Much hope rides on the diagnostic and prognostic ability of biomarkers, exemplified by three papers in the March Archives of Neurology. But such measures may also have other uses. Alison Goate and colleagues at Washington University School of Medicine, St. Louis, Missouri, used CSF amyloid-β (Aβ) as an endophenotype—a measurable intermediate between genotype and overt phenotype—to look for genes that might be associated with late-onset Alzheimer disease (LOAD). In one volunteer, the researchers found a presenilin-1 single nucleotide polymorphism previously identified as an early onset, familial AD mutation. The finding suggests that some cases of apparently late-onset, sporadic AD may actually be inherited in Mendelian fashion. The data is reported in yesterday’s Annals of Neurology online.

It is not clear how many other LOAD cases may be inherited in this manner. “We may find that there has been somewhat of an ascertainment bias, that we were all going out looking for early onset AD families and sequencing their DNA and probably not paying as much attention to the families that had a late age of onset,” suggested Goate in an interview with ARF. But she expects the number of LOAD cases inherited in this fashion to be rare. However, going forward, she is revamping her ideas on AD genetics. “I always really thought of age 60 as a cutoff. If the mean age of onset was above 60, we generally didn’t bother to look for PS1 mutations in families. I’ve changed my view now, and I’ve decided that what is more important is evidence for Mendelian disease. If I find there’s a three-generation history, as in this family, then I’m going to sequence the known genes regardless of the age of onset,” she said.

The discovery came about because Goate and colleagues were interested in using quantitative endophenotypes to search for LOAD risk genes. “We thought if this [strategy] is going to work, then we might expect to find some rare mutations in the known genes amongst people in the extremes of Aβ values,” Goate told ARF. And that’s just what they found. First author John Kauwe and colleagues measured Aβ in cerebrospinal fluid drawn from 191 volunteers. After adjusting the raw data for ApoE status, age, gender, and clinical dementia rating, the researchers identified individuals with extreme (top and bottom 5 percent) CSF Aβ40, Aβ42, and Aβ42/Aβ40, and sequenced their PS1 genes. In one of those individuals, a member of a large family previously classified as having late-onset AD (LOAD), Kauwe found the previously identified missense mutation, which occurs in exon 4 and results in the replacement of an alanine to valine at amino acid 79.

“Though this study serves as proof of concept, what we would really like to do is collect a lot more samples so we can actually do a whole genome association study using Aβ levels as a quantitative trait, to see if we can find the genes that influence the levels of CSF Aβ,” said Goate. She noted that there has been only limited success in finding late-onset AD genes by analyzing sibling pairs in families, and this is an alternative approach that is definitely worth exploring. Endophenotypes have been widely used in the genetic studies of other diseases, for example, glucose levels in diabetes, and blood pressure in hypertension. The AD field has been slower in moving into this area, in part because there haven’t been clear phenotypes that might be useful, Goate explained. “These continuous traits give us a lot more power than simply asking if a person is affected or unaffected by the disease,” she added. Goate says they will be looking at other proteins in the CSF samples in addition to Aβ, including tau and phospho-tau.

Working with CNS samples has several advantages, including the range of proteins than can be measured and the possibility of reanalyzing stored samples. Those advantages were evident in the three Archives of Neurology papers. In one, Anne Fagan (no relation to the author of this piece) and colleagues, also at Washington University (Fagan was a coauthor with Goate on the Annals of Neurology paper), report that the CSF tau/Aβ42 ratio is a predictor of cognitive decline in non-demented older adults. The results suggest that this ratio could be a promising antecedent biomarker for AD. ARF previously reported on this paper when it was published earlier this year online (see ARF related news story).

An antecedent biomarker is somewhat of a Holy Grail for AD clinicians because disease pathology can begin decades before symptoms are recognized. In fact, Fagan and colleagues found that CSF Aβ levels are no different in people with very mild dementia (clinical dementia rating (CDR) score 0.5) than in those with more advanced disease (CDR ≥ 1.0). Instead, it was the tau/Aβ ratio that had more predictive power.

In the second paper, Neill Graff-Radford and colleagues at the Mayo Clinics in Jacksonville, Florida, and Rochester, Minnesota, reported a similar relationship for plasma Aβ42/Aβ40 ratios. They followed 563 cognitively normal, white volunteers for up to 10 years. During that time 53 subjects developed mild cognitive impairment or AD. Those individuals with Aβ42/Aβ40 ratios in the lowest quartile were at highest risk. These volunteers were 3.1 times more likely to suffer cognitive decline, even after correction for age and ApoE status.

The authors suggest that the Aβ42/Aβ40 level is a state variable that simply reflects that extent of Aβ deposition in the brain. Many now believe that as Aβ is deposited in the brain, its presence in the CSF, and subsequently the plasma, fall. However, some reports suggest an association between increased plasma Aβ42 and AD (see Mayeux et al., 2003).

Last but not least, researchers in Denmark and Sweden demonstrate one of the advantages of working with CSF samples, the ability to analyze multiple variables simultaneously. Anja Simonsen, from Ciphergen Biosystems Inc., Copenhagen, and colleagues there and at Danish and Swedish universities, identified a novel panel of CSF biomarkers that appears to predict progression from MCI to AD.

Analyzing CSF fluid from 113 MCI patients, 57 of whom progressed to AD during a 4- to 6-year follow-up, the researchers identified 17 different proteins that are differentially altered in CSF of the converters. Of these, 13 were elevated and four lowered. Five of the proteins were identified: phosphorylated osteopontin C-terminal fragment, ubiquitin, β2-microglobulin, and two components of the complement system-C4a des-Arg and C3a des-Arg.

This panel of biomarkers “may shed light into the pathophysiological process of MCI and have the potential to identify the patients with MCI who will progress to AD,” write the authors. However, they caution that the study needs to be confirmed in larger cohorts.—Tom Fagan


  1. This study adds further to our ability to exploit chemical and imaging biomarkers for the diagnosis of early AD, while emphasizing the need to identify other biomarkers that may predict who will progress to develop Abeta and tau pathologies prio to the onset of these AD pathologies.

    View all comments by John Trojanowski
  2. Aggregation and deposition of amyloid-β (Aβ) in the brain is thought to be the central event in the pathogenesis of Alzheimer disease (AD). Altered cerebrospinal fluid (CSF) levels of Aβ1-40 and Aβ1-42 and the Aβ42/40 ratio are used as biomarkers to support the clinical diagnosis in AD. Several studies have reported on decreased CSF levels of Aβ1-42, and various studies suggest that Aβ levels might be increased in plasma, including in familial AD forms.

    Despite many studies, many factors, such as genetic polymorphisms, which might be associated with disease or even alter the disease phenotype, are not well understood. These studies are complicated by distinct clinical phenotypes of AD, and even more by the fact that frequently, the diagnosis is done on clinical grounds. Therefore, in complex diseases with heterogeneous phenotypes, new approaches are urgently needed. One way to overcome the problem of clinical heterogeneity is to use intermediate traits, or endophenotypes, which are less heterogeneous than clinical diagnoses and might be more directly affected by genetic variations. In this study, Kauwe et al. analyzed if extreme Aβ levels might be used as an endophenotype for AD.

    The study comprised 191 volunteers participating in a study of aging and dementia. Sixty-four percent of them have a positive family history of AD (one or more first-degree relatives). One hundred forty-three were not demented, 33 percent had a very mild dementia (Clinical Dementia Rating 0.5), and 15 had mild dementia (CDR 1.0). They were studied for Aβ levels in CSF and blood, ApoE genotype, and various clinical characteristics. The levels of Aβ1-40 and Aβ1-42 differed by 10-fold between maximum and minimum in subjects studied. Aβ1-42 levels in CSF significantly decreased with age and increased with CDR scale. Aβ1-40 correlated with sex, but not with age, dementia severity, or ApoE status.

    In contrast to common studies on biomarkers, the authors followed a completely new approach. They hypothesized that extreme Aβ1-40 or Aβ1-42 levels might indicate some specific disease phenotypes and therefore can be used as endophenotypes to characterize AD. To study this, individuals with extreme CSF values of Aβ1-40, Aβ1-42, and Aβ42/40 ratio (top and bottom 5 percent values) were identified for DNA sequencing. Eight genetic variants of PSEN1 were identified in the sample. In one individual, a missense mutation in exon 4, A79V, was identified. This is a known mutation in familial Alzheimer disease that has been previously reported in four families. The carrier is nondemented and has the fifth highest adjusted Aβ1-42 value and the third highest Aβ42/40 ratio in the sample studied. To investigate the consequences of this mutation on Aβ levels, the A79V mutation was introduced into the wild-type PSEN-1 sequence and then transfected into PSEN1/2 knockout mouse embryonic fibroblasts. The Aβ levels and the Aβ42/40 ratio in conditioned media from these cells were significantly higher than the wild-type PSEN-1 sequence, thus suggesting a direct effect of the mutation on Aβ levels.

    The results of the study, which follows a completely novel approach to biomarkers, suggests that CSF or potentially blood levels of certain proteins might be a useful endophenotype for genetic studies in AD and might lead to the identification of atypical disease phenotypes or genetic variations. However, the study opens new questions which are related to the diagnosis of AD, which is most always done on clinical grounds. Since the “typical” CSF pattern seen in AD is a decrease of Aβ1-42, patients with abnormally high levels in CSF, which might also display an atypical clinical presentation, might be missed. As a consequence, a careful follow-up of all suspected cases is needed to solve the question of atypical presentations in AD and potentially to detect novel mutations that might cause AD.

Make a Comment

To make a comment you must login or register.


News Citations

  1. Biomarker Roundup: Collecting Clues from MRIs to RNAs

Paper Citations

  1. . Plasma A[beta]40 and A[beta]42 and Alzheimer's disease: relation to age, mortality, and risk. Neurology. 2003 Nov 11;61(9):1185-90. PubMed.

Further Reading

Primary Papers

  1. . Extreme cerebrospinal fluid amyloid beta levels identify family with late-onset Alzheimer's disease presenilin 1 mutation. Ann Neurol. 2007 May;61(5):446-53. PubMed.
  2. . Cerebrospinal fluid tau/beta-amyloid(42) ratio as a prediction of cognitive decline in nondemented older adults. Arch Neurol. 2007 Mar;64(3):343-9. Epub 2007 Jan 8 PubMed.
  3. . Association of low plasma Abeta42/Abeta40 ratios with increased imminent risk for mild cognitive impairment and Alzheimer disease. Arch Neurol. 2007 Mar;64(3):354-62. PubMed.
  4. . Novel panel of cerebrospinal fluid biomarkers for the prediction of progression to Alzheimer dementia in patients with mild cognitive impairment. Arch Neurol. 2007 Mar;64(3):366-70. PubMed.