|Home | About | Journals | Submit | Contact Us | Français|
Although multiple studies have suggested a role for alterations of zinc (Zn) and zinc transport (ZnT) proteins in the pathogenesis of Alzheimer’s disease (AD), the exact role of this essential trace element in the progression of AD remains unclear. The following review discusses the normal role of Zn and ZnT proteins in brain and the potential effects of their alteration in the pathogenesis of AD particularly in the processing of the amyloid precursor protein and amyloid beta peptide generation and aggregation.
Alzheimer’s disease (AD) is the fourth leading cause of death in the United States and in 2000 affected 4.5 million Americans . Estimates indicate that ~3% of Americans between ages 65 and 74, 19% ages 75 – 84, and 47% over age 85 are victims of the disease  and that ~ 60% of nursing home patients over age 65 suffer from AD. Clinically AD is characterized by a progressive decline in multiple cognitive functions and is thought to begin with amnestic mild cognitive impairment (MCI), widely considered to be a transition between normal aging and dementia. Current data suggest that conversion from MCI to dementia occurs at a rate of 10 to 15% per year  with ~80% conversion by the sixth year of followup; although ~5% of MCI subjects remain stable or convert back to normal [13,52]. Progression from MCI leads to early AD (EAD) which is clinically characterized by a) a decline in cognitive function from a previous higher level, b). decline in one or more areas of cognition in addition to memory, c) a clinical dementia rating scale score of 0.5 to 1, d) impaired ADLs, and e) a clinical evaluation that excludes other causes of dementia. The disease eventually progresses to late stage AD (LAD) characterized by severe dementia with disorientation, profound memory impairment, global cognitive deficits and immobility. Without preventive strategies, there may be 14 million Americans with AD by the year 2040 .
Pathologically AD is characterized by an abundance of neurofibrillary tangles (NFT), senile plaques (SP), neuropil thread formation, and Aβ deposition; neuron and synapse loss; and proliferation of reactive astrocytes, particularly in the hippocampus, amygdala, entorhinal cortex, and neocortex. NFT are composed of intracellular deposits of paired helical filaments composed of hyperphosphorylated tau. Senile plaques are present in two forms: a) diffuse plaques (DP) composed of amorphous extracellular deposits of Aβ that lack neurites, and b) neuritic plaques (NP) composed of extracellular deposits of insoluble Aβ surrounded by dystrophic neurites, reactive astrocytes, and activated microglia. Recent studies suggest that, in addition to insoluble Aβ present in SP, soluble Aβ oligomers are present in the AD brain and may represent the main toxic form of Aβ, thus implicating them in the disease process [79,109,192].
Risk factors identified for AD include age , genetic factors including presenilin 1 (PS1) and 2 (PS2) and APP mutations  , single nucleotide polymorphisms in the gene that codes for ubiquilin-1 , a genetic locus on chromosome 10 that includes the insulin-degrading enzyme [15,65,145] that may interact with and degrade Aβ, inherited variants in SORL1 , and the presence of apolipoprotein E4 alleles . Additional risk factors for AD include head injury , low educational attainment and low linguistic ability early in life [176,177], diabetes [33,124,153], hyperlipidemia , hypertension , heart disease , smoking [136,137], elevated plasma homocysteine , and obesity .
The major barrier to treating and eventually preventing AD is a lack of understanding of the etiology and pathogenesis of neuron degeneration and loss. Clearly, AD is a disease of multiple, probably interrelated, etiologic/pathogenic factors. Numerous etiologic/pathogenic mechanisms have been suggested for the cause of AD including genetic defects [179,180], the amyloid cascade hypothesis (reviewed in ), trace element toxicity (reviewed in ), the oxidative stress hypothesis , mitochondrial defects , or a combination of the above. One hypothesis receiving renewed interest is the potential role of altered zinc (Zn) homeostasis in the pathogenesis of neuron degeneration mediated by Aβ processing and oligomerization and SP formation in AD.
Zinc is an essential trace element  that is redox inert with structural, catalytic, and regulatory roles [17,80,186] and functions as a crucial component in over 300 enzymes and transcription factors where it serves as an essential cofactor for catalytic activity  or by conferring structural stability to Zn finger domains of DNA binding proteins  including stimulating protein-1 (sp-1), a transcription factor responsible for ~30% of APP transcription [18,48,49]. In addition to its structural and catalytic roles, recent studies suggest free Zn has important signaling functions including inhibition of gamma aminobutyric acid (GABA-ergic) neurotransmission , modulation of protein kinase C (PKC) signaling pathways , modulation of p53 mediated DNA repair through stabilization of p53/genomic DNA interactions , and modulation of glycogen synthase kinase 3β [2,95].
In the brain, Zn is distributed in 3 pools including: a) a membrane-bound metalloprotein, or protein-metal complex pool involved in metabolic reactions and nonmetabolic functions such as biomembrane structure and support; b) a vesicular pool localized in nerve terminal synaptic vesicles; and c) an ionic pool of free or loosely bound ions in the cytoplasm . Of these sources, the vesicular pool, which is easily chelated, is thought to be the most important [50,69,86,155] because it is released during neurotransmission and may reach concentrations of 300 μM in the synapse. Unless these Zn gradients are immediately sequestered, they could potentially induce neurodegeneration. Overall, mean brain Zn concentrations are ~ 10 times serum Zn levels  and range between 150 and 200 μM [62,160]. Transport of Zn into the brain occurs via the blood/brain and blood/cerebrospinal fluid (CSF) barriers  where brain capillary endothelial cells respond to changes in Zn status by increasing or decreasing Zn uptake . Regionally, Zn concentrations, particularly vesicular Zn, are highest in the hippocampus, amygdala, and neocortex and are relatively low in cerebellum [51,72], a pattern that mirrors the distribution of pathologic features in AD. Zinc concentrations range from < 10−9 M in the cytoplasm in most neurons to > 10−3 M in vesicles of mossy fiber terminals [69,196]. Peak extracellular concentrations may reach 300 μM during seizure activity or during neurotransmission [7,89]. Therefore, it is imperative that cells regulate Zn levels through control of influx and efflux and through chelation to Zn sequestering proteins.
In general, Zn homeostasis is maintained by three families of proteins: a) metallothioneins (MT), b) Zrt-Irt like (ZIP) proteins, and c) zinc transporter (ZnT) proteins. Metallothioneins are a group of low molecular weight (6 – 7 kDa) single polypeptide chains of 60 or more amino acids, 25% – 30% of which are cysteines. These proteins display high Zn binding affinity (KZn = 3.2 X 10−13 M−1 at pH 7.4), can bind 7 atoms of Zn per molecule, and function to sequester Zn immediately after uptake by cells to prevent toxicity . Metallothioneins are ubiquitous and can be induced by metals, glucocorticoids, pro-inflammatory cytokines, oxidative stress, electrophilic compounds, and xenobiotics [100,152,187]. In brain, MT is expressed as 3 functional isoforms including MT-I and II, which are expressed in astrocytes, the perivascular space and pia mater  in a Zn-dependent manner [8,60] and MT-III, which is most abundant in neurons that sequester Zn in synaptic vesicles . The low redox potential of MT (−366 mV) allows mild oxidation to decrease Zn binding and facilitates release of Zn for binding to Zn finger and other transcription factors to modulate DNA binding efficiency and expression of genes involved in antioxidant responses during periods of oxidative stress .
Although the transport of Zn from brain extracellular environments to intracellular compartments in neurons and glia is not completely understood, it is thought to involve members of the ZIP family of proteins . ZIP proteins are predicted to have 8 transmembrane domains with a histidine-rich intracellular loop between domains 3 and 4  and are part of the plasma membrane or membranes of intracellular organelles. To date, 14 mammalian ZIP proteins that function to increase intracellular Zn by increasing Zn uptake (ZIP 1- 5; 7 – 15) or by releasing Zn from intracellular stores when Zn is deficient (ZIP 6 and 7) have been identified using mouse and human sequence analysis. In addition to ZIP proteins, neuronal Zn uptake may also be mediated by Zn-permeable membrane spanning channels including voltage gated L-type Ca2+ , N-methyl-D aspartate (NMDA) receptor gated , Ca2+ permeable AMPA/kainate channels  and Na+/Zn2+ exchangers .
Export and sequestration of Zn are carried out by the ZnT family of proteins that are predicted to have 6 transmembrane domains with a histidine-rich loop between transmembrane domains 4 and 5. To date, 8 ZnT proteins have been described (reviewed ). ZnT-1 is located at the plasma membrane and is expressed in the brain and other organs , whereas the other ZnT proteins are expressed at the membrane of intracellular organelles. ZnT-1 expression is induced in the presence of elevated cytoplasmic Zn through direct binding of Zn to the Zn-finger domain of metal response element-binding transcription factor-1 (MTF-1) (reviewed ). After binding Zn, MTF-1 adopts the appropriate DNA binding conformation and translocates to the nucleus where it binds the metal response element (MRE) in genes for ZnT-1, MT and gamma glutamylcysteine synthetase heavy chain which controls the rate-limiting step in glutathione synthesis (reviewed ). Overexpression of ZnT-1 in baby hamster kidney cells conferred resistance to increased Zn with the rate of Zn efflux increasing as extracellular Zn concentrations increased leading to the conclusion that Zn efflux mediated by ZnT-1 is an energy dependent process and argues against ZnT-1 being a channel or facilitated transporter . In contrast, Chowanadisa et al.  reported that rats provided a Zn-deficient diet demonstrated decreased brain ZnT-1, suggesting low systemic Zn could decrease ZnT-1 to maintain brain Zn stores. These studies are consistent with those of Takeda et al.  who found rats on a Zn-deficient diet showed increased brain Zn. Studies from our laboratory show ZnT-1 protein expression and function can be inactivated by HNE , a neurotoxic aldehydic marker of lipid peroxidation present in MCI and LAD brain [129,197]. Additionally, in the only study of ZnT-1 in AD, our Western blot analyses show significantly decreased ZnT-1 in the HPG of MCI, but significant elevations in EAD and LAD . More recent studies demonstrated that ZnT-1 reduces Zn influx through the L-type calcium channels (LTCC) without increasing Zn efflux [146,149,167].
ZnT-2 is a component of vesicular acid intracellular compartments, and in mice, is predominately expressed in intestine, kidney, and testis and is scarcely detected in brain . In in vitro experiments using baby hamster kidney cells, ZnT-2 overexpression conferred resistance to elevated Zn with sequestration into acidic compartments at higher concentrations . Additionally, co-expression of ZnT-1 suppressed ZnT-2 mediated transport into acidic vesicles suggesting ZnT-2 has a relatively low affinity for Zn and functions only under excessive elevations of Zn as a second line of defense when other ZnTs fail to function properly . ZnT-3 sequesters Zn in vesicles and is expressed only in brain and testis . In mouse brain, ZnT-3 is associated with hippocampal dentate granule cells, pyramidal and intraneurons as evidenced by levels of mRNA . A potential role for ZnT-3 and synaptic Zn in SP formation in transgenic mice was suggested by recent studies that showed crossing mice expressing mutant APP with ZnT-3 null mice led to diminished Aβ deposition . More recently, Friedlich et al.  showed that these mice also demonstrate reduced cerebral amyloid angiopathy that is hypothesized to be due to diminished Zn concentrations in the perivascular space of ZnT-3 null mice. In our studies of LAD and control brain, we find no significant alterations of ZnT-3 in LAD.
ZnT-4 exhibits considerable homology with ZnT-2 and 3 and is present in mammary gland and brain . ZnT-4 functions to sequester Zn in acidic vesicles and is best characterized in its involvement in the transport of Zn2+ into milk during lactation . Our study of ZnT-4 in MCI, EAD, and LAD brain is the first study of this protein in human brain and shows significant elevations of the protein in the HPG and SMTG of EAD and LAD subjects compared to age-matched controls . Although ZnT-5 mRNA is found in most organs in mice, the highest protein expression is in pancreas where is it associated with Zn-enriched secretory granules in insulin containing β cells and is scarcely detected in brain . In contrast to other ZnT proteins, ZnT-5 is predicted to have 15 membrane spanning domains and is ~ twice the size of other ZnT proteins .
ZnT-6 functions to sequester cytoplasmic Zn in the TGN and vesicular compartments  and in mice, ZnT-6 mRNA is present in multiple organs including brain. In the only studies of ZnT-6 in human brain, we show significantly increased ZnT-6 levels in the HPG of EAD and LAD subjects compared to NC and a trend toward a significant elevation in MCI . We also observed a striking association of ZnT-6 with NFT-bearing neurons identified using the modified Bielschowsky stain in LAD and in neurons positive for MC-1, a marker of early NFT formation in MCI . Similar to ZnT-6, ZnT-7 functions to sequester Zn in the TGN but has expression limited to lung and small intestine . Most recently, ZnT-8 has been characterized and is primarily associated with secretory granules of pancreatic β cells [110,162] where it likely plays a role in insulin transport. In mice ZnT-8 has limited expression in brain.
Although Zn is essential for normal brain function, experimental studies show high concentrations of Zn are toxic to neurons in vitro [37,59,203] and in vivo [40,47,114] with increased oxidative stress, and necrotic and apoptotic cell death occurring in as little as 30 min [37,76,133] . Unfortunately, the exact mechanism of Zn-induced cell death remains unclear. One possible mechanism is through the potentiation of glutamate [12,20,50,69,105,181], AMPA [23,37,73,112] or kainic acid [37,170,202] toxicity. Zinc may play a role in mitochondrial dysfunction through inhibition of the cell respiratory chain by blocking the initial step of respiration – the transfer of an electron between coenzyme Q and cytochrome b of the bc1 complex [19,93,111]. At high concentrations, Zn inhibits levels of complex I and II and cytochrome oxidase , although Yamaguchi et al.  demonstrated increased mitochondrial function in rat liver after a single low dose of Zn. More recent studies [32,88,118] show that nM – μM concentrations of Zn inhibit a number of enzymes required for mitochondrial respiration and glycolysis and that exposure to 100 μM Zn caused a 50% loss of intracellular phosphate . A dysfunction in oxidative phosphorylation, coupled with the presence of free radicals, could in turn lead to release of Zn from MT and increased intracellular concentrations of Zn . At relatively low concentrations, Zn inhibits sodium/potassium ATPase (Na+K+ ATPase) activity in isolated protein, inhibits glutamate and GABA uptake in mice synaptosomes  and glutamate transport by human excitatory amino acid transporter (EAAT) 1 in Xenopus laevis oocytes .
Zinc is hypothesized to induce structural abnormalities by influencing assembly and disassembly of tubulin [61,76,77] and several microtubule associated proteins in vitro [9,76,77,117]. In addition, Zn has been shown to mediate tau phosphorylation through modulation of P13/AKT, ERK1/2 and p38/MAPK signaling cascades . Zinc influx into cells by NMDA receptor channels may lead to a depolarization of the cells resulting in a rise of calcium (Ca) that could activate second messenger systems via PKC-mediated phosphorylation of receptor ion channels or voltage-dependent gene expression [8,144,164]. Zinc may also disrupt Ca homeostasis by binding to calmodulin  and through the inhibition of calmodulin-complexed Ca ATPase . Chelatable Zn accumulates in the cell perikarya of apoptotic neurons before and during degeneration after ischemic insult [117,185] and seizure activity , and is suggested to play a pathological role in neuron death. A more recent study using Zinquin, a Zn-specific fluorophore, demonstrated an increase in intracellular Zn levels as an early event in apoptosis that occurs in the absence of exogenous Zn and is consistent with a release of Zn from intracellular stores .
Interest in Zn and its possible role in the pathogenesis of AD was initiated in 1981 when Burnet proposed that Zn deficiencies led to dementia . Although initial studies of AD and control brain showed significantly decreased Zn in the hippocampus, inferior parietal lobule and occipital cortex of LAD subjects [3,4,45,55] later studies using short postmortem interval (PMI) tissue specimens from well characterized LAD and control subjects showed significant elevations of Zn in LAD hippocampus, amygdala, and multiple neocortical areas [44,51,53,63,166,194]. The differences observed between the studies may lie in the fact that formalin-fixed tissues were used in some of the earlier studies that also included control subjects that were not prospectively evaluated. Although multiple studies show alterations of Zn in late stage AD, there have been no studies of Zn concentrations in brain in MCI.
Despite considerable study of Zn at the bulk level, the cellular localization of Zn alterations in the progression of AD is unclear. Studies of Zn distribution in AD have primarily focused on the association of Zn with SP. Using micro-particle induced x-ray emission (micro-PIXE), we initially showed increased Zn in SP compared to adjacent neuropil and an elevation of Zn in LAD neuropil compared to age-matched NC . Several subsequent studies have confirmed those findings in AD [35,71,138,182] and in amyloid plaques in Tg2576 transgenic mice expressing mutant APP [74,122]. Recent studies using Raman microscopy to evaluate the structure and composition of isolated senile plaques Dong et al.  showed Zn2+ and Cu2+ are specifically coordinated with histidine residues and that chelation of Cu2+ led to a disruption of the β-structure of isolated plaques. Despite considerable study of Zn in SP, there have been relatively few studies that measure Zn in individual neurons in AD.
Although the subject of extensive study over the past 25 years, the reasons for elevated brain Zn in AD are unclear. Several studies have attempted to relate changes in peripheral Zn to elevated brain levels, although these studies have been contradictory. Haines et al. , Molina et al. , and Shore et al.  showed no significant differences between AD and control serum Zn, whereas Jeandel et al.  showed a significant decrease in Zn and other nutrients and antioxidant properties in AD serum, although the AD group may have contained malnourished subjects. The study of Haines et al.  may also be questioned because it included control subjects whose Mini Mental Status Examination (MMSE) scores were considered cognitively impaired. In contrast, Rulon et al.  and Gonzales et al.  showed significant elevations of Zn in AD serum. Additionally, Gonzales et al.  showed that serum Zn correlated with the presence of APOE4 alleles and concluded that of the parameters analyzed in their study, only serum Zn appeared to be an independent risk factor associated with the development of AD. In more recent studies of serum Zn in the progression of AD we observed a statistically significant decrease of serum Zn in men with MCI compared to women with MCI or age-matched normal control men . In contrast, we found no significant differences in serum Zn between well characterized LAD subjects and cognitively normal control subjects. Our observation of decreased serum Zn in MCI is of interest in light of previous in vivo rat studies in which systemic Zn deficiencies led to diminished ZnT-1 levels and increased brain Zn [38,147]. Our data support the hypothesis that elevated brain Zn in AD may be due to increased Zn uptake by brain under conditions of diminished extraparenchymal Zn in MCI.
Studies of CSF Zn levels have also been inconsistent with Molina et al.  showing decreased Zn in AD CSF compared to age matched control subjects whereas Basun et al.  showed no significant changes in AD CSF Zn. In addition, recent studies [78,183] show there is an inverse relationship between Zn and copper concentrations and levels of Aβ1-42 in CSF of LAD subjects and that degradation of soluble Aβ is normally promoted by physiologic concentrations of both Cu and Zn . Although the potential variation of Zn through the progression of AD is of interest there have been no published studies of CSF levels of Zn in MCI subjects.
Although several studies indicate alterations in Zn levels in the AD brain, the reason for its elevation and direct evidence for its role in the pathogenesis of AD has been lacking. Although Zn may play a role in multiple pathways relevant to AD, the most widely studied has been the possible role of Zn in processing of APP and aggregation of Aβ. APP synthesis is regulated by Zn-containing transcription factors (NF-κβ and sp1) and although Zn is essential for their activity [200,204,206], it is unclear whether the activity in vivo is regulated by Zn availability. In addition to the potential influence of Zn on APP expression, it also influences processing of the protein. Normal processing of APP by α-secretase cleavage in the Golgi complex leads to formation of sAPP, a neurotrophic factor . Additional proteolytic processing of APP by β-secretase (BACE) at the β-cleavage site [5,31,94,171,189] occurs in endosomes [107,115], where acidic pH necessary for β-secretase activity is possible . Further processing by the γ-secretase complex at the plasma membrane (reviewed ) leads to formation of Aβ, a 40 or 42 amino acid peptide that is the major component of SP in AD . Additionally, APP contains a ligand-binding site for Zn spanning the α-secretase position [24,25]. Zn concentrations less than 50 μM inhibit α-secretase mediated sAPP formation and increase generation of Aβ  perhaps through altered protein conformation. Additionally, high Zn concentrations can inhibit matrix metalloproteinase-2 (MMP-2) , which is able to partially degrade soluble Aβ1-42 in vitro . This inhibition of MMP-2 could lead to increased amyloidogenic Aβ levels. Recent studies suggest that most APP molecules are transported through the TGN where α-secretase cleavage likely occurs. Because Zn can significantly modulate APP processing leading to increased Aβ production, ZnT-6 mediated accumulation of Zn in the TGN could initially diminish α-secretase cleavage of APP. In addition, the presence of elevated Zn in endosomes mediated by ZnT-2 and/or ZnT-4 could further enhance β-secretase activity through modulation of pH.
In addition to its effect on APP processing, several reports indicate that Zn at low physiological concentrations [25–27,134] induces Aβ aggregation, although later studies indicate that higher Zn concentrations are required [41,66] for significant aggregation (fibril formation). More recent studies using atomic force and transmission electron microscopy and Aβ13-21 show Zn2+ specifically controls the rate of fibril assembly and regulates fibril morphology via specific coordination sites .
Multiple studies show that treatment of cortical neuron cultures with Aβ leads to increased levels of reactive oxygen species, increased lipid peroxidation, protein oxidation, mitochondrial dysfunction, caspase activation, and neuron death [29,32,104,201]. Additionally, several transgenic models of AD [82,126,148] including those with mutant APP, mutant APP/PS1 or mutant APP/PS1 and tau show increased Aβ deposition. Although Aβ deposits are associated with AD, the specific Aβ species responsible for neurodegeneration are unclear. Fibrillar Aβ, the predominant component of insoluble amyloid plaques, is neurotoxic [128,158]. However, in vivo, insoluble Aβ deposits do not accurately predict the severity of dementia in AD subjects . In addition, studies of transgenic mice including those with APP mutations show cognitive dysfunction and synaptic damage that preceed amyloid plaque deposition and neuron loss [96,119,140,143,195], leading to the suggestion that soluble oligomeric or protofibril Aβ species may the most toxic.
In vitro studies of synthetic Aβ show monomeric Aβ aggregates in a time-dependent manner that may be accelerated by Zn leading to oligomeric species, which may eventually form fibrils [39,157,191]. Increasing evidence suggests that it is these soluble oligomeric species that are the predominant neurotoxic species for neurons [54,109], leading to inhibition of long term potentiation in synaptic hippocampal slices [120,193] and calcium dysregulation and membrane dysfunction [54,102]. Although the exact Aβ species responsible for mediating neurodegeneration in AD are unclear, several lines of evidence support a role for oxidative damage mediated by Aβ or other oxidative species in the pathogenesis of AD including lipid peroxidation, protein oxidation, and DNA/RNA oxidation.
Based on the potential role of Zn and Cu in the deposition of Aβ in AD brain there has been considerable interest in the use of metal chelators to decrease amyloid pathology . In vitro studies show clioquinol (CQ), an 8-OH quinoline can inhibit aggregation of Aβ by Cu and Zn . In in vivo studies treatment of transgenic mouse models of amyloid deposition (Tg 2576) with CQ for 9 weeks significantly reduced amyloid plaque burden while simultaneously increasing brain Zn and Cu . In pilot, phase 2 double blind placebo controlled clinical trials CQ was shown to significantly slow cognitive decline in AD patients compared to placebo controls . More recently, PBT2, any 8-hydroxy quinoline with increased blood brain barrier permeability has been developed  and in a 12 week phase IIa clinical trial of AD subjects reversed frontal lobe functional deficits and significantly decreased Aβ1-42 levels in cerebrospinal fluid . Together, these data suggest modulation of Zn may be an effective potential therapeutic target in AD.
Although considerable evidence suggests a link between alterations in Zn and the proteins responsible for its uptake and sequestration in the progression of AD, there is a need for further in-depth study of Zn, ZIP and ZnT proteins, particularly early in the progression of AD (MCI) when therapeutic interventions would have greater efficacy. In particular there is a need for quantification of Zn levels in CSF of subjects with MCI and EAD and the correlation of those concentrations with brain ZnT, ZIP and Zn levels and the factors that regulate regional and cellular Zn concentrations. Based on in vivo studies, it is tempting to hypothesize that low extra-parenchymal Zn early in disease progression may lead to increased levels of brain Zn. These elevations of Zn could then become concentrated in subcellular organelles in which Aβ processing occurs leading to increased generation of oligomeric Aβ species and the promotion of oxidative damage associated with AD. As the disease progresses and extra-parenchymal Zn levels normalize, the resulting alterations in multiple ZnT proteins could further promote Aβ aggregation and SP formation.
Supported by NIH grants 1R01-AG16269, 5-P01-AG05119 and 1P30-AG028383, and by a grant from the Abercrombie Foundation. The author thanks Ms. Paula Thomason for editorial assistance.