75

Address correspondence to John Savory, Ph.D., Departmentof Pathology, University of Virginia School of Medicine, P.O.Box 800214,Charlottesville, VA 22908, USA; tel 434 9245682; fax 434 924 2574; e-mail js2r@virginia.edu.

Aluminum-Induced Dendritic Pathology Revisited:Cytochemical and Electron Microscopic Studies ofRabbit Cortical Pyramidal Neurons

Michael S. Forbes,1 Othman Ghribi,1 Mary M. Herman,3 and John Savory 1,21 Department of Pathology, 2 Department of Biochemistry and Molecular Genetics,University of Virginia Health Sciences Center, Charlottesville, Virginia;3 National Institute of Mental Health, National Institutes of Health, Bethesda, Maryland

Abstract. Intracisternal administration of aluminum maltolate induces biochemical and histological changesin the rabbit brain. The primary histological response to this aluminum intoxication is the appearancewithin many neuronal somata and dendrites of intensely argyrophilic masses of fibrillar material.Ultrastructural examination of these bodies in both conventionally-prepared and silver-stained sectionsshows them to be composed of neurofilaments. For this reason, we have elected to call these argyrophilicmasses “neurofilamentous arrays (NFAs).” At their zenith, NFAs in cortical pyramidal neurons comprisethousands of filaments interconnected by periodic crossbridges. NFAs begin to be formed within bothsomata and dendrites as isolated groups of neurofilaments, which apparently go on to assemble en massewithin the cytoplasm. In symptomatic animals, many cortical neurons are rich in NFAs, yet lack classiccytological signs of degeneration, such as nuclear pyknosis. Though silver staining reveals extensive NFAsonly in aluminum-exposed brains, there is a strong degree of immunostaining for phosphorylatedneurofilamentous epitopes in both untreated and Al-injected animals. This suggests that protein subunitsthat are already present in the neurons under normal circumstances are recruited, in the presence of aluminum,to form NFAs through the directed assembly of masses of oriented filaments. (received 30 August 2001;accepted 12 October 2001)

Keywords: Alzheimer’s disease; aluminum neurotoxicity; neurofilamentous arrays, aluminum maltolate

Introduction

The neurotoxicity of the aluminum ion has longbeen established. A seminal finding in aluminumneurointoxication has been the presence in certainneurons of readily-visible cytoplasmic inclusions thatupon examination by various histological stainingregimens and electron microscopy prove to be massesof filaments. This observation has lent credence tothe use of this system for the study of neurofibrillary

pathology and related human neurodegenerativedisorders, including Alzheimer’s disease [1].Although there are distinct differences between thealuminum-induced inclusions and the Alzheimer“tangles,” there nevertheless remains between thetwo pathologies a similarity that involves consid-erable alteration to the cytoskeletal components ofneurons. Apoptosis is another feature that isassociated with severe aluminum neurotoxicity inrabbit brain [2], and this same cell-death pathwaymay play a role in the neuronal loss observed inAlzheimer’s disease [3,4]. It remains to be clarified,however, just how the various abnormalities seen inAlzheimer’s disease brain come eventually to resultin neuronal loss.

0091-7370/02/0100/0075 $3.00. © 2002 by the Association of Clinical Scientists, Inc.

Annals of Clinical & Laboratory Science, vol. 32, no. 1, 2002

76

Although a considerable body of earlier workhas been devoted to the presence, regionaldistribution, and general structural and stainingcharacteristics of aluminum-induced neuronalinclusions [5-9], we have initiated a closerexamination of their microscopic architecture andthe steps by which their development might proceed.We have chosen to term these bodies “neuro-filamentous arrays” (NFAs) to distinguish them fromsuperficially similar aggregates found in otherneuropathologies.

Materials and Methods

Animals. Young female white rabbits of the NewZealand strain were used for this study. The animalsranged from 8-12 mo in age and 2.4-3.6 kg inweight. Experimental animals were administered asingle dose of 50 mM aluminum maltolate in 100ml of sterile saline, injected directly into the cisternamagna with a sterile 25-gauge needle. Controlanimals were left untreated in order to provide anappreciation of the baseline “normal” ultrastructureof rabbit brain. Both in this and our previousbiochemical studies it was determined that one weekof survival is sufficient for histological alterations tobecome visible in spinal cord, brainstem, andcerebral cortex. By that time, aluminum-treatedanimals usually evince a distinct symptomatologythat consists of loss to some degree of motorfunction, particularly in the hind limbs, withaccompanying lethargy and loss of appetite.

Tissue processing. Brains from control and exper-imental animals were prepared for microscopicexamination by vascular perfusion with aldehydefixative. Following administration via the ear veinof a euthanizing dose of pentobarbital, each animalwas rapidly thoracotomized, its descending aortaclamped, and a 16-gauge needle inserted into theapex of the left ventricle. The right atrium was thencut and a gravity-feed perfusion begun, consistingin sequence of 300 ml of glucose/sucrose/salinesolution (0.4%/0.8%/0.8%, respectively), followedimmediately by 500 ml of a solution containing0.05% glutaraldehyde, 4% paraformaldehyde, 15%(v/v) saturated aqueous picric acid, and 0.1 M

sodium phosphate buffer, pH 7.4. Total perfusiontime was approximately 10 min, after which timethe brain was quickly exposed, removed from theskull, and immersed for 3 hr in fresh fixativesolution. The brain was then sliced into four separatecoronal segments (forebrain/midbrain/cerebellum +brainstem/spinal cord), which were reimmersed infixative at 4°C overnight. The segments were washedin several changes of 0.1 M sodium phosphate andstored in phosphate buffer with 0.01% sodium azideto avoid microbial contamination.

For sectioning, the individual segments wereencapsulated in 15% gelatin, and this stabilizingcoating was further hardened by immersion for 24-48 hr in aldehyde fixative. Selected segments werecut into 50-µm coronal sections with injector-typerazorblades mounted in a vibrating microtome (DSKMicroslicer DTK-3000, Ted Pella, Inc., Redding,CA), and returned to azide-phosphate solution untilfurther processing was undertaken. All sections wereprocessed in free-floating form, and not until theconclusion of each procedure were those sectionsintended for light microscopic documentationmounted on slides (see details, below).

Bielschowsky’s silver staining of neurofilament-ous arrays: Selected 50-µm sections were stained infree-floating form by a protocol used in thislaboratory to reveal argyrophilic components inneurons. In summary, sections were immersed indarkness in 20% aqueous silver nitrate at 37°C for30 min, followed by 15 min incubation in“ammoniacal silver” (prepared by adding concen-trated ammonium hydroxide to the silver nitrate,while stirring, until a clear solution appears). Thesilver reaction was developed by adding ~10 drops/50 ml solution of a “developer” consisting of 2%formalin and 0.4% citric acid in distilled wateracidified with concentrated nitric acid (2 drops/250ml).

Whereas in our experience 14-µm-thick,formaldehyde-fixed sections mounted on slidesrequire only about 15-20 min to show a suitabledegree of staining in this solution, the considerablythicker sections used in this study required as muchas 2 hr to display thorough argyrophilia of neuronalcytoskeletal elements (as monitored, during thedevelopment step, by light-microscopic examination

Annals of Clinical & Laboratory Science

77

of the wet sections). The reaction was stopped bytransfer of sections into 1% aqueous ammoniumhydroxide, with a subsequent “fix” in 5% sodiumthiosulfate. Some sections were washed in distilledwater, affixed to gelatin-coated slides, allowed to dry,and then cleared in xylenes and coverslipped for lightmicroscopic examination, while others were directedto a plastic-embedment regimen detailed below.

Immunocytochemical staining for phosphorylatedneurofilaments. Pre-embedding immunocyto-chemistry on brain sections was carried out aspreviously described [10]. To summarize, sectionswere infiltrated 30 min with a cryoprotectantsolution (2% DMSO and 20% glycerol in 0.1 Msodium cacodylate) and collected on segments ofglass microscope slides. These were placed intoaluminum weighing dishes, which were set on aliquid-nitrogen surface for 2 cycles of freezing andthawing; this procedure renders sections morepermeable to antibody penetration.

The sections were washed thoroughly inphosphate buffer, blocked with 10% normal horseserum and 1% BSA (Sigma #7030) in 0.1 M sodiumphosphate, pH 7.4, and then incubated in vials ofprimary antibody solution (see primary antibodyconcentrations below; carrier solution consisted of1% BSA in 0.1 M sodium phosphate buffer, pH7.4). Incubation continued overnight at 4°C withconstant agitation on a Nutator (model 1105, Clay-Adams, Parsippany, NJ).

The next morning, sections were washed inbuffer and then exposed to secondary antibody incarrier (1:200 biotinylated horse anti-mouse IgG,Vector BA-2000) at RT for 2 hr, followed by rinsesand incubation in avidin-solution (Vector VectastainABC Elite Kit) for 2 hr at room temperature.Sections were then rinsed and the reaction developedfor 30 min by the glucose oxidase-diaminobenzidinemethod of Itoh et al [11].

As with the silver-stained sections that wereintended for light microscopic examination alone,immunostained sections were affixed to gelatin-coated slides, allowed to dry, cleared in a series ofxylene solutions, and coverslipped.

A monoclonal antibody that reacts with phos-phorylated neurofilament epitopes (SMI-31, Stern-

berger Monoclonals, Inc., Lutherville, MD) wastested on sections at dilutions of 1:1000 and 1:5000.

Electron microscopy. For conventional TEMexamination, the 50-µm sections were washed in 0.1M Na cacodylate solution and postfixed gradually(to minimize section curling) in ascending concen-trations of osmium tetroxide (final concentration1%) in cacodylate buffer (pH 7.4) for a totalpostfixation period of 30 min. En bloc immersionin saturated aqueous uranyl acetate was carried outfor 30 min. Some of the silver-stained sections wereprepared without osmium postfixation and uraniumblock-staining.

All sections were dehydrated in an ascendingseries of ethanols, passed through propylene oxide,and infiltrated under vacuum in Poly/Bed 812 resin(Polysciences, Warrington, PA). The free-floatingsections were transferred to microscope slides coatedwith liquid-release agent (Electron MicroscopySciences, Fort Washington, PA) and topped withsimilarly-coated coverslips, which were weightedwith coins to flatten the sections. The embedmentswere then cured at 60°C for two days. Sections wereexamined on the slides with a light microscope andareas of interest selected. The coverslip portions overthese regions were removed with a scribing tool anddissecting pin, and a trapezoidal portion of the brainsection was cut out with a razorblade and glued withcyanoacrylate to an epoxy capsule.

Semi-thin (0.25-1.0 µm) sections were preparedwith glass knifes on a Sorvall MT-2B ultramicrotomeand stained with 1% toluidine blue in 1% sodiumborate for orientation and detailed light microscopicdocumentation; for transmission electronmicroscopy (TEM), 7-10 nm “ultrathin” sectionswere cut with a DuPont diamond knife.

Sections were collected on copper-mesh gridsand stained with saturated uranyl acetate in 50%acetone (1-2 min) and 0.5% alkaline lead citrate(1-2 min). All sections were examined in a ZeissEM-10CA transmission electron microscope at anaccelerating voltage of 60 keV. This instrument wascalibrated for crucial magnifications against a replicaof an optical grating, and measurements of cellstructures were made directly from high-magnification photographs with a vernier caliper.

Neurofilamentous arrays in dendrites of Al-treated rabbits

78

Figs. 1-6. Silver-stained (Bielschowsky’s method) coronal 50-µm sections through frontal cortex showing pyramidalneuron cytoskeletal components in untreated young female rabbit (Figs. 1,3) and aluminum-treated female (50 mMaluminum maltolate, 7 days exposure) (Figs. 2,4-6).Fig. 1. Silver staining confers a delicate, threadlike pattern running vertically throughout much of the cortex Althoughthe neuronal nuclei stain lightly with silver (cf. Fig. 3), outlines of the perinuclear cytoplasm are not evident in controltissue. 40x.Fig. 2. An equivalent section from an animal exposed to aluminum for 7 days. A layer of cortical pyramidal neuronsstands out in silhouette because of the argyrophilic staining of their cytoplasm. 40x.Fig. 3. Detail of control cortex. The lightly opacified dots scattered throughout this field are cell nuclei (Nu). The

Annals of Clinical & Laboratory Science

79

Results

Light microscopic observations: The cytoskeletalelements of neurons are clearly delineated with theBielschowsky’s method of silver staining; their three-dimensional interrelationships are particularly welldemonstrated in 50-µm-thick sections stained withthis procedure (Figs. 1-6). In the frontal lobe cortexof untreated animals, the cytoskeleton is revealed asdark threadlike profiles located primarily in axons(Figs. 1,3), whereas equivalent sections of Al-treatedbrain offer a striking contrast, with numerouspyramidal cell bodies standing out in silhouettebecause of prominent silver-stained profiles withinthem (Figs. 2,4-6); cells affected in this manner arefound in varying numbers throughout the frontallobe, but are particularly numerous in the morecaudal frontal lobe, over the medial convexity andin the cingulate gyri, tending to be most concen-trated on either side of the dorsal midline.

In cortical cells from aluminum-exposed brains,a pattern of staining is evident that consists ofelongated apical dendritic aggregates, accompaniedby short, thick basal dendritic arrays (Figs. 2,4-6).The apical dendritic masses may display“corkscrewed” profiles (Fig. 5) that in many casescan be resolved into several separate, parallel helicalskeins within the same process.

In the “whole-mount” sort of display affordedby the thick-section, Bielschowsky’s-stained prep-

arations, some cells (Fig. 6) demonstrate continuitybetween the apical and basal argyrophilic masses.This can be confirmed in toluidine-blue-stained,“semi-thin” sections of plastic-embedded material(Fig. 7), where because of their lack of osmiophiliathe neuronal inclusions appear unstained against thedarker background of contrasted nucleus and cyto-plasm. Although nuclei are difficult to detect inBielschowsky’s-stained cortex of aluminum-treatedrabbits, being largely obscured by the intensely silver-stained masses superimposed on them, they arereadily identified in semi-thin and ultrathin sections,and are normal in appearance (Figs. 7-9).

Electron microscopic studies. In the electron micro-scope, the cytoplasmic inclusions appear in thinsection as lucent profiles of variable size and shape(Fig. 8). TEM examination of Bielschowsky’s silver-stained sections shows that these bodies areselectively and specifically marked by deposition ofelectron-opaque silver deposits (Fig. 9). Eachinclusion is composed primarily of numerousfilaments, which in some cases are arranged in largemasses that can extend for a continuous length ofnearly half a millimeter (as measured in lightmicrographs) and are as much as 15 µm in breadth,with only an occasional mitochondrion and a fewvesicles or tubules captured within the filamentousconfines (Fig. 10).

pattern of silver staining here is limited to extensive thin profiles of cytoskeletal material, the majority associated withaxons. 110x.Figs. 4-7. Frontal cerebral cortex from aluminum-treated animal (treatment as described in Fig. 1).Fig. 4. Aluminum-treated animal. Although some nuclei can be discerned, the picture is dominated by densely-staining cortical cell cytoplasm, made evident by its prominent apical (AD) and basal (BD) dendrites, the latterfanning out from the base of the neuronal cell bodies. 110x.Fig. 5. Aluminum-treated. The elongate silver-positive inclusions within apical dendrites in some instances displaya twisted or “corkscrew” morphology; in places along their lengths, parallel separate strands of stained material can bediscerned. Equivalent masses in the basal dendrites are heavily stained, but are less extensive. 280x.Fig. 6. In this pyramidal cortical neuron, the silver-stained cytoskeletal material is continuous from the apical dendrite,extending as several thin strands through the cell body that pass around the nuclear zone (Nu) to merge with thematerial in each basal dendrite. 565x.Fig. 7. “Semi-thin” (~0.25 µm) section from plastic-embedded cortex, stained with toluidine blue and viewed in anorientation similar to that seen in Fig. 6. Here the nucleus (Nu) is viewed in section, partly surrounded by a contiguousmass of proliferated cytoskeletal material, which in this preparation appears lucent against the basophilia of the othercytoplasmic contents, and extends from the apical dendrite into the cell soma. 1140x.

Neurofilamentous arrays in dendrites of Al-treated rabbits

80

Figs. 8-16. Transmission electron micrographs of cerebral cortex from aluminum-treated rabbit brain, documentingthe ultrastructure of neurofilamentous arrays (NFAs) that characterize the aluminum-induced changes.Fig. 8. Survey of pyramidal neuron (approximately same orientation as previous illustrations); within the apicaldendrite and the perinuclear soma (Nu, nucleus) appear lucent zones (*) that represent sections through masses offilaments. Toward the bottom of the field, similar masses are directed away from the lower portion of the cell into thecytoplasmic extensions that form basal dendrites (BD). No substantial alterations in ultrastructure appear in eitherthe nuclear profile nor do any of the other cytoplasmic contents appear altered. 1800x.

Annals of Clinical & Laboratory Science

81

Inspection at high magnification shows thefilaments to be closely packed and uniform inappearance (Figs. 11,12). On the basis of their 14-nm diameter (14.1±1.7 nm, n =15), similar to thatof neurofilaments that form sparse populations in

dendrites of untreated animals (14.9± 0.9 nm, n =15) and to axonal filaments (13.2± 0.9 nm, n = 15),they can be classified as belonging to the cytoskeletalfibril subcategory commonly known as “neuro-filaments.” Because of this, we designate these

Fig. 9. Cortical neuron in orientation similar to that in Fig. 8, but from a Bielschowsky’s silver-stained section forcomparison. Dominating the field is an opacified neurofilamentous array (NFA) that extends from the cell body upinto the apical dendrite. The fibrillar component of the neuron’s nucleolus (Nuc) is also stained. 3000x.Fig. 10. In this somatic NFA, the majority of filaments are cut in transverse section. The NFA is almost entirelycomposed of 14-nm-diameter neurofilaments (detail shown in inset) with only an occasional profile of a mitochondrion(Mi) or ER tubule. 10,400x; inset 37,000x.Fig. 11. Longitudinal section through NFA in apical dendrite. The closely-packed neurofilaments observe a strictparallel arrangement. Periodic crossbridges extend between adjacent filaments, merging with an amorphous substancethat coats the filament shafts. 80,000x.Fig. 12. Transversely-cut NFA, showing distribution of neurofilaments, many of which are connected laterally bycrossbridges that hold the adjacent filaments in register with a center-to-center spacing of ~33 nm. Amorphousmaterial adheres to the profiles of the individual filaments as well. 134,000x.Fig. 13. One of two small perinuclear concentrations of neurofilaments in a nerve cell found adjacent to otherneurons that contained large NFAs. The component neurofilaments are flanked by Golgi saccules (GA), and, as inthe larger NFAs, are oriented mostly parallel to one another. 24,500x.Fig. 14. A skein of parallel neurofilaments is suspended in the long axis of an apical dendrite. This sort of assembly,along with somatic accumulations like those seen in Fig. 13, appears to represent the beginnings of NFA formation.Although this collection of filaments is less extensive than the NFAs in fully-involved neurons, its components possesssimilar spacing, coatings and crossbridges (see inset). 24,500x; inset 87,500x.

Neurofilamentous arrays in dendrites of Al-treated rabbits

82

particular argyrophilic masses as “neurofilamentousarrays” (NFAs).

While neurofilaments and glial filaments alikeare considered to belong to the general categoryknown as “10-nm” or “intermediate” filaments, it isclear that the neurofilaments, whether found inunaffected dendrites, axons, or NFAs, are typicallylarger in overall diameter, consistently displaying anamorphous coating that is lacking from glialfilaments. In contrast, the glial filaments—asmeasured in the same sections—average about 10nm in diameter (10.2±1.0 nm, n = 15). (Allmeasurements of diameters of different filamentcategories were made on thin sections fromconventionally prepared material convention: ie,aldehyde-fixed and osmium-postfixed, with uranylacetate block-staining and on-section staining withboth uranium and lead.)

Closely-packed neurofilaments that composeNFAs are joined by numerous cross-bridgingstructures (Figs. 11,12) that hook the filamentstogether in a lattice-like array, in which adjacentfilaments commonly observe parallel orientation anda center-to-center spacing of ~33 nm (33.4±2.6 nm,n = 17) (Fig. 12). The degree of NFA developmentis not the same in all cortical neurons, even at thesame coronal level; fully-formed populations ofNFAs may be evident throughout the dendrites andcell body of a neuron, while only small filamentgroups appear in nearby cells (Figs. 13,14). Regard-less of the extent of NFAs, there is no overt ultra-structural sign of degeneration in any of the cellsthat contain them. That is, NFA-containing cells,except for their masses of filaments, are qualitativelysimilar to their unaffected neighbors, displaying nosign of nuclear pyknosis, Golgi swelling, or obviouspathology in the ER and mitochondria.

Figs. 15-18. Immunostaining for phosphorylated neurofilament epitope SMI-31.Fig. 15. Untreated cortex. Staining is evident throughout the cell bodies and dendrites of pyramidal neurons. 270x.Figs. 16,17. Aluminum-exposed cortex. Although casual inspection of Fig. 16 shows these neurons to be similar insize and appearance to those in Fig. 15, the regions of immunostaining are notable in the degree of opacification oftheir contents. This is particularly visible in the apical (AD) and basal dendrites (BD), which at higher magnification(Fig. 17) appear darker and thicker than those in untreated cells. Fig. 16, 270x; Fig. 17, 700x.

Annals of Clinical & Laboratory Science

83

Immunocytochemical observations. Despite theprofound histological differences between normaland aluminum-treated cortical neurons, theimmunochemical pattern for the phosphorylatedneurofilament epitope is similar, with abundantstaining throughout the dendrites and somata inboth untreated (Fig. 15) and aluminum-treatedcortical pyramidal neurons (Fig. 16). The overallstaining appears more intense in aluminum-treatedanimals, and close examination shows this to resultfrom the density of immunoreactive material withinthe apical and basal dendrites (Fig. 17).

Discussion

Fibrillar neuronal inclusions, often called “neuro-fibrillary tangles,” are a hallmark of neurodegen-eration in a variety of pathological situations,including Alzheimer’s disease and aluminum toxicity.In the case of aluminum-treated rabbits, develop-ment of these fibrillar inclusions (neurofilamentousarrays or NFAs) is well-documented, predictable intiming, and age-responsive. However, the presenceof NFAs seems more indicative of a synthetic eventthan of an intrinsically degenerative process. Whilewe do not believe the presence of NFAs necessarilysignifies impending cell death, their developmentin certain brain cells (eg, cortical pyramidal neurons)is nevertheless a bellwether, signaling the occurrenceof a neurological insult (in this case exposure toaluminum maltolate).

The resulting biochemical changes in proteinssuch as cytochrome c, Bax, and Bcl-2 are virtuallyidentical in both cortex and neighboring hippo-campus (12), despite the paucity of NFAs inhippocampal neurons. The time-course ofdevelopment of cortical NFAs is consistent, evenwhen partial neuroprotection is bestowed by glialcell-derived nerve growth factor (GDNF), a proteinwhich when injected simultaneously with aluminumis associated with the animals’ ability to survive wellbeyond a week of aluminum exposure [13]. Afterthis time, many of the NFAs eventually wane(Forbes, unpublished observations) suggesting thatNFA formation may be a largely transitory event,embodying a defensive, “filtering” response bycertain cells to the presence of the aluminum ion.

NFAs in aluminum-treated rabbit brain arecomposed of normal-appearing neurofilaments.The same three polypeptide species (68, 150-160,and 200 kDa) are found in “normal” and NFA-component neurofilaments [14,15], and all suchneurofilaments are immunologically distinct fromglial cell filaments [16], which are composedprimarily of glial fibrillary acidic protein (GFAP).

Commercially available monoclonal antibodiesfor the phosphorylated and non-phosphorylatedepitopes of neurofilaments have been utilized inseveral studies of NFAs in the aluminum-rabbitmodel [5,17-19]. In these studies, SMI-31 (whichdetects phosphorylated NF antigens) intenselystained NFAs of aluminum-exposed brain, while inequivalent neurons of control animals, SMI-31immunoreactivity was restricted to the axons, withno reactivity at all appearing in either dendrites orperikarya. Our present experience with this sameantibody, however, agrees with the findings ofKatsetos et al [20], since it demonstrates aconsiderable presence of phosphorylated NF epitopethroughout cortical pyramidal neurons in controland aluminum-treated rabbit brain alike.

Interestingly, it has been found that the topicalapplication of SMI-31 antibody to acetone-fixedcultured spinal neurons gave results similar to theother studies–ie, axons were strongly labeled, butcell bodies were not [21]. However, actual injectionof antibody into the living cells or immunoreactionof Triton-extracted cells prior to fixation bothresulted in positive staining of perikaryal anddendritic filament complements [21]. In that reportit was proposed that the question of SMI-31immunoreactivity is not dictated by the actual degreeof neurofilament phosphorylation, but rather by therelative accessibility of the epitopes [21]. Thesefindings are consistent with previous findings fromour laboratory (22) that detected no significantchanges between control and aluminum-exposedrabbit brain, either in the degree of phosphorylationor gene expression of any of the neurofilamentisoforms. The immunohistochemical method usedin this and other laboratories for “permeabilization”of sections, then, assumes special significance. Asnoted in our Methods section, we utilize a processof repeated freeze/thaw cycles, in the presence of

Neurofilamentous arrays in dendrites of Al-treated rabbits

84

cryoprotectant solution. It is thought that thisprocedure produces submicroscopic fissures in cellmembranes, which enhance penetration of thesubsequent reagent solutions, in particular theantibodies. This regimen, then, may optimize theimmunological reactivity of phosphorylatedneurofilament epitopes, whether they are incorp-orated into actual filaments or not.

By and large, neuronal dendrites, despite theirnewly-appreciated functionalities [23-25], are notunder ordinary circumstances striking in terms oftheir architecture. However, upon exposure of theCNS to aluminum there rapidly appear, within bothapical and basal dendrites of cortical pyramidalneurons, populations of oriented neurofilamentstightly bound together in massive rods or bundlesthat may nearly fill the dendritic processes, andoccupy substantial somatic cytoplasm besides. Thegroups of filaments appear to begin forming in situfrom nucleating points in both soma and dendritesto create tightly-organized skeins of parallelneurofilaments.

These aluminum-induced cytoskeletal changesmay result largely from utilization of the protein-aceous material—already present in non-filamentousform—which is induced to adopt fibrillarconfiguration and create large three-dimensionalaggregates. Overall, this indicates that phosphor-ylated neurofilament epitopes are already in place,and abundantly so, in the normal brain. Thefundamental difference between control neurons andaluminum-exposed, NFA-containing cells is thestructural form taken by the majority of neuro-filament proteins; ie, they may exist largely in a non-fibrillar form in control animals, while inaluminum-exposed rabbits they are displayed mainlyas large, extensive masses of long filaments. Thereappears, then, to occur a rapid and extensive recon-figuration of existing cytoskeletal subunits into fully-formed cytoskeletal fibrils, interconnected bynumerous crossbridge structures. In vitro treatmentof neuroblastoma cells [26] and fetal rabbit midbrainneurons [27] with aluminum salts has been shownto produce Bielschowsky-silver positive whorlsconsisting of neurofilaments; these whorledaccumulations were not seen in the somata ofuntreated cells, however. This fits in well with our

observations on treated brain, where it also appearsthat aluminum elicits the formation of greatnumbers of “assembled” (polymerized) neuro-filaments. Our finding in untreated neurons of anabundance of phosphorylated epitopes thus suggeststhat a neuro-filament’s architecture, as much as ormore than its degree of phosphorylation, is the basisfor Bielschow-sky’s silver staining in rabbit brain.

Although, in general, neurofibrillary inclusionshave been linked to one or another disease process,we here make the case that NFAs in rabbit brainare transitory and/or reversible in nature, given theright circumstances; under those circumstances,furthermore, their rise and fall can occur quickly.Given the far longer time-frame characteristic ofAlzheimer’s disease, however, it has recently beenproposed that in addition to the lack of a universaloccurrence of neurofibrillary tangles (NFTs), thecells which do develop them may continue to survivein spite of their presence for long periods of time(10-20 yr on average) [25]. Even though there existsubstantial differences between NFTs and NFAs, notthe least of which is the relative lifespan of their host,it appears that the conclusions of Morsch et al [28]regarding Alzheimer’s tangles are equally applicableto aluminum-induced neurofilamentous arrays.

What, then, is the reason for NFA occurrence?Even though there was evidence of hind-limbmalfunction in rabbits 7-8 days after intraventricularor intracisternal aluminum injection, Simpson et al.(29) concluded that no “gross effect on neuronalmetabolism” had been wrought either by aluminumor the presence of NFAs. In the absence of overtmobilization of protein synthetic machinery, suchas might be indicated at the fine-structural level byenlarged Golgi saccules, increased ribosomal rosettesand proliferated endoplasmic reticulum, andconsidering that a great deal of neurofilamentprotein is already present (even in neurons notexposed to aluminum), we have suggested that thefilaments which make up NFAs are assembled inlarge part from proteins already on hand.

It has been proposed that, in Alzheimer’s disease,paired helical filaments are able to form andaccumulate because their composition is so similarto “normal” cytoskeletal proteins as to escape thedegradative and exocytotic mechanisms that

Annals of Clinical & Laboratory Science

85

ordinarily would be triggered by the presence offoreign materials [30]. The situation for aluminum-induced NFAs might be even more subtle.

Aluminum’s effects are multifocal; some studieshave indicated that nuclear chromatin is especiallysensitive to this ion [31], and histochemical stainsspecific for aluminum furthermore localize onchromatin [9,32], implying that aluminum andchromatin have in fact a certain, perhaps deadly,affinity for one another. While aluminum may binddirectly to genetic material, it appears also to be atwork in the cytoplasm. In an early study with theSolochrome-Azurine method for aluminumlocalization, Klatzo et al [6] found staining of bothnucleoli and NFAs. It appears that the presence ofaluminum ion promotes the aggregation ofphosphorylated cytoskeletal proteins [26,33].

It seems likely that there exists some balancebetween buildup and breakdown of variousstructures, including neurofilaments. Goldstein etal [34] found that phosphorylated neurofilamentsare resistant to proteinase action, whiledephosphorylated ones are more susceptible. In theircomparison of aluminum intoxication withAlzheimer’s disease, Yokel et al [35] proposed thatsufficient inhibition of proteinase activity wouldinterrupt the normal cycle of filament breakdownin the neurons, thus leading to the production oflarge filamentous aggregations. One may speculatetherefore that aluminum may subvert the productionof proteinase at the nuclear transcription/translationlevel, while in the cytoplasm it simultaneouslypromotes both the assembly and architecturalstability of neurofilament arrays.

We conclude, therefore, that the formation ofNFAs in aluminum-treated animals can help ourunderstanding of the role of NFTs in the Alzheimer’sdisease brain, and that both—although classicalneuropathological hallmarks of their respectivedisorders— might not cause neuronal death. Rather,these fibrillar accumulations might represent aprotective response by neurons, buffering the effectof neurotoxins in an attempt to minimize apoptotic(and perhaps necrotic) cell death.

Acknowledgements

Supported by a grant (DAMD 1799-1-9552) from theUS Department of the Army to John Savory. The trans-mission electron microscope was provided and main-tained by the Central Electron Microscope Facility ofthe University of Virginia School of Medicine. Theauthors thank Carlo Bruni, Professor Emeritus at theUniversity of Virginia, for helpful discussions.

References

1. Huang Y, Herman MM, Liu J, Katsetos CD, WillsMR, Savory J. Neurofibrillary lesions in experi-mental aluminum-induced encephalopathy andAlzheimer’s disease share immunoreactivity foramyloid precursor protein, Ab, α1-antichymotrypsinand ubiquitin-protein conjugates. Brain Res 1997;771:213-220.

2. Savory J, Rao JKS, Huang Y, Letada P, Herman MM.Age-related hippocampal changes in Bcl-2:Bax ratio,oxidative stress, redox-active iron and apoptosisassociated with aluminum-induced neurodegen-eration: increased susceptibility with aging.NeuroToxicol 1999;20:805-818.

3. Behl C. Apoptosis and Alzheimer’s disease. J NeuralTransm 2000;107:1325-1344.

4. Honig LS, Rosenberg RN. Apoptosis and neurologicdisease. Am J Med 2000;108:317-330.

5. Kowall NW, Pendlebury WW, Kessler JB, Perl DP,Beal MF. Aluminum-induced neurofibrillary degen-eration affects a subset of neurons in rabbit cerebralcortex, basal forebrain, and upper brainstem.Neurosci 1989;29:329-337.

6. Klatzo I, Wisniewski HM, Streicher E. Experimentalproduction of neurofibrillary degeneration. 1. Lightmicroscopic observations. J Neuropathol Exp Neurol1965;24:187-99.

7. Wisniewski H, Karczewski W, Wisniewska K.Neurofibrillary degeneration of nerve cells afterintracerebral injection of aluminium cream. ActaNeuropathol 1966;6:211-219.

8. Yates CM, Gordon A, Wilson H. Neurofibrillarydegeneration induced in the rabbit by aluminiumchloride: aluminium neurofibrillary tangles. Neuro-pathol Appl Neurobiol 1976;2:131-144.

9. Boni UD, Otvos A, Scott JW, Crapper DR. Neuro-fibrillary degeneration induced by systemic alum-inum. Acta Neuropathol 1976;35:285-294.

10. Alheid GF, Beltramino CA, de Olmos JS, ForbesMS, Swanson DJ, Heimer L. The neuronal organ-ization of the supracapsular part of the stria term-inalis in the rat: the dorsal component of theextended amygdala. Neurosci 1998;84:967-996.

Neurofilamentous arrays in dendrites of Al-treated rabbits

86

11. Itoh K, Konishi A, Nomura S, Mizuno N, NakamuraY, Sugimoto T. Application of coupled oxidationreaction to electron microscopic demonstration ofhorseradish peroxidase: cobalt-glucose oxidasemethod. Brain Res 1979;175:341-346.

12. Ghribi O, DeWitt DA, Forbes MS, Herman MM,Savory J. Co-involvement of mitochondria andendoplasmic reticulum in regulation of apoptosis:changes in cytochrome c, Bcl-2 and Bax in thehippocampus of aluminum-treated rabbits. BrainRes 2001;903:66-73.

13. Ghribi O, Forbes MS, ., DeWitt DA, Herman MM,Savory J. GDNF protects against aluminum-inducedapoptosis in rabbits by upregulating Bcl-2 and Bcl-XL and inhibiting mitochondrial Bax translocation.Neurobiol Dis 2001;8:764-773.

14. Selkoe DJ, Liem RK, Yen SH, Shelanski ML. Bio-chemical and immunological characterization ofneurofilaments in experimental neurofibrillarydegeneration induced by aluminum. Brain Res1979;163:235-252.

15. Ghetti B, Gambetti P. Comparative immunocyto-chemical characterization of neurofibrillary tanglesin experimental maytansine and aluminumencephalopathies. Brain Res 1983;276:388-393.

16. Dahl D, Bignami A. Immunochemical cross-reactivity of normal neurofibrils and aluminum-induced neurofibrillary tangles. Immunofluores-cence study with antineurofilament serum.Experimental Neurology 1978;58:74-80.

17. Pendlebury WW, Perl DP, Schwentker A, PingreeTM, Solomon PR. Aluminum-induced neuro-fibrillary degeneration disrupts acquisition of therabbit’s classically conditioned nictitating membraneresponse. Behav Neurosci 1988;102:615-620.

18. Strong MJ, Wolff AV, Wakayama I, Garruto RM.Aluminum-induced chronic myelopathy in rabbits.NeuroToxicol 1991;12:9-21.

19. Savory J, Huang Y, Herman MM, Wills MR.Quantitative image analysis of temporal changes intau and neurofilament proteins during the courseof acute experimental neurofibrillary degeneration;non-phosphorylated epitopes precede phosphoryl-ation. Brain Res 1996;707:272-281.

20. Katsetos CD, Savory J, Herman MM, CarpenterRM, Frankfurter A, Hewitt CD, Wills MR.Neuronal cytoskeletal lesions induced in the CNSby intraventricular and intravenous aluminiummaltol in rabbits. Neuropathol Appl Neurobiol1990;16:511-528.

21. Durham HD. Demonstration of hyperphosphor-ylated neurofilaments in neuronal perikarya in vivo

by microinjection of antibodies into cultured spinalneurons. J Neuropathol Exp Neurol 1990;49:582-590.

22. Savory J, Herman MM, Hundley J, Seward RL,Griggs CM, Katsetos CD, Wills MR. Quantitativestudies on aluminum deposition and its effects onneurofilament protein expression and phosphor-ylation, following the intraventricular administrationof aluminum maltolate to adult rabbits. Neuro-Toxicol 1993;14:9-13.

23. Matus A. Actin-based plasticity in dendritic spines.Science 2000;290:754-758.

24. Hausser M, Spruston N, Stuart GJ. Diversity anddynamics of dendritic signaling. Science 2000;290:739-744.

25. Segev I, London M. Untangling dendrites withquantitative models. Science 2000;290:744-750.

26. Shea TB, Clarke JF, Wheelock TR, Paskevich PA,Nixon RA. Aluminum salts induce the accumulationof neurofilaments in perikarya of NB2a/dl neuro-blastoma. Brain Res 1989;492:53-64.

27. Hewitt CD, Herman MM, Lopes MBS, Savory J,Wills MR. Aluminum maltol-induced neurocyto-skeletal changes in fetal rabbit midbrain in matrixculture. Neuropath Appl Neurobiol 1991;17:47-60.

28. Morsch R, Simon W, Coleman PD. Neurons maylive for decades with neurofibrillary tangles. JNeuropathol Exp Neurol 1999;58:188-197.

29. Simpson J, Yates CM, Whyler DK, Wilson H,Dewar AJ, Gordon A. Biochemical studies on rabbitswith aluminium induced neurofilament accum-ulations. Neurochem Res 1985;10:229-238.

30. De Boni U, McLachlan DR. Senile dementia andAlzheimer’s disease: a current view. Life Sci 1980;27:1-14.

31. Walker PR, Leblanc J, Sikorska M. Effects ofaluminum and other cations on the structure of brainand liver chromatin. Biochemistry 1989;28:3911-3915.

32. De Boni U, Scott JW, Crapper DR. Intracellularaluminum binding; a histochemical study.Histochemistry 1974;40:31-37.

33. Diaz-Nido J, Avila J. Aluminum induces the in vitroaggregation of bovine brain cytoskeletal proteins.Neurosci Lett 1990;110:221-226.

34. Goldstein ME, Sternberger NH, Sternberger LA.Phosphorylation protects neurofilaments againstproteolysis. J Neuroimmunol 1987;14:149-160.

35. Yokel RA, Provan SD, Meyer JJ, Campbell SR.Aluminum intoxication and the victim ofAlzheimer’s disease: similarities and differences.NeuroToxicol 1988;9:429-442.

Annals of Clinical & Laboratory Science