15 October 2007. Plasma cholesterol measurements have done wonders for cardiovascular disease management. Could a blood test do the same for Alzheimer disease? In yesterday’s Nature Medicine online, researchers report that plasma levels of a panel of 18 proteins can identify Alzheimer patients with roughly 90 percent accuracy. The test performed similarly well in predicting which patients with mild cognitive impairment eventually went on to develop AD. “Our study is consistent with the concept that there is a systemic change [in AD] and that this change begins quite early,” said Tony Wyss-Coray, the study’s principle investigator, in an interview with ARF. The study could lead to the development of a diagnostic test and to new insights into the biology of the disease. Alzforum reported preliminary findings from this study, which were presented at a satellite workshop on AD biomarkers at the 2005 Society for Neuroscience meeting (see ARF related news story).
The findings come from a collaboration among Wyss-Coray and colleagues at Stanford University; Sandip Ray and colleagues at Satoris Inc., San Francisco (Wyss-Coray co-founded the company); and numerous collaborating researchers in the U.S. and Europe. “When we started this study it seemed, to some extent, a bit of a crazy idea, that you could study AD in the blood, but there have been several papers since that suggest there is active recruitment of immune cells into the brain and that they may modulate disease,” said Wyss-Coray. Just this year, for example, Joseph El Khoury and colleagues at Harvard Medical School reported that migrating microglia can modulate pathology in mouse models of the disease and that blocking chemokine signals exacerbates pathology (see ARF related news story). Wyss-Coray and colleagues had also opted to focus on chemokines, cytokines, growth factors, and other signaling molecules that might mediate crosstalk between the diseased brain and the periphery. “We decided to focus on a group of factors that have the highest density of information,” said Wyss-Coray.
Joint first authors Markus Britschgi at Stanford and Ray measured plasma concentrations of 120 well-characterized signaling proteins and looked for differences in levels between Alzheimer patients and non-demented controls. “This is one of the strengths of the study. There are advantages in looking for proteins for which there are sensitive assays already up and running, compared to trying to look at the plasma proteome which is a much less well-defined entity,” said co-author Douglas Galasko, University of California, San Diego, in an interview with ARF.
The researchers used a type of computer algorithm called predictive analysis of microarrays (PAM) to narrow the 120 to the smallest number of proteins that could distinguish AD patients from controls. The algorithm compares protein levels between the two groups to find which have the highest significance. Then it compresses the data to find only the markers necessary to discriminate between AD and controls. Using PAM, the researchers found a panel of 18 markers that classify AD and non-demented controls with 95 percent positive agreement and 83 percent negative agreement, respectively.
One might question how specific this test is for AD, since other diseases of the brain might elicit similar changes to plasma signaling molecules. “One of the problems I’ve always perceived is that if one looks at plasma from Alzheimer patients and controls, there may well be differences that don’t have to do with the neurobiology of Alzheimer’s disease but have to do with patients simply having a chronic illness, losing weight, having an inflammatory process, or a number of other questions along those lines, but the study goes a number of ways to address those issues,” said Galasko.
One way was to apply the test to a blinded set of samples to see if it could correctly distinguish AD patients from those with no, or other forms of dementia. The protein panel correctly identified eight of nine postmortem-confirmed AD patients, and also 10 of 11 patients with other dementias, who were classified as non-Alzheimer’s. The test also correctly identified patients with other neurologic diseases (Parkinson disease, ALS, multiple sclerosis, and peripheral neuropathy) and the autoimmune disorder, rheumatoid arthritis, as not having AD. These findings suggest that the panel of proteins is reporting on more than just inflammatory responses or neurodegeneration in general.
The researchers next asked if the test could predict which patients with mild cognitive impairment (MCI) might go on to develop AD; MCI is often, but not always, the precursor to Alzheimer’s. The researchers applied their protocol to 47 plasma samples taken from MCI patients at their time of diagnosis. Of 22 MCI patients who went on to develop AD 2-5 years later, 20 were correctly identified as AD-positive. The test also correctly classified as non-AD all eight MCI patients who subsequently went on to develop other dementias. Of the 17 patients still with an MCI diagnosis, the test predicts that seven will go on to develop AD. “The reason we don’t believe this is an artifact is, first of all, that we have a blinded test for Alzheimer’s and control, and then we use a completely independent MCI data set. It would be unlikely that the data would all fall into place like this,” suggested Wyss-Coray.
Nevertheless, he does acknowledge caveats. The main one concerns the sample sets, since the researchers were not able to always match AD samples perfectly with controls. One reason for this is that many clinics do not collect blood for plasma, but instead freeze it or use it for serum, explained Wyss-Coray. He said that plasma must be immediately prepared from the blood and then frozen. “The ideal study would be to have five centers with each center having 20 cases and control matches. Like every scientific study, this work needs to be reproduced,” he said.
That task will fall to Satoris Inc., which plans to develop an AD diagnostic. Wyss-Coray, for his part, plans to focus on basic biology. “What I’m interested in most, now, is looking at normal aging,” he said. He will compare plasma patterns from normal humans and mice to see if there are any similarities, and plans to study AD mouse models. He also wants to expand the scope to investigate Down syndrome and other dementias and increase the panel of 120 proteins to around 500. Galasko also thinks expanding the test will be worthwhile. “Tony has done a reasonably good job of coming up with the 120, based on assays that were available at the time, but there may be more signaling proteins that can be measured since then, and it may certainly be worth trying to replicate or extending these studies with a broader network,” he said.
One of the most interesting questions coming from this work might be what the panel of 18 proteins is saying about the disease process. “I was surprised that there were quite a few colony stimulating factors in the panel,” said Wyss-Coray. Levels of granulocyte colony stimulating factor (G-CSF), macrophage-colony stimulating factor (M-CSF), and interleukin 3 were all reduced in AD plasma. Wyss-Coray said that there are reports that all of these are neuroprotective and that some of them may promote neurogenesis. In fact, just this year a group from Taiwan reported that G-CSF can rescue cognitive decline in a mouse model of AD (see Tsai et al., 2007). “I find it so interesting that traditional immune molecules may have unexpected functions in the brain,” said Wyss-Coray. G-CSF is currently undergoing clinical trials in Germany for stroke.
Could this panel of 18 proteins serve as the basis for a diagnostic test? Satoris Inc. is planning to reproduce the study. “We will put this on a real clinical platform and do it again with additional hundreds of samples and come out with a researcher test that can be used by pharma, research institutions etc.,” said Patrick Lynn, president and CEO of the company. Lynn said Satoris is currently working with the Memory and Aging Center at UCSF and other institutions to develop prospective sample collections, and also wants to look at retrospective collections. He said after the “research use only” test, they are planning to develop a CLIA- approved test that could be used by clinicians. (The CLIA, or Clinical Laboratory Improvement Amendments, program, run by the Center for Medicare and Medicaid Services, regulates laboratory testing done on humans in the U.S.; see http://www.cms.hhs.gov/clia/). Ultimately, the company wants to develop a diagnostic test that would be approved by the FDA. “It is hard to predict when that might get approval,” said Lynn, “but our hope is that sometime in 2009 we would like to have FDA approval.” The CLIA test could come as early as next year.
A blood test for AD would have advantages over more complex tests, such as measuring cerebrospinal fluid markers or running magnetic resonance imaging or PET scans. “One could envision plasma being used as a screening test to trigger PIB, or MRI or CSF analysis,” suggested Galasko. Wyss-Coray agreed that it would not be a stand-alone test for AD but could trigger further testing, much like prostate-specific antigen measurements can trigger biopsies or other testing for prostate cancer.
Some question the value of a diagnostic test when a treatment is yet to be found that slows or halts the progression of Alzheimer disease. It might be worth keeping in mind that plasma cholesterol measurements were around long before statins. A blood diagnostic for AD could help drug development.—Tom Fagan.
Ray S, Britschgi M, Herbert C, Takeda-Uchimura Y, Boxer A, Blennow K, Friedman LF, Galasko DR, Jutel M, Karydas A, Kaye JA, Leszek J, Miller BL, Minthon L, Quinn JF, Rabinovici GD, Robinson WH, Sabbagh MN, So YT, Sparks DL, Tabaton M, Tinklenberg J, Yesavage JA, Tibshirani R, Wyss-Coray T. Classification and prediction of clinical Alzheimer diagnosis based on plasma signaling proteins. Nat Med. 2007 Nov;13(11):1359-62. Abstract