Posted 15 July 2004

Co-Chairs: David Holtzman, MD, Washington University, St. Louis, MO
Richard Mayeux, MD, Columbia University, New York, NY

Participants:
Neil Buckholtz, PhD, National Institute on Aging, Bethesda, MD;
Michael Conneally, PhD, National Cell Repository for Alzheimer's Disease, Indianapolis, IN;
Matthew Frosch, MD, Harvard Medical School, Boston, MA;
Fran Grodstein, ScD, Harvard Medical School, Boston, MA;
Bradley Hyman, MD, Massachusetts General Hospital, Boston, MA;
Michael Irizarry, MD, Massachusetts General Hospital, Boston, MA;
Adrian Ivinson, Harvard Center for Neurodegeneration and Repair, Boston, MA;
John Morris, MD, Washington University, St. Louis, MO;
Eve Nichols, Fidelity Foundations, Boston, MA;
Elaine Peskind, MD, University of Washington, Seattle, WA;
Leslie Shaw, PhD, University of Pennsylvania Medical Center, Philadelphia, PA;
Meir Stampfer, MD, DrPH, Harvard School of Public Health, Boston, MA;
Jeff Vance, PhD, MD, Center for Human Genetics, Duke University, Durham, NC;
Mark Watson, MD, PhD, Washington University, St. Louis, MO;
Philip Wolf, MD, Boston University, Boston, MA

Introduction
This report summarizes the discussions and recommendations of the second Workshop on Antecedent Biomarkers for Alzheimer's Disease. The first Workshop, chaired by John Morris, MD, of Washington University, was held on May 21, 2003, at the Danielle Hotel in St. Louis, see report. The goal of both workshops was to establish a foundation for developing, evaluating, and refining candidate biomarkers for Alzheimer's disease (AD) in its presymptomatic (preclinical) stages.

Summary of Workshop I
In reviewing the nascent field of antecedent biomarkers for AD, a variety of candidate biomarkers have been proposed. Some are further in development than others, but to date none have been validated as a biomarker for AD. Assays of body fluids such as plasma and cerebrospinal fluid (CSF) evaluate specific proteins that undergo abnormal conformation, which enhances their propensity to aggregate and deposit in brains of individuals with AD. These proteins include the peptide Aβ ?consistently decreased in AD patients compared with controls. Higher plasma levels of Aβ42 in normal individuals have been reported in one study to confer greater risk for AD five years later. These proteins also include tau and its phosphorylated state p-tau, which are elevated in AD. Significant differences have been found in Aβ42, tau, and p-tau in CSF when comparing groups of AD patients with controls, but whether changes in these markers will be useful antecedent biomarkers has not been explored. Up to now, static levels of Aβ and tau have been studied, but future investigations should account for the central nervous system turnover of these proteins.

Other body fluid components that are altered in AD include isoprostanes, a putative marker of lipid peroxidation in the brain. Levels of isoprostanes may reflect white matter damage and have been reported to be elevated in the CSF of individuals with very mild AD. They also include sulfatides, a prominent component of cerebral white matter, which have been reported to be decreased in individuals with very mild AD. These lipids have not yet been explored as antecedent biomarkers.

Neuroimaging markers also were reviewed, including structural changes (e.g., regional and whole brain volumetric loss), functional changes (e.g., hypometabolism), and radiolabeled tracers that bind to Aβ deposits with high affinity and can be detected by positron emission tomography (PET). Data suggest that imaging changes may precede cognitive deficits; however, as with fluid biomarkers, imaging modalities have not yet been systematically explored for their use as antecedent biomarkers.

Also reviewed were the potential application of genomics (gene chip microassays/cluster analysis of mRNA transcript abundance in tissue) to detect predictive gene patterns or signatures in individuals with AD and proteomics (recognition of complex protein patterns associated with AD, for example by unbiased comparison of protein levels between different clinical samples using mass spectrometry). Both genomics and proteomics have been shown to be highly sensitive and specific for certain cancers, where they are useful for prognosis.

It was recommended that these multiple candidate biomarkers be simultaneously investigated, as a battery of markers may prove superior to any single marker alone.

To proceed with the exploration of putative antecedent biomarkers, it was recommended that four key issues be addressed:

1. Achieve consensus on the clinical distinctions between non-demented aging and dementia of the Alzheimer type (including the earliest symptomatic stages of AD) to permit meaningful group classifications;

2. Support efforts to explore candidate biomarkers for their predictive value and their limitations; these efforts should include (but not necessarily be limited to) genomics, proteomics, fluid analysis, and neuroimaging;

3. Standardize methods for sample collection, storage, and analysis to allow results to be compared across sites; and

4. Learn from other fields (e.g., cancer; cardiovascular disease; Huntington disease) involved in antecedent biomarker studies.

The Workshop concluded with specific suggestions: 1) to collaborate with clinical investigators, database managers, and biostatisticians to standardize the information to be collected across institutions; 2) design longitudinal studies to evaluate candidate biomarkers and to address potential confounds (e.g., medications; lifestyle factors; renal function); 3) add younger cases to the clinical samples, as antecedent AD changes may occur years or decades prior to the symptomatic expression of the disease; and 4) plan to store samples for future analysis as new candidate biomarkers or investigative techniques become available. In addition, there was strong support to develop plans to standardize data acquisition and methods for sample collection, storage, and analysis. This latter recommendation prompted the convening of Workshop II and formed the basis of its agenda.

Workshop II
This Workshop was co-sponsored by the Alzheimer's Disease Research Center at Washington University, the Taub Institute at Columbia University, and the Harvard Center for Neurodegeneration and Repair. Drs. Holtzman and Mayeux organized the Workshop and served as co-chairs.

1. State of the Field
The clinical diagnosis of AD is limited because it relies on the presence of symptoms (i.e., dementia), whereas the pathology may begin years before AD symptoms are expressed. There thus is strong impetus to develop antecedent biomarkers for AD to detect its preclinical stages. Either a disease risk factor or a disease surrogate can be developed as a biomarker. Exposure to a risk factor and its dose determine its biological effect in producing disease; surrogate markers reflect altered structure or function that reflects the clinico-pathologic state. Markers need to be evaluated for their relationship to the disease (validity). It cannot always be inferred that markers themselves are causal, as they may relate to an unknown confounding factor that is more directly involved in the disease process. Developing biomarkers involves field methods, evaluation of dose response and possible modifiers, determinations of sensitivity and specificity, and assessment of variation within populations.

Advantages of a true biomarker include objectivity, precision, reliability, validity, relationship to disease mechanisms, and less bias than clinical diagnosis. Disadvantages include the potential for fixed "windows" (a limited period of time during which a biomarker operates within the disease process), expense, determination of "normal" range, storage of samples, laboratory errors, and ethical responsibilities (e.g., how should the presence of a biomarker for a lethal disease for which no treatment exists be communicated to an asymptomatic individual?).

2. The Neuroimaging Initiative
The National Institute on Aging (NIA) plans to award a grant (funding to begin in September 2004) to a consortium of AD programs that will enable structural neuroimaging measures to be obtained serially for up to three years in non-demented elderly and individuals with mild cognitive impairment and early-stage AD. Samples of blood, urine, and CSF also will be obtained and banked for future use by investigators. In addition to the NIA, the Initiative is supported by the Alzheimer's Association, the pharmaceutical industry, foundations, and the Food and Drug Administration.

Other AD programs with tissue banks include the Framingham Study, the Honolulu-Asia Aging Study, the Religious Orders Study, the Nun Study, the Guam Study, the Northern Manhattan Study, and a clinical trial of vitamin E in adult Down syndrome.

3. The Framingham Study
The original study began in 1948 to study cardiovascular risk factors; the Offspring Study (5,124 individuals enrolled in 1972 and followed to 2008) and the Third Generation Study (3,500 grandchildren of original participants enrolled in 2002) are newer cohorts. Any participant agreeing to brain donation is examined annually. A total of 6,500 cell lines have been established. It is possible to examine specimens collected 10-30 years ago for antecedent biomarkers for AD. Other cardiovascular studies include the Cardiovascular Health Study and the Atherosclerosis Risk in Communities.

4. What to Collect and How
a. Recommended specimens to collect include: CSF, plasma, serum, and DNA; urine and buffy coat are optional (can obtain DNA from buffy coat but labor intensive).

b. Collection methods must be standardized to allow comparisons across sites and studies. Recommendations:

1. Fasting state (allow water, medications)

2. Sampling time: 7-10 a.m.

d. Whenever possible, sample collection should be paired (e.g., blood and CSF).

e. Samples should be divided and stored among freezers and locations. Freezers should be -70°C or -80°C with backup power. Liquid nitrogen storage is preferred (less freezer failure but more costly).

f. Samples should be bar coded; maps should be made of the sample boxes. Note was made of potential risk in de-identified samples of two samples collected at different times from the same subject being mistakenly considered as from two different subjects.

g. Quality control measures are essential.

h. CSF

1. Lumbar puncture (LP): Use adequate local anesthesia and atraumatic Sprotte needle (24 gauge) with 20 gauge introducer to minimize risk of post-LP headache. Using the 24 gauge needle, the CSF often needs to be drawn with a sterile syringe. Alternatively, a 22 gauge Sprotte needle can be used; although there is a slightly greater risk of post-LP headache with this gauge needle, a syringe generally is not required to remove the CSF. Note the subject's position (i.e., lateral decubitus or sitting). Post-LP precautions include lying down for 1 hour after the procedure, avoiding exertion for 24 hours, and good hydration and perhaps modestly increasing salt intake.

2. Recommend collection of up to 30ml of CSF (minimum: 8-10ml); collect in 5ml volumes at a time, recording the batch number on the storage tube. CSF analytes may or may not have a rostral-caudal gradient (i.e., concentration may increase as CSF is removed, reflecting greater brain contribution). Send first 2ml (unfrozen) to lab for routine studies (cell count, glucose, protein).

3. Methods to address possible blood contamination of CSF samples remain controversial. There are two options.

• a. Option #1: place samples on wet ice (4° C). Cold centrifuge for 10 minutes at 3000 rpm; aliquot CSF into storage tubes without disturbing rbc pellet. Freeze and store at -70° to -80° C. This method removes rbcs but samples will remain contaminated by plasma proteins.

• b. Option #2: Aliquot and freeze samples immediately on dry ice. As described below, samples with lab cell count indicating >10 rbcs will not be usable for proteomics studies.

4. Exclude samples with >500 rbcs or >10 wbcs. Samples with <10 rbcs can be used for proteomics.

5. Divide samples into 0.25ml or 0.5ml aliquots (maximal yield of never-thawed samples versus available freezer space). Place samples immediately on dry ice for transport to freezer; store at -70° or -80° C. Thawing and refreezing CSF may change biomarker concentrations.

1. Use EDTA tubes for plasma; place tubes on wet ice (-4°C). Cold centrifuge for 10 minutes at 3,000 rpm. Aliquot immediately after centrifuging and freeze (-70°C) within 1 hour of collection.

2. Use plain glass or serum separator tubes for serum. Allow blood to clot at room temperature, then centrifuge as above. Aliquot and freeze samples within 1 hour of collection. The pellet (wbcs) can be recovered as the buffy coat.

3. Do not add antioxidants or protease inhibitors to all aliquots.

4. Standard serum measures (e.g., glucose, creatinine) should be obtained in all subjects.

k. DNA (both nuclear and mitochondrial) from blood (wbcs) in EDTA tubes is preferred; buccal DNA is variable and unstable.

l. Exclude urine if >10,000 wbcs, >1+ protein or glucose, or if bacteria are present. Place sample on ice, centrifuge for 10 minutes at 4°C at 3000 RPM, and freeze within 1 hour.

5. Large Repositories
Advantages include economies of scale from batching (for example, in genotyping for ApoE), backup facilities (multiple freezers, emergency power, liquid nitrogen), and a single site to distribute data and tissue to legitimate investigators. Disadvantages are a lack of uniformity in sample collection methods among contributing sites, reduced motivation to contribute to the national repository (rather than the investigator's own), and material transfer agreements (MTAs). The MTA codifies who has rights to discoveries based on repository data and tissue and can indemnify the repository from errors made at individual sites.

Issues for NIA-sponsored repositories include determining access to data and tissue (e.g., requests from commercial entities), data management, and quality control.

6. Analytical Methods
Large scale longitudinal studies (e.g., the Physicians Health Study II, the Nurses Health Study, and the Women's Health Study) collect specimens for assays that might predict disease development or aid in the understanding of disease etiology and risk factors. These studies generate very large data sets and demand a large amount of quality control. Methods at individual sites can vary widely, including different conditions for freezing samples (-20°C versus -70°C). Moreover, new techniques (often complex), new equipment, new reagents, or new technicians (e.g., staff turnover) may result in differences between assays. The time course also may be important: estrogen levels do not change appreciably in blood from time of collection to 48 hours later, whereas testosterone levels change because of ongoing metabolism in the tube. All these sources of variability can affect reproducibility. A small coefficient of variation (<10%) is desired but requires a large and critical amount of ongoing quality control. Assay issues for quality control include determinations of validity and variations: within person, between persons, batch-batch, and study-study.

7. Conclusions
The Workshop ended with the following goals:

• Identify specific questions to be addressed in studies of antecedent biomarkers for AD.
• Inventory existing studies that have collected biological samples in AD to learn what is available and what will be needed for future studies.
• Standardize sample collection and storage methods.