Cholesterol levels as an AD risk factor
High levels of cholesterol, and particularly low-density liopoprotein (LDL) cholesterol, are a well-established risk factor for developing coronary artery disease and stroke. There is an emerging impression that these may also be an AD risk factor, based on human epidemiological studies, human neuropathology studies, and experiments using animal models of AD, each of which we will review in order in this first part.
As regards human epidemiological studies, high total cholesterol in serum has been reported to increase the risk of developing AD. In one study, this relationship was strengthened when apolipoprotein E genotype was controlled for2
. In another, high levels of LDL or total cholesterol were reported to correlate with lower Modified Mini Mental State Exam scores in clinically nondemented patients3
. High total cholesterol levels at midlife have been associated with a nearly threefold increase in the likelihood of developing AD (OR=2.8, 95% CI=1.2–6.7), even after controlling for apo E genotype4
. It has been further suggested that cholesterol levels exert some influence over the well-established correlation of the ApoE4 allele with risk of AD5
. Solomon and colleagues6
found that total cholesterol levels at midlife (mean age = 50.4 years) were higher in those who ultimately developed mild cognitive impairment (MCI)/dementia than in those who did not, and this relationship was unaffected by ApoE status. Similarly, Whitmer et al.7
concluded that high total cholesterol at midlife (mean age = 42 years) substantially increased the risk of “late-life” (range=61–83 years) dementia (HR=1.42, 95% CI=1.22–1.66). AD patients with high total or LDL cholesterol can experience faster cognitive decline than those with normal cholesterol measures, as can those with an ApoE4 allele8
. For ApoE4 non-carriers, high HDL levels also contributed to faster cognitive decline8
. Triglycerides did not significantly impact rate of decline in the latter study population, regardless of ApoE4 status. In another epidemiological study, high levels of neutral lipids in peripheral blood mononuclear cells were much more common in AD patients than in healthy age-matched controls, and plasma HDL levels were reduced9
. Zambon et al.10
concluded that individuals with familial hypercholesterolemia were more likely than healthy persons to develop MCI, the amnestic form of which is commonly a harbinger of AD.
Although these diverse studies suggest a consensus that high cholesterol levels confer an increased risk of developing AD, some reported results do not fit with this conclusion. Reitz et al.11
, for example, found that high total cholesterol levels in people 77 years and older decreased
the risk of AD (HR=0.48, 95% CI=0.26–0.86). The individual effects of HDL and LDL cholesterol, however, were not significant. This group did not find any significant effect of total cholesterol, HDL, LDL, or triglycerides on cognitive ability in healthy elderly persons12
. A later study by Reitz and coworkers13
found that high total cholesterol or LDL levels were associated with a decreased risk of developing MCI in people 65 years and older (although this was dependent on which factors were adjusted). Mielke et al.14
similarly found that high total cholesterol levels between the ages of 70–79 reduced the risk of developing dementia between the ages of 79–88.
Two major caveats must be considered when evaluating the epidemiological studies of cholesterol and AD risk. First, many such studies measured only total cholesterol and thus could not detect differing effects of HDL and LDL. Second, when participants in a study already have AD at entry, it can be difficult to determine whether changes in cholesterol levels are having an effect on the progression of the disease or conversely whether the pathophysiological changes that accompany AD alter cholesterol levels. This second consideration is particularly important, as many published studies have been conducted relatively late in the patient’s life, when substantial AD-type neuropathology may already be present. As shown in , studies finding a negative correlation between cholesterol levels and dementia risk (gray) were principally conducted late in the patients’ lives, whereas those finding a positive correlation (red) tended to be conducted earlier (left end of each balloon represents mean age at study start; horizontal dimension of balloon indicates length of follow-up).
Relationship between outcome and age of participants in studies of cholesterol’s effect on dementia risk and progression
In addition to clinical epidemiology, correlative human neuropathological studies have been conducted. Early work in this area demonstrated that amyloid pathology was more common in patients with heart disease than in healthy subjects15
. One neuropathological study16
noted differences in the distribution of AD-type pathology associated with high cholesterol at different ages: high HDL levels at both midlife and in late-life were associated with greater neurofibrillary tangle pathology in the neocortex, but only late-life HDL levels were associated with increased plaque pathology in hippocampus or neocortex. This is an interesting observation, as HDL is generally considered to be atheroprotective. The authors speculated that alterations in HDL metabolism could change neuronal membrane composition, and in this context, they raised the possibility that HDL might affect amyloid β-protein (Aβ) production, aggregation, and/or clearance, possibly through an ApoE-mediated mechanism. They also noted that ApoE genotype didn’t influence the effect of cholesterol on AD brain pathology. Another study17
found a link between high cholesterol levels and brain amyloid pathology in 40–55 year old subjects but not in older people, suggesting that cholesterol levels during the presymptomatic stage -- particularly in middle age -- are important.
Some of these human neuropathology results appear to be borne out in animal models of AD. Feeding amyloid precursor protein (APP) transgenic mice a high-fat, high-cholesterol diet increased the number and/or size of amyloid plaques18,19
and led to higher Aβ levels in formic acid extracts of brain18
. One of these studies observed a positive correlation between plasma levels of murine ApoE and Aβ deposits in mice expressing the Swedish mutation of human APP19
. Feeding rabbits a high-cholesterol diet doubled Aβ levels in the hippocampal cortex (although this effect did not reach statistical significance) and also caused damage to the BBB20
. Interestingly, returning rabbits to a normal diet after several weeks of a high-cholesterol diet reversed the increase in Aβ levels15
Potential mechanisms for cholesterol’s apparent adverse effect on the development of AD
There are several possible mechanisms that could explain the studies reviewed in the previous section that seem to connect high cholesterol levels with the development of AD neuropathology. Cholesterol may increase the activity of the β- or γ-secretase enzymes that generate Aβ from APP, decrease the flux of APP through the non-amyloidogenic α-secretase pathway, or affect various non-amyloid factors such as local inflammation or tau metabolism.
Partial repression of the non-amyloidogenic α-secretase pathway is one route postulated from animal studies by which high cholesterol levels could increase AD risk. Application of exogenous cholesterol to human embryonic kidney (HEK) cells overexpressing human APP decreased the α-cleavage product of APP, APPsα21
, as did feeding mice a diet high in fat and cholesterol18
. Conversely, treatment of HEK cells, human neuroglioma cells, or APP-overexpressing astroglioma cells with the cholesterol-extracting agent methyl-β-cyclodextrin (MβCD) increased the secretion of APPsα22
. Another study found that progesterone, which decreases transport of cholesterol from the plasma membrane to the cytosol, did not alter the cholesterol-mediated inhibition of APPsα secretion23
. This result suggests that cholesterol may act on APP primarily at the cell surface.
Cholesterol could also exert its effects more directly by influencing the β- and γ-secretase cleavages that produce Aβ. Depletion of cholesterol from human APP-expressing rat hippocampal neurons by applying lovastatin and MβCD strongly decreased the amount of Aβ produced24
. This effect occurred without changing the levels of APPsα and was reversed by restoring cholesterol. The authors suggested two possible reasons why cholesterol depletion could affect β-cleavage but not α-cleavage: intracellular transport of APP might be affected in such a manner that APP colocalizes less with β-secretase, or β-secretase may be principally active in cholesterol-rich lipid rafts24
. The application of cholesterol-lowering agents to APP-overexpressing HEK cells has also been shown to inhibit β-cleavage of APP; conversely, addition of exogenous cholesterol enhanced β-cleavage and led to increased secretion of both Aβ40 and Aβ4225
. Interestingly, treating neuronal cells with compounds that interfere with intracellular cholesterol transport inhibited β-secretase activity but promoted γ-secretase activity26
. In a separate study, removal of cholesterol appeared to decrease γ-secretase activity, an effect that could be reversed by restoration of cholesterol27
Once Aβ has been produced, cholesterol levels could influence its aggregation state. Increased interactions of Aβ with the cell membrane under low-cholesterol conditions may allow greater internalization and degradation of the peptide; conversely, high cholesterol levels could make it more difficult for Aβ to associate with the cell surface, leading to its accumulation and aggregation in the extracellular space28
. Furthermore, a group that had previously described a novel 5 kDa form of Aβ (Aβ monomer is 4 kDa) having unique properties that were dependent on the presence of cholesterol demonstrated that this Aβ isoform could act as a seed for fibrillar aggregation when it was incubated with synthetic Aβ peptides29
. Treating cells with compactin (an HMG-CoA reductase inhibitor) or filipin (an antibiotic that binds to cholesterol) inhibited this seeding effect, and that inhibition was countered by application of exogenous cholesterol. The authors suggest that this special 5 kDa “seeding form” of Aβ is likely to be produced in lipid rafts.
In addition to its apparent effects on APP processing and Aβ, cholesterol also exerts a range of pleiotropic effects on neuronal physiology. For example, cholesterol depletion had adverse effects on dendrite growth and axonal branching, even when the chemical used to deplete cholesterol didn’t interfere with the synthesis of downstream isoprenoids30
. In view of this and many other reports that well-regulated cholesterol helps maintain healthy neurons, there has been great interest in whether statins, which lower LDL cholesterol by inhibiting HMG-CoA reductase, or other lipid lowering agents (LLAs) may alter the risk of developing AD or be beneficial for its treatment.
The current state of knowledge on statins and AD
Because statins are already in widespread use, the possibility that they might be useful for AD treatment or prevention must be rigorously confirmed or denied. In the years since the connection between cholesterol and AD was uncovered (above), many studies have investigated the potential use of statins as AD-modulating compounds. However, the results of these studies have often been inconsistent, in large part due to major differences in study design and data analysis. While these methodological differences make it difficult to synthesize the results from various studies, we have endeavored to identify factors that could explain the observed variation in study outcomes.
Several studies in cell culture () have indicated that statins can reduce Aβ levels. Application of lovastatin (along with MβCD) to rat hippocampal neurons expressing human APP decreased Aβ production without affecting the quantity of APPsα or the p3 peptide that arises from α- followed by γ-secretase cleavage24
. Another study also found that inhibition of cholesterol synthesis with lovastatin decreased Aβ formation31
. Yet another study reported that treatment of various cell lines with lovastatin increased α-secretase activity, and that treatment of human astroglioma cells with lovastatin lowered Aβ production22
. Furthermore, Aβ-induced release of lactate dehydrogenase from human neuroblastoma cells was abolished by application of mevastatin32
. In hippocampal or mixed cortical neurons from rats, treatment with either simvastatin or lovastatin reduced levels of both intracellular and extracellular Aβ40 and Aβ4233
. This study also demonstrated that lovastatin and MβCD increased levels of C-terminal fragment-alpha (CTFα) in primary neurons carrying the Swedish missense mutation in APP, suggesting a stimulation of α-secretase processing. Taken together, these cell culture studies suggest that statins can lower Aβ generation.
Preclinical studies of the efficacy of LLAs for the treatment and prevention of AD and dementia
Animal models of AD have generally yielded complementary findings (). Administering simvastatin to guinea pigs for three weeks resulted in decreased brain and cerebrospinal fluid (CSF) levels of Aβ, an effect reversed by discontinuing the treatment33
. In another study, lovastatin and pravastatin each reduced the amount of Aβ in the brains of transgenic mice, while simultaneously increasing levels of APPsα34
. Transgenic mice treated with simvastatin also performed better on the Morris water maze test than their untreated counterparts35
The large body of literature on a putative connection between elevated cholesterol levels and increased AD risk suggests that lowering cholesterol might be a viable strategy for AD treatment or prevention, although differences among studies in the age of participants are a complicating factor. Based on this evidence, numerous researchers have investigated the potential therapeutic effects of LLAs, with a focus on statins. The results of preclinical research on this topic (above) are encouraging, but human studies have been far more inconsistent in outcome. In the forthcoming second part of this review, we will examine these clinical studies in detail, elucidate possible reasons for this observed variability, and make recommendations for future human studies of this important topic.