27 October 2009. It is a sign of progress in Alzheimer disease research that longitudinal observational studies are beginning to converge on when and how a person’s cognition shows the first subtle signs of trouble on the way to dementia (see Part 3 of this series). That’s dandy for research, but clinicians urgently need robust tools to diagnose presymptomatic dementia in one clinic visit. Part of the toolkit for that will come from biomarkers, and at the 7th Leonard Berg Symposium, held 1-2 October 2009 at Washington University, St. Louis, scientists shared some of the latest news in this burgeoning area of study. As throughout this conference, talks toggled between what’s known in LOAD and eFAD, comparing all the while how well knowledge on these forms of AD matches up.

John Morris of WashU started the topic with an update on his center’s Antecedent Biomarker Study, which has been seeking to find a predictive combination of biomarkers in cognitively normal adult children of a parent with AD since 2005. (These are not families with dominantly inherited AD.) Dispensing with cautious qualifiers, Morris summed up the bottom line of this work: “We can detect preclinical AD in cognitively normal older adults.” How long before dementia? About four years. And this is how it works, claimed Morris: When people have reduced CSF Aβ42, elevated CSF tau/phospho-tau, and amyloid in their brains, they will subsequently develop dementia. Their Aβ42 drops first, brain amyloid shows up soon after, and tau starts rising just prior to symptoms (for comparison, see Johnson diagram in Part 6). Almost all people whose CSF Aβ42 is abnormally low also have amyloid in their brains.

This website has covered many individual studies on the way to this conclusion (e.g., Skoog et al., 2003; Sunderland, 2003; Fagan et al., 2006; Fagan et al., 2007; Li et al., 2007; Shaw et al., 2009), as well as the broader literature showing that damage to the brain is extensive by the time a person is diagnosed. Hence, this story will focus on the latest data presented in St. Louis. For example, Morris reported that the amyloid in people’s brains is driven by the two leading risk factors for late-onset AD—age and ApoE. In a PIB PET series of 241 cognitively normal people age 45 to 89, brain amyloid started showing up around age 55 and became more and more frequent in older folks (Morris et al., 2009). People with ApoE4 were highly overrepresented relative to their allele frequency among the PIB-positive group, whereas almost no one with ApoE2 had brain amyloid even up into the highest ages. In the sixties, seventies, and eighties, the percentages of people with ApoE4 being positive for PIB were 37, 53, and 75, respectively; for people without ApoE4, these numbers were 8, 16, and 16 percent by comparison. “This suggests that expression of the E4 phenotype is very strongly associated with presence of cerebral Aβ deposits as people age,” Morris said. In essence, this makes brain amyloid a phenotype of sorts of ApoE4 (see also Reiman et al., 2009). And an upcoming paper in Archives of Neurology reports that, among a group of 159 cognitively normal research volunteers, having amyloid in the brain predicts that the person would develop AD symptoms when followed for up to 5.5 years (mean of 2.4 years; Morris et al., 2009).

Besides ApoE and age, scientists are currently looking to relate the two linked biomarkers of abnormal CSF/brain amyloid to additional known risk factors and markers of AD. The hope is that a comprehensive picture might emerge of how multiple parts of the biology fit together in the preclinical phase. One recent step in this effort was a paper showing that in cognitively normal people, the abnormal CSF signature is statistically linked with brain atrophy; at later stages, when dementia sets in, tau drives the CSF-atrophy relationship as it continues to rise while the brain continues to shrink with progressing illness (Fagan et al., 2009). More than just whole brain atrophy, brain amyloid in cognitively normal older adults is associated with thinning of the cortex in regions known to be vulnerable to AD pathology (Dickerson et al., 2009). This suggests that a preclinical CSF signature is beginning to match up with an imaging signature composed of amyloid and cortical thinning.

In his talk, David Holtzman of WashU noted that a new Dutch/MGH study reporting reduced CSF Aβ40 concentrations in cerebral amyloid angiopathy (CAA) provided yet another independent confirmation for the general idea that as amyloid deposits in the brain, in this case on blood vessels, it becomes trapped and Aβ concentrations in the CSF drop. CAA is a common cause of strokes (Verbeek et al., 2009).

An upcoming paper from the WashU group further tightens the connection between brain amyloid and CSF Aβ by analyzing CSF Aβ versus PIB and age in 189 cognitively normal people, Holtzman said (Fagan et al., in press). In this series, everyone whose brain binds PIB also has low CSF Aβ42, but the opposite is not true. Some people, especially the youngest participants between 45 and 55 years of age, already have low CSF Aβ42 but no PIB. The scientists interpret this to mean that brain amyloid deposition begins in a conformation that may initially be invisible to PIB. The subsequent drop in CSF Aβ42 would consequently be the earliest detectable biomarker at present. When the brain amyloid later becomes fibrillar, it binds PIB. “This is our impression so far, but we do not have proof yet,” said Holtzman. “We have to follow these cohorts longer to see in which order these markers come up.”

The same paper also contains more data connecting the CSF combination of high tau/low Aβ42 with brain amyloid, in essence predicting that this CSF signature reflects ongoing neurodegeneration and will predict onset of symptoms in the next three to five years. Building on a smaller previous study, this finding extends into cognitively normal people in a widely cited study from Kaj Blennow’s group three years ago, in which virtually everyone with this CSF signature among a large cohort of MCI patients converted to AD within five years (Hansson et al., 2006). In this study and an independent recent one, these biomarkers predicted not only whether people would develop AD, but also how fast their cognitive decline would progress (Snider et al., 2009).

Now, for the first time, a pharma company, Bristol-Myers Squibb, targets people at this pre-dementia stage for a drug trial. The new twist, compared to previous MCI trials (which all failed), is that not an MCI diagnosis but a low CSF Aβ42 level plus a subjective memory complaint determine whether a person can enter the trial. CSF Aβ/tau, and brain atrophy are the outcome measures listed just below safety. If the distinction between CSF Aβ42 and CSF tau changes holds up in larger studies, i.e., if the former truly precedes the latter by two years or so, then future trials could push back to treating asymptomatic people by screening for low CSF Aβ42 and enrolling people just at the point when their tau is beginning to nudge up but before they have symptoms. CSF tau could then conceivably become an outcome measure to see if the drug is effective. That, then, would constitute a prevention trial, and it may soon come within reach, said Reisa Sperling of Brigham and Women’s Hospital.

These data come from research with volunteers who have, or may develop, the common forms of AD. Establishing the order of antecedent biomarkers in dominantly inherited (aka autosomal-dominant or early-onset familial) AD with sufficient statistical power to support drug trials is part of what the Dominantly Inherited Alzheimer Network (DIAN) is aiming to accomplish. Small biomarker studies with individual families have already begun to pave the way. For example, researchers led by Dan Pollen of University of Massachusetts Medical Center in Worcester, who described the first reported presenilin 1 family, reported CSF Aβ42 decreases in six presymptomatic mutation carriers (Moonis et al., 2005). Last year, John Ringman of UCLA reported the same thing, plus that CSF tau was increased. In this study, CSF isoprostanes were up, too, as was plasma Aβ (Ringman et al., 2008).

At the Leonard Berg Symposium, Raquel Sanchez-Valle of the Hospital Clinic in Barcelona, Spain, reported new CSF and plasma data on 14 relatives from four different families with presenilin mutations. Of the eight participants who carried the AD mutation, half were symptomatic, half not yet. These Spanish investigators offer genetic counseling, testing, and observational research to families with genetic neurodegenerative diseases including eFAD (see eFAD studies). In St. Louis, Sanchez-Valle presented the first cross-sectional data of what is to become a longitudinal study of these volunteers. Using Innogenetics’ Innotest for CSF, and a cutoff value of 495 pg/ml (as per van der Vlies et al., 2009), CSF Aβ42 levels were normal, i.e., high, in those presymptomatic carriers who were still more than a decade away from their family’s mean age at onset, and low in even mildly symptomatic carriers, Sanchez-Valle reported. CSF tau was elevated only by the time carriers became clearly symptomatic, and it then correlated strongly with a person’s clinical dementia rating or MMSE. This Spanish group found no differences in amyloid plasma between carriers and non-carriers in this initial study.

“Overall, this indicates that the same kind of changes are occurring in dominantly inherited AD as in late-onset AD,” Holtzman said, but cautioned that these studies are all very small.

That said, speakers in St. Louis agreed that neither DIAN, nor the Alzheimer's Disease Neuroimaging Initiative (ADNI), nor the Adult Children Study and other cohorts should restrict their analyses to the usual suspects Aβ and tau. A multitude of other markers are coming out of proteomics analyses of CSF and plasma, and some bear close watching. For example, Eric Portelius, working with Henrik Zetterberg and Kaj Blennow at University of Gothenberg in Sweden, has developed combined immunoprecipitation/MALDI-TOF mass spec protocols to explore the proteomic diversity of Aβ and APP species in CSF of LOAD and FAD. The Swedish scientists have found some 20 different Aβ species in CSF; Aβ42 was one of the least abundant ones. It is important to disease because it is hydrophobic and aggregates readily, but as biomarkers, other species may be easier to use and more informative. Variability among centers, particularly in Aβ42 measurements, has been dogging the field for some time (Verwey et al., 2009; Mattsson et al., 2009). At ICAD in Vienna, Zetterberg included in his plenary lecture unpublished data suggesting that in the CSF of some familial AD cases, Aβ37, 38, and 39 were all particularly low, whereas an Aβ1-16 fragment was abnormally high. This pattern differed starkly between PS-mutant AD and sporadic AD, though both forms had similar, and expected, findings on Aβ40 and 42. Aβ1-16 popped out of that work as a novel biomarker candidate for both sporadic and familial AD. “DIAN should look at these other species, too,” Holtzman said.

Also at ICAD, Zetterberg presented his group’s detection in CSF of AD patients of a set of truncated Aβ forms, as well as a set of APP fragments, both of which point to the existence of a new, yet-to-be-defined cleavage sequence of APP (see also Portelius et al., 2009; Portelius et al., 2009; Portelius et al., 2009). Beyond APP and Aβ species, a wealth of potential markers are being discovered. To quote but one example from the Leonard Berg Symposium, the poster session featured a study by Rawan Tarawneh and colleagues at WashU on the neuronal injury marker VILIP-1, an intracellular calcium-sensor that tracked with dementia severity in a small study of nine AD patients and 15 controls.—Gabrielle Strobel.

This is Part 4 of a seven-part series on presymptomatic detection. See also Parts 1, 2, 3, 5, 6, and 7.