For decades, scientists have yearned for a blood test that can tell who accumulates amyloid in the brain. A simple blood draw would help fill clinical trials much more quickly and cheaply than PET scans or lumbar punctures. The dream looks set to become a reality. In the January 31 Nature, Japanese and Australian researchers unveil a sensitive and reproducible blood assay that can predict amyloid status with up to 90 percent accuracy. The test, which uses immunoprecipitation of Aβ peptides followed by mass spectrometry, works as well as existing CSF biomarkers at picking up brain amyloid. Katsuhiko Yanagisawa, National Center for Geriatrics and Gerontology in Aichi, Japan, joined with Shimadzu Corporation scientists and Colin Masters of the University of Melbourne in Australia to develop and test the serum-based method using two different research cohorts. 

  • New plasma Aβ assay predicts brain amyloid with 90 percent accuracy.
  • The assay works in independent clinical cohorts.
  • Blood tests promise to speed clinical trial recruitment, cut costs.

The report comes on the heels of work from Randall Bateman, Washington University in St. Louis, who presented and published a comparable blood test just last year (Jul 2017 conference news and Ovod et al., 2017). “It is remarkable that this group totally independently developed a very similar kind of assay, and found pretty much identical results,” Bateman told Alzforum. “This is really good news, and it tells me this approach is definitely going to work.”

Paul Aisen, University of Southern California, agreed. “The two groups’ efforts have clearly yielded plasma assays that are indicative of brain amyloidosis. That is a tremendous advance,” Aisen said. 

Ups and Downs. Quantitation of plasma Aβ peptides by mass spec reveals candidate biomarkers for amyloidosis in Japanese (NCGG) and Australian cohorts (AIBL). [Nakamura et al., Nature.]

The new assays need validation, including replication and optimization in different populations, but Bateman thinks that can happen quickly. It may only take a few years for blood biomarkers to make their way into practice, he predicted. 

The idea of a blood test to detect brain amyloid has been around almost as long as researchers have known about amyloid (Rumble et al., 1989). Despite many attempts over the years, no one had been able to overcome the technical challenge of precisely quantitating sticky peptides present in picomolar concentrations in a stew of thousands of other molecules. As it turns out, Koichi Tanaka, the Nobel Prize-winning chemist who helped perfect mass spec for proteins, has been quietly working on the problem. Tanaka and colleagues at Shimadzu Corporation immunoprecipitated APP-derived peptides from human plasma, then identified and quantitated them using matrix-assisted laser desorption/ionization time-of-flight mass spectrometry (Kaneko et al., 2014). Although MALDI-TOF is not normally considered a quantitative technique, Tanaka and colleagues managed to develop an accurate and reproducible assay for Aβ peptides, with variances less than 10 percent for repeat measures.

They focused on three Aβ peptides: Aβ42, Aβ40, and APP669-711. The latter comprises Aβ40 with an additional three amino acids at its N terminus. Its concentration in blood does not change in people with Alzheimer’s disease, so it serves as an internal control for fluctuations in Aβ42. In a small sample of 62 people, the APP669-711/Aβ42 ratio pinpointed who had brain amyloid with more than 90 percent accuracy (Kaneko et al., 2014). 

For the new study, first author Akinori Nakamura and colleagues refined and further tested the assay using binding of the amyloid PET ligand PiB as a gold standard. Initial testing in a discovery cohort of 121 Japanese, 50 of whom were amyloid-positive by PiB PET, was followed up with samples from a second group comprising 252 participants in the Australian Imaging, Biomarker, and Lifestyle Study of Aging. In AIBL, about half the participants had brain amyloid detected by a PiB, florbetapir, or flutemetamol scan. Each cohort contained a balance of people clinically classified as being cognitively normal, having mild cognitive impairment, or having AD.

Plasma Markers. Analysis reveals high sensitivity and specificity for plasma biomarkers in Japanese (left) and Australian cohorts (right). [Courtesy of Nakamura et al., Nature.]

The results were striking. In both cohorts, the researchers found a small but significant decrease in serum Aβ42 in amyloid-positive people compared with their amyloid-negative counterparts. Receiver operator characteristic (ROC), a statistical analysis that approximates accuracy, showed that by itself, Aβ42 moderately predicted PET positivity, with area under the curve (AUC) values of 87.2 percent in the Japanese and 71.8 in the Australian cohort. Ratios of Aβ40/42, APP669-711/42, or a composite biomarker that incorporated both ratios all demonstrated better predictive power, which was highest in the PiB-PET groups and slightly lower when other tracers were used. The Aβ40/42 ratio reached 90.9 percent accuracy in the Japanese cohort, 88.9 percent in the Australian PiB-only group, and 83.7 percent in the whole Australian cohort. The performance in the Australian cohort was virtually identical to what the St. Louis team achieved with their IP-MS assay in their AD cohort, Bateman told Alzforum.

In optimizing their biomarkers, the investigators broke with tradition and flipped the ratio, using Aβ40/42 as their readout, rather than the more common 42/40. They did that because putting Aβ42 in the denominator yielded data with a smoother distribution, which is important when trying to detect small differences in plasma Aβ between groups. “It’s a very small point but I think it could be a critical factor contributing to the high performance,” Masters said.

The most accurate predictor of amyloidosis turned out to be the composite, which the investigators calculated by averaging the Aβ40/42 and APP669-711/Aβ42 measures. Indeed, the composite proved equal to CSF Aβ at predicting PET positivity, showing about 80 percent accuracy in a small cohort of 46 people from AIBL who had both CSF and PET analysis.

Some commentators raised concerns about combining the two ratios in that way. “Ideally, a composite would come from two independent markers,” said Henrik Zetterberg, University of Gothenburg, Sweden. “The APP669-711/Aβ42 ratio is not independent from the Aβ40/42 ratio. In effect, the Aβ1-42 changes are counted twice." 

Sid O’Bryant, University of North Texas Health Science Center, Fort Worth, shared that concern. He wrote to Alzforum that the composite score could have skewed the data given that Aβ42 is the marker that was most impactful in the first place, and the only one that was normally distributed. But he feels the method might be valuable. “While the statistical approach is not what I would have anticipated, if others can replicate the use of this approach using these analytic methods, it is possible that the authors are onto something very novel and highly important," he wrote in an email to Alzforum. 

The investigators did see a systematic difference in the Aβ42 measurements between the cohorts, with values in the Australian group coming in approximately 25 percent lower. Since all the assays were done at Shimadzu, they believe the variation arises mainly during sample handing, an issue that has flummoxed CSF biomarker analysis as well (Apr 2017 conference news). Nonetheless, Nakamura was able to combine data from the two cohorts to derive a single, optimized cut-off value for the composite biomarker of 0.376. In the individual cohorts, this cut-off value still gave over 85 percent accuracy. In a separate, follow-up group of 51 Japanese subjects, the predetermined cut-off was 90 percent accurate at assigning amyloid status.

Erik Portelius, University of Gothenburg, Sweden, would like to see a comparison of the different methods being developed for serum amyloid assays. “The authors show convincingly that plasma Aβ40/Aβ42 ratio indeed reflects CNS-derived Aβ, since the plasma Aβ levels correlated with both CSF Aβ concentrations and amyloid PET. However, as the authors point out, the finding needs to be replicated. More importantly, studies comparing available methods (mass spectrometry-based and immune-based, such as ELISA) should be conducted in order to fully understand this novel promising plasma biomarker,” he wrote in an email to Alzforum (Dec 2017 conference news and Fandos et al., 2017). Portelius had some of the earliest success in measuring plasma Aβ42 by IP/mass spec (Pannee et al., 2014). 

Kaj Blennow, also from UGothenburg, said he, too, would like to see comparison with ELISA and the more sensitive single-molecule array (SIMOA) approach.

If blood tests can be validated, they will pay off by reducing the number of PET screens needed to identify people with preclinical AD for prevention trials, Masters said. “At the moment, to get 100 people into a preclinical study, we have to screen nearly 700 with a PET scan. If we use a blood test first, we need to scan only about 150. That’s a huge savings in cost and efficiency,” he said. A serum biomarker might eventually be useful for differential diagnoses of dementia, and for monitoring Aβ clearance in clinical trials, though both will take a lot more study, he said.

AIBL researchers will continue to collaborate with Shimadzu, and Masters said they have many experiments planned to look at the marker longitudinally, and to ask if it correlates with cognitive impairment.

Bateman reported his group just received a grant to accelerate the development and validation of their test. “We are moving forward with a large clinical study, looking at samples from existing studies, and collecting new samples, too,” he told Alzforum. C2N Diagnostics, a St. Louis company co-founded by Bateman and his WashU colleague David Holtzman, is leading efforts to develop his IP/MS assay for commercial use.

The authors said they are now working on an automated version of the assay, to standardize the analysis and increase throughput. They are open to accepting samples from other labs outside of Japan, Tanaka wrote in an email to Alzforum.—Pat McCaffrey

Comments

  1. This is a great paper! For many years, it was hard to get a reliable amyloid biomarker signal from blood and many inconsistent results were published. Many researchers, including us, believed that peripheral production of Aβ (from platelets, hepatocytes, and other cell types) made Aβ a poor biomarker for cerebral beta-amyloidosis when measured in blood. It was also hard to quantify Aβ in blood because of its low concentration and stickiness to other blood proteins. Nevertheless, during recent years, a number of studies have shown that plasma Aβ concentrations can be reliably quantified and actually do reflect amyloid deposition in the brain, if the methods are sensitive and specific enough. These include Janelizde et al., 2016, who quantified plasma Aβ using an ultrasensitive single molecule array (SIMOA) method, and Kaneko et al., Panee et al., and Ovod et al., who used mass spectrometry (Janelizde et al., 2016; Kaneko et al., 2014; Pannee et al., 2014; Ovod et al., 2017). Now, this beautiful paper further corroborates this and shows even stronger associations, most likely because of further refinements of the method.

    References:

    . Plasma β-amyloid in Alzheimer's disease and vascular disease. Sci Rep. 2016 May 31;6:26801. PubMed.

    . Identification and quantification of amyloid beta-related peptides in human plasma using matrix-assisted laser desorption/ionization time-of-flight mass spectrometry. Proc Jpn Acad Ser B Phys Biol Sci. 2014;90(3):104-17. PubMed.

    . The amyloid-β degradation pattern in plasma--a possible tool for clinical trials in Alzheimer's disease. Neurosci Lett. 2014 Jun 24;573:7-12. Epub 2014 May 4 PubMed.

    . Amyloid β concentrations and stable isotope labeling kinetics of human plasma specific to central nervous system amyloidosis. Alzheimers Dement. 2017 Aug;13(8):841-849. Epub 2017 Jul 19 PubMed.

  2. The authors should be congratulated on developing a highly specific blood test for amyloidosis and demonstrating it so well. It was a fantastic paper that found nearly identical results in performance of a mass spectrometry-based blood test for amyloid plaques to us (Ovod et al., 2017), and matches with our findings presented at AAIC that these results are highly reproducible across studies. The authors have done a top-notch job and the extensive studies across several cohorts confirming the findings and their work are impressive. The fact that two labs have independently accomplished this level of accuracy provides very high confidence this will be a robust way forward.

    References:

    . Amyloid β concentrations and stable isotope labeling kinetics of human plasma specific to central nervous system amyloidosis. Alzheimers Dement. 2017 Aug;13(8):841-849. Epub 2017 Jul 19 PubMed.

  3. It has long since been believed that measures of cerebrospinal fluid Aβ concentrations more likely reflect Aβ metabolism in the central nervous system (CNS). Previous studies on plasma Aβ have been contradictory and most groups have reported no change in plasma Aβ42 concentrations between Alzheimer´s disease patients and controls. In 2014 we published a paper in which we developed an immunoprecipitation method that we combined with mass spectrometric analysis to quantify several Aβ peptides in plasma (Pannee et al., 2014). In a small sample set including nine AD patients and 10 controls we were not able to detect a significant difference between the groups, but there was a trend toward lower concentrations of Aβ1-42 in AD.

    In this paper by Nakamura and colleagues, the authors have developed and optimized a highly sensitive method for relative quantification of three Aβ peptides: Aβ1-42, Aβ1-40, and APP669–711. The method is based on selective enrichment of the Aβ peptides by immunoprecipitation and analysis by MALDI TOF. For normalization of the data they used a stable-isotope-labeled Aβ1-38. The method was applied on two independent sample sets—one discovery data set comprising 121 subjects and one validation set including 252 subjects of which 111 individuals were diagnosed using amyloid positron-emission tomography (PET). The authors show convincingly that the plasma Aβ40/Aβ42 ratio indeed reflects CNS-derived Aβ since the plasma Aβ levels correlated with both CSF Aβ concentrations and amyloid PET. However, as the authors point out, the findings need to be replicated. More importantly, studies comparing available methods (mass spectrometry-based and immuno-based, such as ELISA) should be conducted in order to fully understand this novel promising plasma biomarker.

    References:

    . The amyloid-β degradation pattern in plasma--a possible tool for clinical trials in Alzheimer's disease. Neurosci Lett. 2014 Jun 24;573:7-12. Epub 2014 May 4 PubMed.

  4. As the authors state carefully and clearly, these findings need to be replicated, the analytic methods need to be confirmed, ROC performance in other brain diseases is needed, and validation in clinical trials of AB lowering agents is needed.

    That said, if we had an inexpensive, minimally invasive, and repeatable biomarker of cerebral amyloidosis, it would represent a major contribution to the field. This would allow acquisition of multiple measures in trials, for example, during a run-in period as well as during the treatment period. The gain in statistical power would be tremendous and would facilitate rapid go-no/go decisions inexpensively and allow us to address many more trial hypotheses quickly (e.g., dosing, combination therapy).

  5. These results are very promising, and lend further support to the use of blood Aβ as a screening biomarker for brain amyloidosis. The results add to the recent paper published last year that also used immunoprecipitation and mass spectrometry for measurements (Ovod et al., 2017) and showed high concordance for the plasma Aβ42/Aβ40 ratio with amyloid PET, and outperform concordance obtained using immunoassay techniques such as SIMOA (Janelidze et al., 2016). For future possible clinical applications, it would be highly interesting to see a direct method comparison of the mass spectrometry and immunoassay techniques for quantification of blood Aβ.

    References:

    . Amyloid β concentrations and stable isotope labeling kinetics of human plasma specific to central nervous system amyloidosis. Alzheimers Dement. 2017 Aug;13(8):841-849. Epub 2017 Jul 19 PubMed.

    . Plasma β-amyloid in Alzheimer's disease and vascular disease. Sci Rep. 2016 May 31;6:26801. PubMed.

  6. This is very interesting and promising discovery work. The authors provided an initial replication study in AIBL, which further supports this very early work. Much remains to be done, but I am excited by these findings and leveraging both of these studies was an excellent way to do the work. I really do applaud the authors for this effort. This is one of the largest such studies to date; however, the sample size is still relatively small so replication across additional studies and laboratories is still needed.

    The composite ratio score that is created from normalized intensities with Ab42 utilized twice in the formula is an interesting approach that could have skewed the data, given that this is the marker that was most impactful in the first place and the only one that was normally distributed. When you look at the bar graphs, there is tremendous overlap across groups. While the statistical approach is not what I would have anticipated, if others can replicate the use of this approach using these analytic methods, it is possible that the authors are onto something very novel and highly important. I would assume that others will be seeking access to the data set for independent analyses very soon to see if they can replicate the findings with different analytics.

    The real validation of any work in this area is prospective application within a primary care setting where such a biomarker is most needed: that is, a blood test to tell the primary care doctor if this specific patient should or should not be referred for confirmatory PET scans or LP for amyloid assay. The base rates and statistics will work out very differently than within a clinic-based cohort where 50 percent of those in the sample are amyloid-positive; in a primary care setting the majority of patients will not be amyloid-positive so you are looking for a minority of patients and need to cost-contain by screening out as many patients as possible (not to mention providing patients with a sense of ease with a negative finding). To date, no such study has been completed globally on any of the promising biomarkers, but this looks promising for such a setting.

    Overall, I am very encouraged by these findings. They are great for the field, as this particular line of work is very important and needed. It is still early discovery work and requires additional validation, but I am cautiously optimistic.

  7. This is a well-planned, impressive study that relates brain Aβ amyloidosis with plasma Abeta and related peptide parameters. The methods used, IP-MS, are relatively simple, so there are going to be additional studies that try to confirm the observations made here.

    It was a smart idea to use isotope Aβ1-38 as an internal control, although it remains unclear to me why the authors avoided using the Aβ1-42/Aβ1-38 ratio for the prediction.

    The other question is "Where do all the Aβ and related peptides come from?" If they are derived from vascular endothelial cells and platelets, then the cause-and-effect relationship and mechanism become unclear.

    References:

    . Brain endothelial cells produce amyloid {beta} from amyloid precursor protein 770 and preferentially secrete the O-glycosylated form. J Biol Chem. 2010 Dec 17;285(51):40097-103. PubMed.

  8. This plasma assay used IP-MS methodology, with standardization of plasma levels of Aβ42, 40, and a novel peptide by inclusion of Aβ38. The assay methods are well described, and the CV’s for the various plasma measures were low (around 5 percent). There was high sensitivity and specificity when calibrated against amyloid PET imaging and (in a much smaller subset of subjects) against CSF Aβ42 measured by ELISA. The findings represent a step forward in developing a plasma screening test for CNS amyloid deposition, and the study design, with discovery and replication cohorts, is commendable.

    Several recent studies on plasma Aβ are not mentioned in this manuscript. In particular, Ovod et al. used a similar IP-MS approach with internal standards to quantify plasma Aβ42 and 40, and found comparably strong relationships with markers of CNS amyloid, albeit in a single Center study (Ovod et al., 2017). Another study measured soluble and bound forms of Aβ in plasma in the AIBL cohort, and found estimates of sensitivity and specificity in that cohort that appear similar to those for AIBL in the present study (Fandos et al., 2017). The differences between the extremely high sensitivity/specificity in the Japanese cohort and the lower sensitivity and specificity (and overlap in values presented in Figure 1 for AIBL) is not well explained. As a general comment, for many biomarkers, initial performance through best fits to develop cutoffs in discovery cohorts is often higher than performance in follow-up studies.

    Given the expense of amyloid PET imaging, the use of a plasma marker to improve the efficiency of ordering a confirmatory test is supported by this study, and replication, extension, and comparison with other methods will take us much closer to this goal.

    References:

    . Amyloid β concentrations and stable isotope labeling kinetics of human plasma specific to central nervous system amyloidosis. Alzheimers Dement. 2017 Aug;13(8):841-849. Epub 2017 Jul 19 PubMed.

    . Plasma amyloid β 42/40 ratios as biomarkers for amyloid β cerebral deposition in cognitively normal individuals. Alzheimers Dement (Amst). 2017;8:179-187. Epub 2017 Sep 12 PubMed.

  9. This is an important paper and a significant advance in the field. The key question remaining concerns the relevance of Aβ40 and/or 42 in the etiology of the disease and whether this reflects a downstream event in the pathological process. 

  10. The article by Nakamura et al. is really good news for those who have been working on developing Aβ blood-based biomarkers for a long time. 

    The feeling in this particular research area, traditionally overwhelmed by controversy, started to change when the validation gold standard was changed from the clinical diagnosis to Aβ burden in the brain (as determined by PET or CSF analysis).

    New approaches are always welcome and Nakamura’s paper, together with the previous by Ovod et al. and by Janelidze et al., have dramatically helped elucidate that inexpensive blood-based biomarkers are viable and should be a top research priority for advancement in the fight against AD. In our own studies (Fandos et al., 2017), we have shown that validated ELISA tests for plasma Aβ (Pérez-Grijalba et al., 2016), at reference laboratories, can also be developed into a useful prescreening tool for secondary prevention clinical trials.

    However, we should be aware that an early and accurate diagnosis of such a complex disease as AD will most likely require a combination of biomarkers that reflect the different pathological mechanisms driving the disease progression.

    References:

    . Plasma amyloid β 42/40 ratios as biomarkers for amyloid β cerebral deposition in cognitively normal individuals. Alzheimers Dement (Amst). 2017;8:179-187. Epub 2017 Sep 12 PubMed.

    . Plasma β-amyloid in Alzheimer's disease and vascular disease. Sci Rep. 2016 May 31;6:26801. PubMed.

    . Amyloid β concentrations and stable isotope labeling kinetics of human plasma specific to central nervous system amyloidosis. Alzheimers Dement. 2017 Aug;13(8):841-849. Epub 2017 Jul 19 PubMed.

    . Validation of Immunoassay-Based Tools for the Comprehensive Quantification of Aβ40 and Aβ42 Peptides in Plasma. J Alzheimers Dis. 2016 Sep 6;54(2):751-62. PubMed.

Make a Comment

To make a comment you must login or register.

References

News Citations

  1. Finally, a Blood Test for Alzheimer’s?
  2. Are CSF Assays Finally Ready for Prime Time?
  3. Automated CSF Tests: Check. Blood Tests: In the Works

Paper Citations

  1. . Amyloid β concentrations and stable isotope labeling kinetics of human plasma specific to central nervous system amyloidosis. Alzheimers Dement. 2017 Aug;13(8):841-849. Epub 2017 Jul 19 PubMed.
  2. . Amyloid A4 protein and its precursor in Down's syndrome and Alzheimer's disease. N Engl J Med. 1989 Jun 1;320(22):1446-52. PubMed.
  3. . Identification and quantification of amyloid beta-related peptides in human plasma using matrix-assisted laser desorption/ionization time-of-flight mass spectrometry. Proc Jpn Acad Ser B Phys Biol Sci. 2014;90(3):104-17. PubMed.
  4. . Novel plasma biomarker surrogating cerebral amyloid deposition. Proc Jpn Acad Ser B Phys Biol Sci. 2014;90(9):353-64. PubMed.
  5. . Plasma amyloid β 42/40 ratios as biomarkers for amyloid β cerebral deposition in cognitively normal individuals. Alzheimers Dement (Amst). 2017;8:179-187. Epub 2017 Sep 12 PubMed.
  6. . The amyloid-β degradation pattern in plasma--a possible tool for clinical trials in Alzheimer's disease. Neurosci Lett. 2014 Jun 24;573:7-12. Epub 2014 May 4 PubMed.

Further Reading

No Available Further Reading

Primary Papers

  1. . High performance plasma amyloid-β biomarkers for Alzheimer's disease. Nature. 2018 Feb 8;554(7691):249-254. Epub 2018 Jan 31 PubMed.