Patterns of Compensation and Vulnerability in Normal Subjects at Risk of Alzheimer’s Disease

Abstract

Alzheimer’s disease (AD) is the most frequent form of dementia in elderly individuals and its incidence and prevalence increases with age. This risk of AD is increased in the presence of genetic and demographic factors including apolipoprotein E 4 allele, lower education, and family history of AD. There are medical risk modifiers including systemic hypertension, diabetes mellitus, cardiovascular disease, and cerebrovascular disease that increase the vulnerability for AD. By contrast, there are lifestyle risk modifiers that reduce the effects of AD risk factors include diet and physical and cognitive activity. Our research has consistently shown that it is the interactions among these risk factors with the pathobiological cascade of AD that determine the likelihood of a clinical expression of AD—either as dementia or mild cognitive impairment. However, the association between “vulnerability” and “protective” factors varies with age, since the effects of these factors on the risk for AD may differ in younger (age < 80) versus older (age > 80) individuals. The understanding of the dynamic of these factors at different age periods will be essential for the implementation of primary prevention treatments for AD.

INTRODUCTION

Alzheimer’s disease (AD) is the most common form of dementia in the elderly, and age is the most important risk factor. The prevalence of AD increases exponentially between the ages of 65 and 85, approaching 50% in the oldest old (>85 years old) [1, 2]. After 90 years of age, the incidence of AD increases dramatically with older age, from 12.7%/year in the 90–94 age group, to 21.2%/year in the 95–99 age group, and to 40.7%/year in those >100 years old [3].

This risk of AD is increased in the presence of genetic and demographic factors including apolipoprotein E (APOE) 4 allele, age, female gender, lower education, and family history of AD [1, 4, 5]. Medical risk modifiers include systemic hypertension (HTN), diabetes mellitus (DM), cardiovascular disease, and cerebrovascular disease (CVD) [6-11]. Lifestyle risk modifiers that reduce the effects of AD risk factors include physical and cognitive activity, and diet [12-14]. Our research over the past 10 years has consistently shown that it is the interactions among these risk factors with the pathobiological cascade of AD determine the likelihood of a clinical expression of AD—either as dementia or mild cognitive impairment (MCI).

AGING AND DEMENTIA

There are two critical assumptions that must be understood in the study of risk factors for AD. First, it has become increasingly apparent that the AD pathology is present several years before the development of the clinical symptomatology [15]. This distinction is critical because there are few, if any, unambiguous genetic, metabolic, or vascular factors that alter the risk to express the pathology. However, there is an ever-increasing set of variables that alter the risk to express the dementia. The core of our research has been to examine how those variables modify the risk for the clinical syndrome either by increasing or decreasing vulnerability. The second assumption is that if we accept the position that age itself is less important as a risk modifier then the various medical factors that change as a consequence of age [16]. For example, in our analysis of 924 cognitively normal individuals participating in the Cardiovascular Health Study-Cognition Study (CHS-CS), we found the commonly observed correlation between performance on the digit symbol substitution task (DSST) and chronological age (r = −0.32, p < 0.001) [17]. However following a series of regression analyses, we found that DSST performance was uniquely predicted by brain volumes in the hippocampus and frontal cortex. The structural integrity of these two brain regions was predicted not only by the presence of the APOE4 allele, but also by the presence of white matter lesions (WMLs), an indicator of the presence of small vessel disease. Age did not have a direct effect on DSST performance; it acted only by increasing the risk for WMLs and independently decreasing the volume of the hippocampus. Thus, one of our goals in our research program has been to try to “minimize” the effects of age by being able to more fully decompose its effects into the component parts (e.g., vascular disease, metabolic syndromes, etc.).

VULNERABILITY FACTORS

Aging

Older age has a global impact on brain structure [18], but the effects are not homogenous [19]; some areas are more vulnerable than others [20]. Atrophy in the prefrontal, frontal, and calcarine cortices [19, 21, 22] and hippocampal/parahippocampal regions [23-25] is common with increasing age. However, the cross-sectional studies of age effects are limited because we do not know antecedent information and, most important, we do not know what happened after the scan; it is possible that within a short period of time they progressed to dementia (e.g., [26]) and, consequently, we may attribute to age effects those that are actually due to neurodegeneration.

We took advantage of the longitudinal CHS-CS data by examining the magnetic resonance imaging (MRI) scans of 169 cognitively normal subjects who remained cognitively normal 5 years after the scan, and 33 AD patients [27]. We found that with older age, there were lower gray matter (GM) volumes in sensorimotor and heteromodal association areas in frontal, temporal, occipital, and parietal lobes, as well as in the cerebellum. While age was associated with atrophy in the posterior hippocampus, thalamus, and middle cingulate gyrus, AD was associated with atrophy in the anterior hippocampal/parahippocampal regions and also in the precuneus. Normal aging and AD strongly overlapped in the hippocampus, specifically in the entorhinal cortex, and Cornu Ammonis. Age and AD are independently “connected” with GM atrophy but overlap substantially in the hippocampus, including entorhinal cortex.

The size of the hippocampus decreases 0.86% per decade, increasing to 1.85%/decade after age 70 [28]. However, the rate of atrophy was associated not only with baseline brain parenchymal fraction, but also hemoglobin A1c and severity of the WMLs [29], which themselves are more common with advancing age. The Atherosclerosis Risk in Communities study showed that DM, HTN, history of stroke, and the presence of the APOE4 allele contributed to cognitive decline in late middle age subjects followed during 14 years [30]. Studies that examined ventricular expansion over time noted that this process is faster in subjects who will develop MCI [31], although Carmichael and colleagues, using the CHS-CS database, found that the rate of expansion can be accelerated by DM and HTN in normal individuals [32]. Taken together these observations, it seems that vascular risk factors play a role in the point of inflection of the slope that leads to cognitive deficits during the aging process.

Vascular disease

The relationship between vascular disease and AD pathology has not been clearly established, although ischemic vascular lesions may modulate AD clinical manifestation by expressing clinical dementia with fewer AD pathological changes [33, 34]. Cerebral infarctions increased the odds of clinical dementia in persons with AD pathology [35, 36] and arteriosclerosis in the Circle of Willis was more severe in AD patients than in non-demented subjects [37]. By contrast, other studies have reported no association between AD and brain infarcts [38, 39]. These pathological studies provide important insights into the relationship between AD and CVD but they are cross-sectional, do not provide a temporal relationship between the two pathologies, and they only address severe CVD (i.e., infarcts). Nevertheless, they have suggested that the critical vascular disease influence on AD pathology occurs in early stages of the clinical syndrome, but not in more advance stages [35].

Although the prevalence of dementia continues to increase with age, pathological studies found that AD pathology can be seen in a high proportion of autopsied cognitively-normal subjects after age 80 + [40-42] suggesting that some individuals have sufficient “brain reserve” to tolerate this pathology. On the other hand, it has been found that the association between neo-cortical neuritic plaques and dementia in subjects age 95+ was attenuated [41], which was due primarily to increased pathology in the cognitively normal subjects, and the density of AD pathology is reduced in subjects age 90+ compared to those age 80 or younger [40]. A recent study showed that the primary cause of dementia in subjects age 90 + was still AD pathology [43]. These findings suggest that the threshold to express dementia is not the same in all individuals, some can manifest the symptoms with less pathology than others, and the border zone between normalcy and dementia is very “narrow” in subjects 90+.

WMLs, as a manifestation of a subclinical CVD (see below), have been reported as risk factors for MCI [44] and AD [6]. WMLs are associated with hippocampal and amygdala volumes [45, 46]. WMLs correlate with age, HTN and DM, urinary incontinence, and loss of strength in lower limbs [47-51]. Sub-clinical systemic vascular disease has also been shown to increase the risk for dementia. Markers of peripheral atherosclerosis [52] and measures of pulse pressure and pulse wave velocity, as markers of arterial stiffness, have been associated with cognitive impairment [53]. We have shown that incident AD was associated with prevalent peripheral artery disease (i.e., ankle-arm index), and with cardiovascular disease, defined as internal carotid artery wall thickness (in mm) [52]. These findings indicated that subjects with vascular disease other than stroke and MRI-infarcts had higher risk of AD than did those without vascular disease.

HTN is very common in old individuals, and the fact that subjects with HTN have a diminished cerebrovascular dilative response to physiological stimuli has served as the basis for the studies of the effects of the regional cerebral blood flow (rCBF) responses on cognitive stimulation [54-57]. Several studies have found a pattern of decreased MRI volume and altered metabolism/perfusion in middle-aged and elderly subjects with HTN [45, 55, 58-60], as well as cognitive deficits [56, 61]. A longitudinal study conducted in 14 highly selected hypertense, cognitively normal subjects (mean age: 60), with biannual H₂¹⁵O-PET scans, showed decreased rCBF overtime in the middle and inferior prefrontal cortex, anterior cingulate gyrus, and occipital-temporal cortex [60]. A recent study showed an association between HTN and amyloid deposition, especially in APOE4 allele carriers [62].

Perfusion MRI studies conducted by the CHS-CS in cognitively healthy individuals showed that there was diminished rCBF in heteromodal association areas usually affected in AD in hypertensives compared to normotensives [63]. These findings were consistent with previous observations in untreated, relatively young (mean age 61), and cognitively normal subjects with HTN [55, 61], who had reduced rCBF and less cortical grey matter in the hippocampus, anterior cingulate gyrus, and middle temporal and parietal lobes. Mid-life HTN is associated with hippocampal volume after age 65 [64], and low cerebral blood flow, as measured with transcranial Doppler, and is associated with the development of dementia [65]. We have reported a relationship between dementia and prevalent and incident stroke, MRI-identified infarcts, and the severity of WMLs in CHS [66].

HTN and small vessel ischemic disease increase the risk for AD; how they affect brain structure in cognitively-normal subjects is not clear. Increased aortic stiffness, even in subjects without overt HTN, was associated with lower memory scores, MRI silent infarcts, and WMLs in non-demented subjects [67]. Such insights are important in understanding how these factors raise brain vulnerability for neurodegeneration. We found that severity of WMLs inversely correlated with GM volumes, especially in the frontal cortex [17], and that age-related atrophy was observed in the hippocampus and posterior cingulate gyrus. Regression analyses revealed links among age, HTN, WMLs, APOE4 allele, GM, and digit symbol substitution scores. Both advancing age and HTN predicted WMLs, which is itself associated with GM atrophy. These data demonstrated that small vessel disease, as indicated by WMLs, is associated with GM tissue loss that results in decreases in cognitive function in cognitively normal subjects. This observation showed that multiple factors have to be taken into account to interpret the effects of WMLs on brain function and in normal individuals.

There is an association between DM, abnormal glucose metabolism, and incident dementia [10, 11, 68-73], especially when DM is uncontrolled [74], and in APOE4 allele carriers [75]. The CHS-CS found that DM was associated with atrophy in the frontal, temporal, and subcortical regions, while insulinemia was associated with atrophy of the frontal lobes, hippocampus, and corpus callosum [76]. However, these associations were attenuated when we controlled for body mass index (BMI).

Renovascular disease and markers of inflammation

The CHS has found that MRI-infarcts and WMLs were predictors of incident dementia [6], and that elevated CRP and IL-6 (as markers of low grade inflammation) were associated with myocardial infarction, WML, and death [77-79]. This indicated that inflammatory markers were present in individuals with cardiovascular and cerebrovascular diseases, suggesting that their relationships with AD pathology could be related to underlying vascular disease. CRP and IL-6 levels have been found associated with silent infarcts in the brain [80, 81], and it has been suggested that inflammatory markers (CRP) were involved in the pathogenesis of small vessel disease, in particular the development of WMLs [82, 83]. Furthermore, several studies conducted in non-selected populations found that CRP was associated with WMLs, which subsequently led to executive function deficits [83, 84].

There are not many studies that have examined the relationship between brain volumes with inflammatory markers in normal or demented subjects. A study conducted by the Framingham Heart Study on 1926 participants age 35–85 (mean 60) found that markers of inflammation were associated with MRI total brain volume, even after excluding participants with incident cardiovascular disease [85]. This study confirmed that inflammatory markers are associated with greater brain atrophy than expected for age. However, it did not examine the specific brain regions affected by these markers, and there is no knowledge about selective brain vulnerability to inflammatory markers in elderly subjects (age > 75).

Cystatin C is considered a marker of glomerular excretion function [86, 87], and it has been associated with the presence of peripheral vascular disease [88], cardiovascular disease [89], stroke [90], subclinical brain infarction [91], markers of inflammation [92], and death [93, 94]. The CHS has also found a correlation between plasma amyloid-β (Aβ)_1-40 and Aβ_1-42 levels and cystatin C levels in normal subjects [95], which it may explain the relationship between reduced insulin clearance and plasma Aβ levels in AD patients [96], and the modulation of plasma Aβ levels by drugs (e.g., non-steroidal anti-inflammatory drugs) that affect glomerular function in normal subjects [97]. However, the relationship between cystatin C and AD is not well established. Some studies found that low cystatin C levels predicted AD [98], and they speculated that this was secondary to an increased deposition in the brain parenchyma, as part of a physiological process of neuroprotection [99]. By contrast, other studies have found no association [100] or increased levels of cystatin C in the cerebrospinal fluid of AD patients [101]. The current state of the research suggests that cystatin C in a marker of renovascular disease, which it should be increased in individuals with vascular disease in general.

These studies have shown sub-clinical vascular disease is a widespread phenomenon that affects kidney and cardiovascular function, and brain structure. Therefore, it is expected that measures of cardiovascular disease, cystatin-C levels, and markers of inflammation will be higher in normal subjects who will convert to MCI/AD.

Cholesterol metabolism

Epidemiological studies have shown that total cholesterol levels in mid-life were associated with dementia in old age [102], and the association was attenuated with statin use [103]. However, these findings have not been replicated in other studies [104, 105]. Similarly, the relationship between cholesterol and AD in the elderly was not consistent; some studies found that low [106] or high [107] cholesterol levels were associated with AD in elderly subjects. It is possible that the discrepancy was related to the age of the cohorts and to timing in the use of lipid-lowering therapy, although a trial of pravastatin in subjects at risk for cardiovascular disease found no difference between the placebo and treatment arms in cognitive function after 3-year follow-up [108].

Although the relationship between AD and systemic cholesterol could be explained by increased CVD [37], it is important to know that the central nervous system (CNS) cholesterol is produced in the brain, and lipid metabolism is essential for neuronal functioning. Therefore, markers of CNS metabolites (e.g., 24S-hydroxychoelestrol) that are formed only in the brain and that can cross the brain blood barrier could be better markers of brain cholesterol metabolism than those of peripheral etiology [109]. Finally, the role of the APOE as a lipid transporter and in lipid delivery for growth and repair of axons after an injury underscores the importance of the relationship between lipids and AD [110]. Subjects carrying the APOE4 allele, a risk factor for AD [111], have low levels of APOE and consequently decreased cholesterol transportation from neuronal bodies to axons with the subsequent reduction of axonal extension and sprouting [110]. The relationship between cholesterol metabolism and neurodegeneration is not well-understood, and it is possible that systemic and CNS cholesterol-associated factors converge to create a vulnerability state for AD.

Neuroimaging in normal subjects at risk for MCI and AD

PET, SPECT, and MRI have demonstrated abnormal cerebral structure and metabolism in the mesial temporal lobes and other heteromodal association cortices in MCI subjects who progressed to AD [112-124], and in normal subjects at risk for AD [125-127], especially those who are homozygous for APOE4 allele with family history of dementia.

While perfusion studies showed that an association between temporal, parietal, and posterior cingulate regions and subsequent conversion to AD in normal subjects, structural volumetric studies showed that hippocampal, amygdala, and the ventral striatum volumes were predictors of AD [26, 128, 129]. The CHS-CS examined the MRI volumetric characteristics of the normal subjects who progressed to AD during a 4.5 years follow-up using modulated voxel-based morphometry found that the ventral striatum and hippocampal volumes were associated with incident AD [26]. This finding is particularly important since amyloid ligand studies have found amyloid deposition in the ventral striatum and frontal lobes in patients with early AD [130, 131] and in cognitively-normal subjects [132]. In another study, we found that hippocampal volume and MRI-infarcts were independent predictors of conversion to dementia in normal subjects [128]. These data showed that both the ventral striatum (which includes the basal forebrain) and hippocampal volumes were associated with conversion to AD; and the conversion was faster in those subjects who had greater basal forebrain atrophy. Therefore, the perfusion and volumetric studies indicated that there is a widespread compromise of cortical and subcortical structures in normal subjects destined to develop AD.

COMPENSATORY MECHANISMS

One important aspect of the transition from normalcy to dementia is that in spite of the presence of progressive neuronal loss and atrophy, these subjects are able maintain a certain level of functional capacity. For example, PET studies that used activation paradigms [133, 134] have found that subjects with mild/moderate AD have a greater activation than normals in the brain regions usually involved in memory tasks, as well as in others (frontal lobes) that only activate with increasing task difficulty. This compensation response occurs without necessarily resulting in improved cognitive function. Similarly, fMRI studies have found that MCI subjects have a greater activation of the hippocampus compared to normal controls and AD subjects in memory tasks [135]. Studies that used attentional/executive tasks found increased activation in the parietal lobes, with decreased activation in the prefrontal cortex, and anterior cingulate gyrus [136].

We have extensively examined the effect of physical activity and obesity on brain volume in cognitively normal individuals in the CHS-CS [13, 137-140]. Because there is a closed association between educational level and physical activity in elderly subjects, we examined whether educational level or physical activity (based on leisure-time energy expenditure) was associated with brain tissue expansion in cognitively normal subjects in CHS-CS [139]. After controlling for age, gender, and physical activity, higher educational levels were associated with ~2–3% greater tissue volumes, on average, in the temporal lobe GM. When controlling for age, gender, and education, greater physical activity was associated with ~2–2.5% greater average tissue volumes in the white matter of the corona radiata extending into the parietal-occipital junction. However, the effects of education and exercise on brain structure were attenuated when including BMI in the statistical model. This study showed a dynamic role of vulnerability (obesity) and protective (physical activity) factors on the aging brain, although BMI may be the dominant explanatory factor mediating the observed relationship among education, physical activity, and brain structure; there is an negative association between education level and obesity [139].

We have also conducted longitudinal studies to examine the effect of physical activity and brain volume in the CHS-CS. Erikson and colleagues examined whether physical activity (measured in 1991–92) was associated with greater GM volume (MRI in 1998–99) after a 9-year follow-up, and whether a threshold could be identified for the amount of walking (measured in blocks walked per week) necessary to spare GM volume, and whether GM volume associated with physical activity would be associated with a reduced risk for cognitive impairment (MCI/dementia) 13 years later (after 2002–03) in cognitively normal subjects [13]. After controlling for WMLs, MRI-identified infarcts, and other variables (including BMI), greater physical activity predicted greater volumes of frontal, occipital, entorhinal, and hippocampal regions 9 years later. Walking 72 blocks per week was necessary to detect increased GM volume, and greater GM volume associated with physical activity reduced the risk for MCI/dementia 2-fold between 1998–99 and 2002–03. Therefore, in this longitudinal model, physical activity had a long-term effect, as it was associated with greater GM volume, which is in turn associated with a reduced risk of cognitive impairment. These studies have shown the positive effects of physical activity on the brain, although vascular factors can attenuate these effects.

Perfusion MRI studies have also shown that early AD patients can also have areas of increased rCBF (inferior frontal and anterior temporal cortices) compared to controls [141, 142]. The CHS-CS found that there were decreases and increases in rCBF in specific brain regions in early AD and MCI (e.g., anterior cingulate gyrus, ventral striatum, hippocampus) [142]. A study with perfusion-weighted MRI showed increased rCBF in the medial temporal region, amygdala, and anterior cyngulate gyrus in a subgroup of MCI subjects [143]. In addition, this latter study found a tendency to increased rCBF in the frontobasal region when both groups (AD/MCI) were examined together. This is particularly important in view of studies conducted in Pittsburgh where we examined the relationship between amyloid ligand of Aβ deposition with cerebral metabolism in amyloid-positive control subjects and patients with MCI or AD [144]. Glucose metabolism in parietal and precuneus cortices of AD patients was negatively correlated with Pittsburgh Compound-B (PiB) retention. By contrast, in MCI patients with Aβ deposition, there was a lack of negative correlations between Aβ and metabolism, but there were frequent positive correlations. This suggested that, at the MCI stage of the disease, Aβ deposition was independent of brain metabolism in specific brain areas, which was similar to the pMRI studies conducted in CHS-CS.

The hypothesis that educational level, through the development of increased synaptic formation, is a plausible explanation for the “protective” effect in cognitively-normal subjects. However, there are other critical factors that are associated with education and incident AD. For example systemic vascular disease is highly prevalent in the elderly and the treatment and care of this condition is closely associated with education level [145-147]. Indeed, pathological studies have found an association between vascular disease and education level, but not with AD pathology [148], although this was not confirmed by others [149]. Subjects with low education level may seek diagnosis and treatment after those with high education level, and they may not be able to receive the proper treatments. Therefore, the role of the educational factor in the clinical manifestation of AD may not be related to the years of education, but to treatment and diagnosis of vascular disease. Finally, we have shown that cerebral ventricular volumes progressed faster in elderly subjects with HTN and DM than in those without these conditions [32, 150].

SUMMARY

The studies conducted at the CHS-CS as well as those described in the literature suggested that vascular disease creates a vulnerability state to AD pathology and modulates its clinical presentation. Our studies have been oriented to examine whether vascular disease can accelerate the transition from normal to abnormal cognition, either MCI or AD, by attenuating the effect of physiological compensatory mechanisms. In addition, we are examining the relationship between vascular disease and amyloid deposition using PiB. The understanding of this process will determine when and how the cerebral vascular process could be treated, and how this process influence future primary prevention therapies for AD.