHEMATICAL EXPRESSION NOT REPRODUCIBLE IN ASCII] exchange by band 3 has been extensively studied, and the 2 processes correlate rigorously.[14,15] Therefore, our Michaelis-Menten analysis (Figure 4) quantitatively corroborates the findings of Bosman and colleagues[13] for the comparison of AD to age-matched control. In contrast to their findings in advanced AD, Bosman and colleagues[32] detected no abnormality of [MATHEMATICAL EXPRESSION NOT REPRODUCIBLE IN ASCII] exchange in erythrocytes from patients with mild AD. Using EDTA as an anticoagulant, we inadvertently confirmed this finding as well (see "Subjects and Methods"). Altered band 3 kinetics is, therefore, the first red blood cell abnormality in AD to be corroborated by 2 independent laboratories. Furthermore, erythrocyte band 3 kinetics display a similar abnormality in Down syndrome, in which the brains of patients that attain middle age exhibit pathologic changes similar to AD.[33]

[Figure 4 ILLUSTRATION OMITTED]

Against this background, the questions raised by our data divide into 2 broad categories. One is empirical, treating the model parameters as statistics without regard for units and optimizing the capacity for classifying subjects with an eye toward clinical utility. The other is pathophysiologic, dealing with possible mechanisms by which band 3 kinetics may be coupled to the underlying biochemical lesions of AD. We will focus on each category in turn.

The general predictions of the model for [Cl.sup.-]/[MATHEMATICAL EXPRESSION NOT REPRODUCIBLE IN ASCII] exchange as contained in equations 2 and 3 (on p 1041) fit the data from both the AD and control groups (Figure 1). In principle, therefore, the effect of AD on band 3 kinetics should be traceable to changes in the theoretical rate constants, although our current technique does not generate enough information to do this. In contrast, the paradigm of red blood cell aging, which provided a framework for discussing earlier studies, cannot accommodate the behavior of band 3. The changes in red blood cell electrophoretic mobility,[24] [MATHEMATICAL EXPRESSION NOT REPRODUCIBLE IN ASCII] transport,[13] and band 3 immunoelectrophoresis[13] found in AD are distinct from those seen in red blood cell aging. This difference between AD and normally aging red blood cells may serve as another illustration for Berg's thoughtful arguments that AD is not simply an exaggeration of the normal aging process.[34]

Furthermore, model parameters assigned to subjects by our technique discriminate well enough to suggest their usefulness in classifying individuals (Figure 2). The linear discriminant function [Lambda] clearly has no physical significance but classifies 82% of the subjects in accordance with the NINCDS-ADRDA criteria (Figure 3). Mahalanobis's generalized distance[35] between the 2 groups is 1.7, an improvement over the generalized distance of 0.7 obtained by using Michaelis-Menten kinetics. Receiver operating characteristic analysis (Figure 5) indicates that cutoffs may be chosen to achieve high specificity or sensitivity, depending on the intended clinical use.[27] Although the comparison of patients to age-matched controls is a useful starting point,[36] further studies must be done with patients experiencing other forms of dementia whose diagnosis has been confirmed by autopsy before the actual clinical utility of this approach can be determined.

[Figure 5 ILLUSTRATION OMITTED]

Because of the small number of subjects in this study, caution must be used in exploring correlations between [Lambda] and other biological variates. Nonetheless, [Lambda] does satisfy some of the necessary conditions to serve as a marker for the underlying lesion in AD. Unlike those with familial forms of AD, patients with the sporadic form may be assumed to have had normal physiologic processes beforehand. This consideration leads to the hypothesis that patients with AD had normal anion exchange kinetics ([Lambda] [is less than] 0) before the onset of disease, which in turn requires that measurements of young adults differ from those of patients. The data show that [Lambda] fulfills this necessary condition (Figure 3). If [Lambda] begins to increase before the onset of clinical AD, then [Lambda] in the elderly controls would tend to be higher than in young adult controls, because the elderly group presumably has some individuals with preclinical AD. The current study shows this trend ([Lambda] for the young control = -0.055; [Lambda] for the age-matched control = -0.038) but was not designed to detect this difference.

Changes in band 3 could explain the 2 other erythrocyte abnormalities reported in AD: acanthocytosis and altered cell electrophoretic mobility. A variant of band 3, 5HT, promotes the formation of acanthocytes,[37] presumably through the binding of band 3 to the cytoskeleton. However, lipoprotein binding sites have been described in the red blood cells,[38] and acanthocytosis raises the possibility that these might play a role as well. Trypsin treatment abolished the altered cell electrophoretic mobility seen in AD,[24] which suggests that a protein component of the membrane underlies the effect. Although it would be premature to hypothesize that band 3 is the primary cause of these effects, we will proceed on the assumption that it is an important contributing factor.

There are enough data in the literature to consider some possible causes of abnormal band 3 kinetics more or less likely than others. Several cell types from patients with AD have been reported to have altered membrane fluidity, both increased and decreased.[12,39,40] Although increased fluidity might lead to higher rate constants for membrane transport, it is unlikely that fluidity makes a major contribution to the abnormal kinetics of band 3, because 2 laboratories have independently found no difference in erythrocyte membrane fluidity between patients with AD and controls.[41,42] Chauhan et al[43] found that synthetic A[Beta] stimulates casein kinase I in erythrocytes, leading to increased phosphorylation of membrane proteins. Phosphorylation of membrane-associated cytoskeletal proteins affects red blood cell shape, and a disruption of the normal pattern of phosphorylation may contribute to the formation of acanthocytes.[44] Although casein kinase I phosphorylates the cytoplasmic domain of band 3,[45] such phosphorylation would not be expected to lead to the kinetics seen in AD.[15,46] Furthermore, 2 groups have failed to detect AB in erythrocytes from patients with AD,[28,47] and phosphorylation of band 3 by this mechanism is not likely to contribute to abnormal anion exchange.

That there is no known physiologic mechanism intrinsic to red blood cells for regulating anion exchange[15] points toward extracellular causes of the abnormal band 3 kinetics. Differences in blood levels of homocysteine[48] and an iron-binding protein p97[49] have been reported in AD, but there is no known mechanism to couple these findings to band 3 kinetics. The differential effects of anticoagulants suggest that either platelet activation[50] or complement[51] may play a role. Recent work has emphasized that red blood cells are not passively trapped in thrombi but interact with platelets to modulate their function.[52] Most studies have focused on the modulation of platelet function, and little is known about possible reciprocal effects of platelet activation on erythrocyte physiology. Immunoelectrophoresis of red blood cell membrane proteins from patients with AD demonstrates a different pattern of proteins cross-reacting with antibodies against band 3 in comparison to age-matched controls.[13] This finding suggests altered proteolysis of band 3, which could certainly cause altered anion exchange, but the mechanism of this proteolysis is not known. Platelet activation differs in AD[53] as does the ratio of amyloid precursor protein isoforms in platelet membranes.[54,55] The hypothesis that proteases associated with activated, abnormal platelets cleave erythrocyte band 3, thereby causing altered anion exchange kinetics, can explain all the findings to date. It is consistent with previous red blood cell experiments that show no intrinsic pathway for regulating anion exchange.[15] This hypothesis also explains why the alteration seen in AD differs from that of normal red blood cell aging, the only nonpharmacologic process known to affect red blood cell anion exchange in vivo. It accounts for the anticoagulant effect observed by us and others.[24] Platelet abnormalities become more pronounced the more advanced the dementia,[53,55] which could further explain why EDTA might mask the abnormality of erythrocyte [SO.sub.4] = exchange in mild[33] but not advanced AD.[13] This hypothesis also points to a line of investigation that could lead to a deeper understanding of events within the central nervous system. It has long been recognized that the platelet may reflect metabolic disturbances of the central neuron.[50] There is, however, a growing understanding of the connection between neurons and proteins classically associated with the red blood cell cytoskeleton: ankyrin,[56,57] spectrin,[57-60] band 4.1,[61] and band 3.[33,62] The generality emerging is that neurons express several isoforms of these proteins and appear to commit the red blood cell isoform to particular functions. For example, the erythrocyte isoform of spectrin has been detected only in the red blood cells, brain, and heart.[60] The isoform known as brain spectrin is mostly limited to axons, whereas the red blood cell isoform is found in nerve cell bodies and dendrites.[58] Therefore, membrane abnormalities in the red blood cells, whether resulting from interaction with platelets or some other mechanism, have the potential to serve as a model for pathologic events within the central nervous system. It would otherwise be a singular occurrence for red blood cell phenomena to be remote from the underlying lesions of AD and yet so clearly to distinguish it.

This work was supported in part by the Annie Laurie Charitable Trust Pilot Research Grant (IIRG93-026) from the Alzheimer's Association (A.K.S.), by a grant from the American Medical Association Education and Research Foundation (F.G.), by grant AG09301 from the National Institute on Aging (A.S.), and by a gift from Philomen T. Marvell, MD.

We thank Agnes Janoshazi and Ladislav Volicer for critical discussions, Sandra Cole for technical assistance, and Professor Solomon Snyder for kindly sending us a preprint of his manuscript.

References

[1.] Hyman BT. New neuropathological criteria for Alzheimer disease. Arch Neurol. 1998;55:1174-1176.

[2.] The National Institute on Aging and Reagan Institute Working Group on Diagnostic Criteria for the Neuropathological Assessment of Alzheimer's Disease. Consensus recommendations for the postmortem diagnosis of Alzheimer's disease. Neurobiol Aging. 1997;18(S4)S1-S2.

[3.] Cummings JL, Vinters HV, Cole GM, Khachaturian ZS. Alzheimer's disease. Neurology. 1998;51 (suppl 1):S2-S17.

[4.] Blacker D, Tanzi RE. The genetics of Alzheimer's disease: current status and future prospects. Arch Neurol. 1998;55:294-296.

[5.] Relkin N. Diagnostic markers in Alzheimer's disease. In: Folstein MF, ed. Neurobiology of Primary Dementia. Washington, DC: American Psychiatric Press Inc; 1998:271-288.

[6.] Gandy S, Greengard P. Processing of Alzheimer A[BETA]-amyloid precursor protein: cell biology, regulation, and role in Alzheimer's disease. Int Rev Neurobiol. 1994;36:29-50.

[7.] Goedert M. Tau protein and the neurofibrillary pathology of Alzheimer's disease. Trends Neurosci. 1993;16:460-465.

[8.] Selkoe DJ. Normal and abnormal biology of the [Beta]-amyloid precursor protein. Annu Rev Neurosci. 1994;17:489-517.

[9.] Roses AD. Apolipoprotein E alleles as risk factors in Alzheimer's disease. Annu Rev Med. 1996;47:387-400.

[10.] Vickers JC. A cellular mechanism for the neuronal changes underlying Alzheimer's disease. Neuroscience. 1997;78:629-639.

[11.] de la Torre JC. Cerebromicrovascular pathology in Alzheimer's disease compared to normal aging. Gerontology. 1997;43:26-43.

[12.] Scott RB. Extraneuronal manifestations of Alzheimer's disease. J Am Geriatr Soc. 1993;41:268-276.

[13.] Bosman GJCGM, Bartholomeus IGP, De Man AJM, Van Kalmthout PJC, De Grip WJ. Erythrocyte membrane characteristics indicate abnormal cellular aging in patients with Alzheimer's disease. Neurobiol Aging. 1991;12:13-18.

[14.] Passow H. Molecular aspects of band 3 protein-mediated anion transport across the red blood cell membrane. Rev Physiol Biochem Pharmacol. 1986;103: 61-223.

[15.] Jennings ML. Cellular anion transport. In: Seldin DW, Giebisch G, eds. The Kidney: Physiology and Pathophysiology. New York, NY: Raven Press Ltd; 1992: 113-145.

[16.] Alper SA. The band 3-related anion exchanger (AE) gene family. Annu Rev Physiol. 1991;53:549-564.

[17.] Tanner MJA. Molecular and cellular biology of the erythrocyte anion exchanger (AE1). Semin Hematol. 1993;30:34-57.

[18.] Cohen CM. New trends in understanding the regulation of red cell membrane skeleton organization. Adv Cell Biol. 1988;2:1-24.

[19.] Bosman GJCGM, Kay MMB. Alterations of band 3 transport protein by cellular aging and disease: erythrocyte band 3 and glucose transporter share a functional relationship. Biochem Cell Biol. 1990;68:1419-1427.

[20.] Greco FA, Solomon AK. Kinetics of chloride-bicarbonate exchange across the human red blood cell membrane. J Membr Biol. 1997;159:197-208.

[21.] McKhann G, Drachman D, Folstein M, Katzman R, Price D, Stadlan EM. Clinical diagnosis of Alzheimer's disease: report of the NINCDS-ADRDA Work Group under the auspices of Department of Health and Human Services Task Force on Alzheimer's Disease. Neurology. 1984;34:939-944.

[22.] Hachinski VC, Iliff LD, Zilhka E, et al. Cerebral blood flow in dementia. Arch Neurol 1975;32:632-637.

[23.] Folstein MF, Folstein SE, McHugh PR. Mini-mental state: a practical method for grading the mental state of patients for the clinician. J Psychiatr Res. 1975; 12:189-198.

[24.] Walter H, Widen KE. Differential electrophoretic behavior in aqueous polymer solutions of red blood cells from Alzheimer patients and from normal individuals. Biochim Biophys Acta. 1995;1234:184-190.

[25.] Fisher RA. The use of multiple measurements in taxonomic problems. Ann Eugenics. 1936;7:179-188.

[26.] England JM. Discriminant functions. Blood Cells. 1989;15:463-473.

[27.] Beck JR, Schultz EK. The use of relative operating characteristic (ROC) curves in test performance evaluation. Arch Pathol Lab Med. 1986;110:13-20.

[28.] Goodall HB, Reid AH, Findlay DI, Hind C, Kay J, Coghill G. Irregular distortion of the erythrocytes (acanthocytes, spur cells) in senile dementia. Dis Markers. 1994;12:23-41.

[29.] Hanin I, Reynolds CF, Kupfer DJ, et al. Elevated red blood cell/plasma choline ratio in dementia of the Alzheimer type: clinical and polysomnographic correlates. Psychiatry Res. 1984;13:167-173.

[30.] Greenwald BS, Edasery J, Mohs RC, Shah N, Trigos GG, Davis KL. Red blood cell choline, I: choline in Alzheimer's disease. Biol Psychiatry. 1985;20: 367-374.

[31.] Inestrosa NC, Alarcon R, Arriagada J, Donoso A, Alvarez J, Campos EO. Blood markers in Alzheimer's disease: subnormal acetylcholinesterase and butylcholinesterase in lymphocytes and erythrocytes. J Neurol Sci. 1994;122:1-5.

[32.] Bosman GJCGM, Renkawek K, Van Workum FPA, Bartholomeus IGP De Grip WJ. Involvement of neuronal anion exchange proteins in cell death in Alzheimer's disease. Gerontology. 1997;43:67-78.

[33.] Bosman GJCGM, Visser FE, De Man AJM, Bartholomeus IGP, De Grip WJ. Erythrocyte membrane changes of individuals with Down's syndrome in various stages of Alzheimer-type dementia. Neurobiol Aging. 1993;14:223-228.

[34.] Berg L. Does Alzheimer's disease represent an exaggeration of normal aging? Arch Neurol. 1985;24:737-739.

[35.] Mahalanobis PC. On the generalized distance in statistics. Proc Natl Inst Sci India. 1936;2:49-55.

[36.] The Ronald and Nancy Reagan Research Institute of the Alzheimer's Association and the National Institute on Aging Working Group. Consensus report of the Working Group on Molecular and Biochemical Markers of Alzheimer's disease. Neurobiol Aging. 1998;19:109-116.

[37.] Bruce LJ, Tanner MJ. Structure-function relationships of band 3 variants. Cell Mol Biol (Noisy-le-grand). 1996;42:953-973.

[38.] Hui DY, Noel JG, Harmony JA. Binding of plasma low density lipoproteins to erythrocytes. Biochim Biophys Acta. 1981;664:513-526.

[39.] Mecocci P, Beal MF, Cecchetti R, et al. Mitochondrial membrane fluidity and oxidative damage to mitochondrial DNA in aged and AD human brain. Mol Chem Neuropathol. 1997;31:53-64.

[40.] Muller WE, Eckert GP, Scheuer K, Cairns NJ, Maras A, Gattaz WF. Effects of [Beta]-amyloid peptides on the fluidity of membranes from frontal and parietal lobes of human brain: high potencies of A[BETA] 1-42 and A[BETA] 1-43. Amyloid. 1998; 5:10-15.

[41.] Markesbery WR, Leung PA, Butterfield DA. Spin label and biochemical studies of erythrocyte membranes in Alzheimer's disease. J Neurol Sci. 1980;45: 323-330.

[42.] Hajimohammadreza I, Brammer MJ, Eagger S, Burns A, Levy R. Platelet and erythrocyte membrane changes in Alzheimer's disease. Biochim Biophys Acta. 1990;1025:208-214.

[43.] Chauhan A, Chauhan VPS, Murakami N, Brockerhoff H, Wisniewski HM. Amyloid [Beta]-protein stimulates casein kinase I and casein kinase II activities. Brain Res. 1993;629:47-52.

[44.] Olivieri O, De Franceschi L, Bordin L, et al. Increased membrane protein phosphorylation and anion transport activity in chorea-acanthocytosis. Haematologica. 1997;82:648-653.

[45.] Wang CC, Tao M, Wei T, Low PS. Identification of the major casein kinase I phosphorylation sites on erythrocyte band 3. Blood. 1997;89:3019-3024.

[46.] Bursaux E, Hilly M, Bluze A, Poyart C. Organic phosphates modulate anion self-exchange across the human erythrocyte membrane. Biochim Biophys Acta. 1984;777:253-260.

[47.] Schlossmacher MG, Ostaszewski BL, Hecker LI, et al. Detection of distinct isoform patterns of the [Beta]-amyloid precursor protein in human platelets and lymphocytes. Neurobiol Aging. 1992;13:421-434.

[48.] Clarke R, Smith AD, Jobst KA, Refsum H, Sutton L, Ueland PM. Folate, vitamin B12, and serum total homocysteine levels in confirmed Alzheimer's disease. Arch Neurol. 1998;55:1449-1455.

[49.] Kennard ML, Feldman H, Yamada T, Jefferies WA. Serum levels of the iron binding protein p97 are elevated in Alzheimer's disease. Nature Med. 1996;2: 1230-1235.

[50.] Bush Al, Tanzi RE. Alzheimer disease-related abnormalities of amyloid [Beta] precursor protein isoforms in the platelet. Arch Neurol. 1998;55:1179-1180.

[51.] Taylor RP, Horgan C, Hooper M, Burge J. Dynamics of interaction between complement-fixing antibody/dsDNA immune complexes and erythrocytes: in vitro studies and potential general applications to clinical immune complex testing. J Clin Invest. 1985;75:102-111.

[52.] Rocca B, FitzGerald GA. Simply read: erythrocytes modulate platelet function: should we rethink the way we give aspirin? Circulation. 1997;95:11-13.

[53.] Davies TA, Long HJ, Tibbles HE, et al. Moderate and advanced Alzheimer's patients exhibit platelet activation differences. Neurobiol Aging. 1997;18:155-162.

[54.] Rosenberg RN, Baskin F, Fosmire JA, et al. Altered amyloid protein processing in platelets of patients with Alzheimer's disease. Arch Neurol. 1997;54: 139-144.

[55.] Di Luca M, Pastorino L, Bianchetti A, et al. Differential level of platelet amyloid [Beta]-precursor protein isoforms. Arch Neurol. 1998;55:1195-1200.

[56.] Peters LL, Birkenmeier CS, Bronson RT, et al. Purkinje cell degeneration associated with erythroid ankyrin deficiency in nb/nb mice. J Cell Biol. 1991; 114:1233-1244.

[57.] Lambert S, Bennett V. Postmitotic expression of ankyrin[.sub.R] and [Beta][.sub.R]-spectrin in discrete neuronal populations of the rat brain. J Neurosci. 1993;13:3725-3735.

[58.] Riederer BM, Zagon IS, Goodman SR. Brain spectrin(240/235) and brain spectrin(240/235E): two distinct spectrin subtypes with different locations within mammalian neural cells. J Cell Biol. 1986;102:2088-2097.

[59.] Siman R, Ahdoot M, Lynch G. Ontogeny, compartmentation, and turnover of spectrin isoforms in rat central neurons. J Neurosci. 1987;7:55-64.

[60.] Hu R, Watanabe M, Bennett V. Characterization of human brain cDNA encoding the general isoform of B-spectrin. J Biol Chem. 1992;267:18715-18722.

[61.] Walensky LD, Shi ZT, Blackshaw S, et al. Neurobehavioral deficits in mice lacking the erythrocyte membrane cytoskeletal protein 4.1. Curr Biol. 1998;8: 1269-1272.

[62.] Kay MMB, Goodman J. Brain and erythrocyte anion transporter protein, band 3, as a marker for Alzheimer's disease: structural changes detected by electron microscopy, phosphorylation, and antibodies. Gerontology. 1997;43:44-66.

Accepted for publication December 28, 1999.

From the Biophysical Laboratory, Harvard Medical School, Boston, Mass (Drs Greco and Solomon), and the Consolidated Department of Psychiatry, Harvard Medical School, and Geriatric Psychiatry Program, McLean Hospital, Belmont, Mass (Dr Satlin).

A preliminary report of these data appeared as a poster presentation at the Partners Symposium on Alzheimer's and Other Neurodegeneratire Diseases in January 1998.

Reprints: Frank A. Greco, MD, PhD, Edith Nourse Rogers Memorial Veterans Hospital, Research Service (151), Bedford, MA 01730 (e-mail: [e-mail]).