Unlike APP and presenilin, BACE1 mutations have never been found to cause familial forms of AD; however, the protease has nonetheless assumed a central place in AD research. The field has widely accepted the finding made by different groups over the past six years that BACE1 activity and protein levels (but not its mRNA) are increased in the brains of people with late-onset AD (e.g., Holsinger et al., 2002; Fukumoto et al., 2002; Tyler et al., 2002; Yang et al, 2003). This finding has raised the question of what regulates BACE1 translation. Vassar approached the issue starting from the broader premise that both aging and Aβ42 contribute to AD pathogenesis early on. The lab began a research program to see if age-related stressors in a person’s physiology might integrate with specific AD processes through BACE1. Broadly speaking, BACE1 levels go up in response to various stressors—published work exists on apoptosis, hypoxia, ischemia, and oxidative stress, as well as traumatic brain injury.

At Keystone, Vassar focused on energy production, hypothesizing that age-related energy impairments induce a stress response that drives up BACE1 in the brain. Brain imaging has shown that glucose use is down in AD, MCI, and even asymptomatic middle-aged people whose ApoE4 genotype puts them at risk of AD. And work going back to 1994 showed that blocking energy metabolism boosted the amyloidogenic processing of APP in vitro (Gabuzda et al. 1994). To model where in all this there could be a connection to BACE1, Rod Velliquette in Vassar’s lab blocked energy production in Tg2576 APP-transgenic mice with injections of various energy production inhibitors. Published work indicates that single injections of either insulin (which causes hypoglycemia), 2-deoxyglucose (which mimics hypoglycemia), or 3-nitropropionic acid (which reduces ATP production), all increased BACE1 levels and Aβ levels (Velliquette et al., 2005).

To get at how this might work, Tracy O’Connor in the lab studied energy inhibition in a cell line overexpressing BACE1 and in cultured primary cortical neurons from Tg2576 mice. This work ruled out the possibility that BACE1’s half-life grew longer under conditions of energy starvation. But it did point to a change in eIF2α. This translation factor normally helps initiate global translation. Under conditions of stress it becomes increasingly phosphorylated, arrests global translation, and switches to supporting selective translation of certain stress-response proteins instead. The BACE1 mRNA 5’ untranslated region contains signature features of mRNAs that are translated with increased efficiency following physiological stresses, an effect induced by phosphorylated eIF2α. Experiments driving up eIF2α phosphorylation pharmacologically with an inhibitor of its phosphatase increased BACE1 protein levels, whereas experiments driving down eIF2α phosphorylation genetically prevented the stress-induced BACE1 increase. “We think we have nailed the mechanism. We can modulate eIF2α phosphorylation up and down and get a corresponding change in BACE1,” Vassar told the audience.

With those data, O’Connor went back in vivo and measured whether eIF2α was more phosphorylated in Tg2576 mice that had undergone chronic energy inhibition. Three months of weekly injections of either 2-deoxyglucose or 3-nitropropionic acid likewise increased both eIF2α phosphorylation and BACE1 protein levels but, again, not BACE1 mRNA. Aβ levels and amyloid plaque deposition went up in parallel in these energy-deprived mice. And a series of postmortem samples of AD brain from Rush University Medical Center confirmed not only the previously reported BACE1 increase but also a corresponding increase in eIF2α-P in AD patients.

When the scientists looked in a particularly aggressive transgenic mouse model of autosomal-dominant AD (Oakley et al., 2006), they noted an eIF2α-P increase even without depressing the mice’s energy production, Vassar reported. This suggested to him that amyloid itself can drive up eIF2α phosphorylation and subsequent selective BACE1 translation as part of a positive feedback loop, in line with a localized BACE1 increase near plaques the lab had published last spring (Zhao et al., 2007).

The data available to date comes down to a working hypothesis by which different kinds of age-related stress may all funnel into impaired brain energy metabolism, Vassar said. The stressors could be age, high cholesterol, cardiovascular disease, traumatic brain injury, ApoE4. Each of these could cause physiological stress, which leads to eIF2α phosphorylation and a subsequent BACE1 increase. This is initially a protective response to short-term stress, but if the chain of events becomes chronic, then the increased Aβ production will lead to amyloid accumulation. If amyloid degradation and clearance cannot keep up, amyloid itself can feed back to fuel eIF2α phosphorylation, more amyloid, and then downstream tau pathology and sporadic AD.

This presentation generated numerous questions, including why some people get AD without having had energy deprivation (e.g., a stroke) while others who sustain trauma do not go on to develop AD. In response, Vassar said that sporadic AD is a syndrome and this pathway may only operate in a fraction of patients. On the other hand, age-related chronic physiological stress is likely to be low level, and therefore difficult to detect, but could increase amyloidogenesis over a long period of time. There is at present no epidemiological data on energy inhibition in a subset of AD samples. Another questioner noted that when eIF2α switches from global translation to selective translation, other proteins may get selectively upregulated and other disease states may follow. A third area that needs follow-up research is whether vascular disease, which constricts the brain’s blood supply and often comes with amyloid deposits in blood vessels, might act in part through this change in translational regulation, as well (Cole and Vassar, 2008).—Gabrielle Strobel.