This study demonstrated the unexpected result that a brief treatment with a low carbohydrate/high saturated fat diet reduced total Aβ levels in a mouse model of Alzheimer's disease. Previous studies had suggested that diets rich in saturated fats or cholesterol increased both the production and deposition of Aβ in mouse models of AD, leading to the suggestion that diets rich in lipids were a factor in AD [10, 12–14]. However, these diets were not low carbohydrate diets. In the high cholesterol diets, cholesterol was added to the diet without reduction in other components [10, 14]. In the studies of high fat diets, carbohydrate content was still relatively high. For example, Ho et al. used a diet of 60% fat, 20% carbohydrate, 20% protein. This diet was sufficiently high in carbohydrate to cause large increases in body weight in the animals .
The interaction of different macronutrients, in particular fats and carbohydrates, is known to influence the metabolic state of the animal. For example, Marsset-Baglieri et al. examined if fat in the diet alone was sufficient to shift energy balance toward fat storage. Yet, rats fed ad libitum high fat (50%) diets devoid of carbohydrates did not increase energy intake and did not gain in body adiposity, while animals fed high fat diets (30%) in the presence of carbohydrates (56%) increased energy intake and gained in body fat . Such studies support the view that when fat and carbohydrates are consumed simultaneously, the carbohydrates stimulate insulin secretion and thereby promote storage of fat (for recent review see ). Therefore, it is important to consider the macronutrient profile of the diet when examining the effects of dietary fat on biological processes.
In the present study, transgenic animals were fed ad libitum a very high fat (79%) diet that was practically devoid of carbohydrates (0.76%). The KD resulted in ketone body production, weight loss, and decreased Aβ levels. Hence, the data presented here suggests that it may not be fats in the diet that increases Aβ levels, but perhaps levels of total calories, carbohydrates, or the metabolic state of the animal.
Epidemiological studies in humans have implicated saturated fats in the diet as a risk factor for Alzheimer's disease. For example, Kalmijn et al. correlated eating habits with incidence of dementia after a two year follow-up in a large study of 5,386 subjects in Rotterdam, NL. The results from this analysis led the author to suggest that diets rich in saturated fats and cholesterol increased the risk of several types of dementia . However, after a 6 year follow-up of this same population, no correlation between dementia and fat intake could be identified, leading the authors to conclude "High intake of total, saturated, and trans fat and cholesterol and low intake of MUFA, PUFA, n-6 PUFA, and n-3 PUFA were not associated with increased risk of dementia or its subtypes." . More recent studies have also examined the link between dietary fat and cognitive decline. In a study of 2,560 participants ages 65 and older in the Chicago Health and Aging project, fat intake was measured by food questionnaire and correlated with cognitive testing examined after a 3 and 6 year follow- up. This large study found only weak trends between saturated fat and cholesterol intake and cognitive decline . Both the rodent and human studies highlight the complications of trying to link complex environmental factors, such as eating habits or macronutrient intake, with dementia and Alzheimer's disease. In particular, one complicating factor in the human studies is the normal consumption of large amounts of carbohydrates in modern diets.
The present study demonstrates that, contrary to expectations, transgenic mice fed ab libitum a very low carbohydrate/high saturated fat diet present lower levels of Aβ after only 43 days of dietary change. The KD group exhibited low levels of both Aβ40 and the more amyloidic Aβ42, suggesting that the KD diet did not change or increase the efficiency of cleavage sites within APP. Instead the data suggests the KD regime either reduced processing of APP or increased degradation of Aβ species. Most of the animals administered the ketogenic diet lost body weight as well as exhibited reduced Aβ levels. However, the reduced Aβ levels may not have been due to a general lowering of protein content. Total brain protein levels did not differ between the groups and Aβ levels did not correlate with weight loss. Interestingly, despite change in diet, weight loss, and Aβ levels, no change in cognitive performance was observed (Table 2). This observation agrees with the general finding that KD diets are not harmful to mice . Also, the finding that reduction in Aβ did not improve cognitive performance may be due to the modest lowering of levels under these conditions and longer treatment may be required.
The KD diet was developed to mimic a starvation response in animals without reducing calories to harmful levels . In this way a KD is similar to caloric restriction (CR) regimes that have been used in many species to alter aging and increase some forms of stress resistance. CR typically reduces calories 30–40% compared to ad libitum fed animals and has numerous positive effects on animal health . In the present study we did not attempt to restrict calories in any way and the animals had free access to the ketogenic chow at all times and intake was self limited. However, since the animals were reluctant at first to eat the KD chow and we observed weight loss in the KD group, we cannot rule out the possibility that the Aβ lowering effects were due to CR.
Yet, CR and KD may work through similar mechanisms. KD are well known to reduce insulin signaling and mimic starvation, thereby increasing fatty acid oxidation and promoting a catabolic state . Similarly, CR is well known to reduce serum insulin and IGF levels and much of the benefit of CR may derive from this reduction in insulin/IGF signaling (for review see ). For example, decreased insulin/IGF-like signaling inhibits protein synthesis and promotes protein degradation, which may lead to the clearing of degradation-sensitive proteins, such as amyloidic peptides.
Increasing evidence suggests a role for insulin/IGF-1 in regulating APP and modulating Aβ levels. Receptors for both insulin and IGF-1 are highly expressed in brain, especially in hippocampus and cortex, where they may influence learning and memory . Insulin signaling in the brain increases extracellular levels of Aβ by promoting secretion  and inhibiting degradation by insulin-degrading enzyme . This view has also gained recent support in humans. Fishel et al. demonstrated that induced hyperinsulinemia in healthy elderly subjects elevated both serum and spinal fluid Aβ levels, suggesting insulin plays a role in elevating Aβ, especially under conditions such as type II diabetes .
Such an interpretation is consistent with recent studies demonstrating similar Aβ lowering effects of a low carbohydrate, caloric restriction (CR) regime in mice expressing the "Swedish" form of APP (Tg2576 mice). In these animals, lower levels of Aβ 40 and 42 were detected in animals fed 30% less carbohydrates than ad libitum fed animals. The Aβ lowering effect may have been due to increased α-secretase and insulin degrading enzyme activity in the CR animals .
Alternatively, other physiologic changes may have reduced Aβ levels in this study. The high levels of ketone bodies alone may have contributed to increased protein turnover. Ketone bodies added to cell culture have been shown to lead to increased oxidation of proteins prone to oxidative damage. The presence of damaged protein triggers proteolysis of normally long lived proteins via chaperone-mediated autophagy . Such a mechanism may be at work in animals fed a ketogenic chow and exposed to high levels of ketone bodies. Some support for this model comes from the observation that elevated ketone body levels correlated better with lowering of Aβ species than did weight loss. In addition, ketone bodies may serve as an efficient substrate for neuronal metabolism. Previous studies have shown that acute elevation of ketone bodies may improve cognitive performance in some individuals with mild to moderate AD .