13 September 2008. What started as a trickle of data is turning into a rivulet, and may become a stream before long as more and more researchers are testing the idea that β-secretase enzyme activity in cerebrospinal fluid could be a sign of early Alzheimer disease. Recent work from Henrick Zetterberg and Kaj Blennow at Goteborg University in Molndal, Sweden, published in the August issue of the Archives of Neurology, indicates that CSF BACE activity is significantly increased in people with AD, and even more so in people with mild cognitive impairment (MCI) compared to an age-matched healthy control population. In the longitudinal study, the Swedish scientists show that the people who progressed to AD during a three- to six-year follow-up period had higher baseline BACE1 activity levels than those who remained stable or developed other forms of dementia. BACE1 levels are not diagnostic of AD; there is too much overlap between the groups for that. But with further study, the measure could turn out to be useful for tracking early stages of disease.
Their work follows on recent studies from Yong Shen at Sun Health Research Institute, Sun City, Arizona, and Harald Hampel at Ludwig Maximilian University in Munich, Germany, showing a similar pattern of increased BACE protein and activity in MCI and a subsequent decline later when AD sets in. There are some differences in the results between the studies, particularly on the question of whether BACE activity is up, down, or unchanged in advanced disease. Even so, the results so far will only serve to increase interest in this new potential biomarker.
BACE1 is a membrane-bound protease that catalyzes the initial and rate-limiting cleavage event of the amyloid precursor protein to form the β amyloid peptide. (It also has other substrates.) BACE1 activity increases in the brain during the course of AD, and its levels there correlate with amyloid load in both humans and mouse models. The first studies to look at BACE protein or activity in CSF of AD patients came from researchers in Australia who reported slight but significant elevations (Holsinger et al., 2004 and Holsinger et al., 2006). At the same time, a group from the Netherlands developed a highly sensitive enzyme assay, and reported the presence of a soluble, truncated form of BACE in the CSF, and the elevation of BACE activity and protein in a small number of AD patients (Verheijen et al., 2006).
The first data on CSF BACE in MCI came from a study by Shen, Blennow, and Hampel, who studied a group of 208 Swedish and German participants, which included 80 with AD, 59 with MCI, and 69 healthy controls. They found BACE1 protein and activity were both increased in patients with MCI, but patients with AD were no different from controls (Zhong et al., 2007). This spring, the same researchers published additional results on a subset of those patients (51 with MCI, 61 with AD, and 37 healthy controls), where they found that the presence of an ApoE4 risk allele in either AD or MCI subjects was associated with increased CSF BACE1 activity (Ewers et al., 2008). In this group, they also looked at CSF Aβ42 and got the expected result: levels were lower in AD than MCI, which was lower than healthy controls.
Most recently, Hampel presented data at the ICAD meeting in Chicago, showing the longitudinal follow-up of 47 MCI patients. In addition to CSF BACE1, the investigators measured CSF tau and p181tau, and determined ApoE genotype and baseline performance on neuropsychological tests. After an average follow-up time of 2.3 years, 15 MCI patients converted to AD, and regression analysis suggested that BACE1 protein levels and ApoE genotype were the strongest predictors of conversion.
The work corroborates that from Zetterberg and colleagues, who collaborated with researchers at Merck Research Laboratories in West Point, Pennsylvania, to determine CSF BACE1 activity using a new and highly sensitive assay (Wu et al., 2008). That assay, from Adam Simon and colleagues, uses a florescent ELISA readout to measure the appearance of a new epitope after cleavage of an optimized peptide substrate. The assay can detect BACE1 in the 30-50 pM range, which in the CSF seems to come from a C-terminally truncated soluble form of BACE1.
Zetterberg and coworkers used the Merck assay to determine CSF BACE1 activity in a group of people from Malmo, Sweden, comprising 87 patients with AD, 33 healthy controls, and 113 with MCI who were followed for four to six years. In contrast to the results of Shen and colleagues, Zetterberg reports a “slight but significant” elevation of CSF BACE1 in AD patients. But more importantly, they found that the 45 people with MCI who progressed to AD during the follow-up had significantly higher baseline BACE1 (35 pM) than those with MCI who did not progress (29 pM), or who progressed to another form of dementia (20 pM).
Looking at other biomarkers, the Zetterberg group found no correlation between BACE1 levels and Aβ42, but there was an association with levels of soluble APP fragments sAPPα and sAPPβ. In all the groups, BACE1 strongly correlated with total tau levels. The results extend the findings of Zhong et al., the authors conclude, “by showing that elevated BACE1 activity in MCI is confined to subjects with incipient AD and that it is particularly pronounced in subjects with AD having high CSF levels of total tau.”
One point of contention among the different studies is what happens to the concentration of BACE1 in AD. The increase documented by Zetterberg and colleagues stands in contrast to the results of Shen and Hampel and to data presented by the Merck researchers in Wu et al., 2008. There, the researchers assayed samples from 27 pathologically confirmed cases of AD and 29 controls, and the results indicated that BACE activity increased in an age-dependent way. However, when adjusted for age, the AD patients showed a 56 percent decrease in CSF BACE1 activity. In another study presented in abstract form at ICAD, Jonathan Williams and coworkers from Oxford University, United Kingdom, also report decreased BACE1 activity in a small sample of AD patients (Williams et al., 2008). Among other differences, Zetterberg finds no correlation of BACE1 activity with age, and unlike Hampel, he detects no influence of ApoE status. These differences will need to be sorted out with larger studies in other patient groups.
The nature and source of BACE1 activity in the CSF also remain somewhat nebulous to date. Some studies report finding a C-terminal truncated form of BACE1 in CSF, consistent with the proteolytic cleavage of the transmembrane domain and shedding of the catalytic domain. Alternatively, BACE1 could be released during neurodegeneration. In either case, the increase in CSF BACE1 could reflect a higher brain load of the enzyme early in disease, said Robert Vassar of Northwestern University, Chicago, Illinois. “PIB imaging shows a significant amount of amyloid in people with MCI, so that plaque burden is high even before dementia. We and others have seen that BACE levels are increased in AD brain and in transgenic mice with amyloid plaques. We know that BACE is increased around plaque, and it is increased in neurons and in structures that look like dystrophic neurites,” he said. “Maybe during neurodegeneration, these neurons with high levels of BACE are dying, and some gets out. Alternatively, maybe as levels of BACE1 increase in AD, a certain amount is processed, released, and secreted. It’s not clear what the increase in BACE in CSF might mean for the progression of disease, but it sounds like a reasonable biomarker for early AD or MCI.”
Why would levels of BACE1 peak in MCI, and then go down in AD? Vassar speculates that it could be related to neuron loss. “Maybe as you lose neurons, BACE declines. AD has an early phase with no neuron loss but synaptic loss and dystrophic neurites. Maybe that is the phase where BACE is increasing, and later it declines with neuron loss. That’s just a guess, but it’s a hypothesis that might be testable in animal models,” he said.
On the other hand, Blennow has previously proposed that CSF tau is a marker for axonal degeneration (Blennow, 2004). If BACE1 shedding occurs during neuronal loss, then that might explain the correlation Zetterberg and colleagues observe between CSF levels of the two proteins.
BACE1 levels should be intimately related to Aβ42 levels, yet the kinetics of change over disease progression appear quite different. The preliminary data show BACE1 peaking in MCI, while Aβ declines during the progression from healthy aging to MCI to AD. Low CSF Aβ42 correlates well with amyloid load, even before dementia appears. “Looking at correlations between BACE and Aβ42 may not be meaningful in the presence of plaques,” says Anne Fagan of Washington University in St. Louis. “I predict the plaques acting as a sink would trump any effect of BACE levels on Aβ production,” she told ARF in an e-mail.
“These are early days for BACE1. Further studies are needed to know its diagnostic utility,” said John Trojanowski of the University of Pennsylvania in Philadelphia. Trojanowski, who was not involved in any of the CSF BACE1 work to date, but is co-leader with Les Shaw of the biomarker core for the Alzheimer Disease Neuroimaging Initiative (ADNI). That study has accumulated CSF samples from more than half of ADNI study participants at baseline and continues to collect CSF from these participants at year 1, 2, and 3. “CSF BACE1 looks promising enough for us to include it in the proposal for our competing renewal of the ADNI study next year,” Trojanowski told ARF in an e-mail.—Pat McCaffrey.
Zetterberg H, Andreasson U, Hansson O, Wu G, Sankaranarayanan S, Andersson ME, Buchhave P, Londos E, Umek RM, Minthon L, Simon AJ, Blennow K. Elevated cerebrospinal fluid BACE1 activity in incipient Alzheimer disease. Arch Neurol. 2008 Aug;65(8):1102-7. Abstract
Author Q&A with Harald Hampel. Questions by Pat McCaffrey.
Q: In your published results, you see a definite increase in CSF BACE1 protein and activity in MCI, but not in AD. That is, the AD patients are the same as healthy controls?
A: This was our initial finding in our first wave of patients published in the Archives of General
Psychiatry (Zhong et al., 2007) and Brain (Ewers et al., 2008). We have now assessed CSF BACE activity in a second, independent sample and found BACE1
activity also increased in AD when compared to healthy controls.
Although the reason for the different findings is not clear, we note
that the AD patients in the current study were less strongly impaired (MMSE = 24) than those in the first sample (MMSE = 20). Taken together with the findings of dramatically increased BACE in MCI, this may mean that BACE1 is increased especially in early stages of AD. But this hypothesis needs to be tested in further
Q: In your ICAD talk last July, you showed data on hippocampal atrophy
predicted by CSF BACE levels. Can you clarify that experiment?
When was BACE measured?
A: MRI scans were obtained in the sample of mild AD patients mentioned above. We
found that increased BACE1 activity was correlated with decreased hippocampal
volume. Hippocampal volume was measured at the time of CSF measurement so
the correlation is cross-sectional. We do not have sufficient MRI follow-ups
currently to look at rates of atrophy. The current findings suggest that BACE1
is a good predictor of the actual state of focal brain atrophy in AD. It will
need to be examined in a follow-up study whether acute BACE1 levels predict
progression of atrophy. Such a longitudinal study, along with additional investigations of DTI-assessed white matter changes and functional MRI assessed alterations of brain activation in association with BACE1 activity, is underway.
Q: What are the important next steps you see for this work?
A: To test at the multicenter level the validity of BACE1 as
a marker of early stages of AD. According to the recent recommendations for the
revision of NINCDS-ADRDA criteria (Dubois et al., 2007) we will
need to investigate whether BACE has sufficient accuracy for prediction and
differential diagnosis of AD in order to add to the early diagnosis of AD. We
are interested in whether BACE1 may pick up particular aspects of the AD
pathology which are not reflected by already well-investigated marker
candidates such as p-tau and Aβ in CSF (reviewed in Blennow and Hampel, 2003), and may thus provide additional
predictive power for AD. Together with our collaborators Yong Shen, Sun Health Research Institute,
and Michael Ewers from my team, I am currently working on approaches to address these challenges.
Q: How does this relate to BACE as a drug target?
A: Since BACE1 inhibition is now being tested as a potential treatment of
AD, it is desirable to assess in longitudinal studies whether BACE assessed in
CSF is a sensitive marker for disease progression. As a marker for a specific mechanism of action (MoA9 BACE1 may reflect the initial steps in the amyloidogenic processing of Aβ and therefore may account for total Aβ production), BACE1 may become an
interesting marker in clinical trials (reviewed in Hampel et al., 2008). It could serve both for Phase 2 enrichment of AD, MCI, or preclinical trials by selecting
those patients with pathologically increased amyloidogenic processing and thus potentially high likelihood of AD, as well as an outcome
parameter to assess disease-modifying treatment effects in Phase 3 confirmatory trials.
In addition to the current in-vivo assessment of BACE in CSF, we are also
pursuing BACE1 measurement in postmortem brain and CSF to understand
better to what extent BACE1 activity in CSF reflects BACE1 activity in the