Blood Based Biomarkers in Alzheimer Disease: Current State of the Science and a Novel Collaborative Paradigm for Advancing from Discovery to Clinic

1.1. Methodological Considerations

One key advancement produced by the international Professional Interest Area (PIA) on Blood Based Biomarkers was the generation of the first-ever guidelines for pre-analytic processing of specimens[8]. This initial effort was the result of a tremendous work spanning industry and academic investigators from across the globe. It provided a basic set of pre-analytic processing variables to be followed (and refined) and a minimum set of information that should be provided within publications to allow for appropriately designed cross-validation efforts. More recently, this workgroup published data comparing biomarkers from the same blood draw (person, date and time) across assay platforms and blood fraction (serum and plasma)[9]. Results indicated that individual markers, while often statistically significantly correlated, may share minimal variance across platform or tissue indicating that direct comparisons are regularly not possible. Differences in concentration for specific analytes on different technological platforms can be due to a number of things including (1) calibrators, (2) neat biological samples or different dilutions may not have the same immunoreactivity with the antibodies included, (3) differences in antibodies as well as (4) differences in overall sensitivity and reliability of the instrument. Additionally, the use of different assay design can impact findings[10]. Together, this work clearly demonstrated methodological factors that must be considered when comparing across studies, cohorts, and biorepositories. Andreasson, Blennow and Zetterberg[11] provided an update and overview of ultrasensitive technologies to measure AD-related biomarkers in blood as well as CSF. While still early in the process, these novel assay technologies have the capacity to detect very low-levels of markers that may be of significant advantage when seeking to move from research-grade to “pharmaceutical-grade” kits in future attempts to take research use only (RUO) methods towards laboratory developed tests (LDTs) and in vitro diagnostics (IVDs)[12, 13]. As evident from the continued progress of the Global Biomarkers Standardization Consortium of CSF biomarkers (GBSC), the blood-based biomarker field will need to address additional methodological barriers in order to produce clinically useful and applicable biomarkers.

1.2 Blood Biomarkers of AD Risk

An important potential COU for blood-based AD biomarker science is the identification of individuals at greatest risk, which can take several forms: (1) risk of incident cognitive impairment and AD, (2) risk of progressing from mild cognitive impairment (MCI) to AD, and (3) risk for rapid progression within AD. Biomarkers related to these specific COUs have tremendous potential for clinical intervention trials aimed at preventing AD, halting progression from MCI, as well as slowing progression among patients with manifest AD. Enrichment of these specific subjects into trials has the benefit of reducing the diluting effect of enrolling those subjects not likely to progress. Indeed, an important potential of AD blood biomarkers could be to increase the likelihood of subjects being positive on more expensive (e.g. PET imaging) or invasive (lumbar puncture for CSF sampling) biomarkers used later to determine trial eligibility.

A substantial amount of work has been conducted examining blood-based markers within the COU of predicting progression of AD[14], conversion from MCI to AD[15] as well as risk for future AD[16]. Mapstone et al[17] recently examined plasma lipidomic and metabolomic markers from 525 community-dwelling older adults in an effort to identify a signature of risk for incident aMCI/AD. The authors identified a signature of 10 metabolites that yielded approximately 80% accuracy in discriminating controls from MCI/AD and 90% or greater accuracy in detecting those normal controls who converted to aMCI/AD over time. However, cross-validation attempts have been unsuccessful. Casanova et al[18] examined these same 10 metabolites in the Baltimore Longitudinal Study of Aging (BLSA) and the Age, Gene/Environment Susceptibility-Reykjavik Study (AGES-RS). In that work, these metabolites yielded an AUC=0.64 (BLSA) and an AUC=0.40 (AGES-RS) in these independent cohorts. Additionally, examining data from the Atherosclerosis Risk in Communities (ARIC) study, Li and colleagues[19] were unable to cross-validate the cross-sectional discrimination capacity of the 10 metabolites in discriminating normal controls from MCI/AD. This work and cross-validation attempts, is important to propel the field forward. Hye et al[20] analyzed plasma proteomics from 452 cognitively normal elders, 169 MCI non-converters, 51 MCI converters and 476 AD cases from across three independent cohorts, AddNeuromed (ANM), Kings Health Partners-Dementia Case Register (KHP-DCR) and Genetics AD Association (GenAD). A set of 10 proteins predicted progression from MCI to AD (average time of conversion approximately one year)(AUC=0.78).

There has also recently been a surge in research devoted towards the potential utility of exosome markers in predicting and detecting AD and other neurodegenerative diseases[21–23]. Recently, Rissman[24] examined the utility of neuronally-derived exosomes (NDEs) in predicting conversion from MCI to dementia. Alterations in plasma NDE levels of P-tau, Aβ1–42, neurogranin (NRGN) and repressor element 1-silencing transcription factor (REST) were found among AD and MCI cases that converted to AD within 36 months compared to stable MCI cases and normal controls. Additionally, when injected into the right hippocampus of wild-type (C57/BL6) mice, the NDEs from MCI cases that converted to AD caused increased P-tau when compared to NDEs from normal controls and stable MCI cases. This work significantly advances the utility of exosome biomarkers in AD and, critically, back-translates these findings into animal models for additional study, which is rarely done. A significant amount of work remains to standardize methods to effectively understand and work with exosome biomarkers; however, strong signals have been identified and confirm the need for this effort.

An example of a blood-based biomarker that has received a great deal of attention for predicting future risk is plasma clusterin (aka ApoJ). Levy[25] examine plasma clusterin from 1,532 non-demented subjects of the Framingham Study Offspring cohort to determine whether this putative biomarker could predict incident dementia and stroke. Among older adults (age>80), plasma clusterin was associated with increased risk for dementia; however, plasma clusterin was related to a reduced risk of dementia (age 60–69) and stroke (age < 80) among younger participants. These results suggest the importance of considering age when interpreting the predictive utility of this putative biomarker.

While still early in development, the above-described studies provide sufficient evidence for ongoing research investigating the potential use of blood-based biomarkers when considering the specific COU of predicting future risk. However, a great deal of additional work is required including, but not limited to, independent cross-validation, rigorous standardization of methods and assay technologies, and prospective studies designed to explicitly test the COU (with direct application of specific cut-scores). This COU may, in fact, be the “Holy Grail” of AD biomarkers and blood-based biomarkers provide an optimal first step in a multi-stage approach to addressing this COU. It is also possible that blood-based biomarkers may serve as the first-line in a multi-stage approach where the biomarker-specific COU is to rule out those least likely to progress, thereby screening out those who are not in need of more costly and invasive procedures, not only in clinical trial contexts but also in general medical practice. If this is the most valuable COU and market strategy, the design of the studies should be appropriately tailored.

1.3 Biomarkers of AD Diagnosis

The most studied potential COU for blood-based biomarkers in AD are diagnostic biomarkers. Some of this work seeks to identify screening tools for primary care clinics as part of a multi-stage approach[5] while others seek to identify diagnostic tools[26, 27].

One biomarker investigated in this potential COU is plasma total tau (T-tau) concentration. One study suggested that plasma tau was higher in the dementia stage of AD but the data were less clear in the MCI stage of the disease[28]. The Mayo Clinic Study of Aging recently reported that higher levels of plasma tau were cross-sectionally associated with worse memory performance and lower cortical thickness in an AD-signature region among non-demented individuals[29]. However, in analyses comparing eight groups defined by cognitive status and amyloid and neurodegeneration imaging markers there was high overlap between groups, suggesting that plasma tau may not be a good diagnostic AD biomarker. Unfortunately, there is no clear correlation between plasma and CSF T-tau concentrations[30] and this low correlation between CSF and blood biomarkers is quite common. In fact, different isoforms are present for almost every protein, in addition to the fact that the assay design can have an impact on the equilibrium between bound and free analyte in the sample. Another biomarker receiving a significant amount of attention in this COU is neurofilament light (NF-L). In contrast to tau, there is excellent correlation between CSF and plasma concentrations of NF-L[31]. CSF NF-L concentration is increased in both dementia and MCI stages of AD[28]. Additionally, these findings were recently replicated on serum and plasma samples[32].

Martins et al[33] examined baseline and 18-month follow-up plasma ApoJ (aka clusterin) concentrations in the AIBL cohort. The authors found that ApoJ levels were significantly higher among MCI and AD cases at both time-points and were also correlated with standardized uptake value ratio (SUVR) PET amyloid levels and hippocampal volume. Recently, specific glycosylated forms of ApoJ have been found to be more robust markers within this group. Nagele et al have conducted a series of studies examining the potential utility of autoantibodies in detecting AD and other neurodegenerative diseases[34, 35]. Recently from this lab, DeMarshall[36] examined serum autoantibodies from 236 participants (50 MCI with low CSF Aβ42 levels, 25 early stage Parkinson’s disease [PD], 25 mild-to moderate PD, 50 mild-moderate AD, 25 multiple sclerosis, 11 breast cancer, 50 controls). The top 50 differentially expressed autoantibodies were utilized for the classification analyses. The authors found >95% (96–100%) sensitivity and specificity for discriminating MCI from all other diagnostic categories within this cross-sectional cohort. Using the top 10 markers, excellent accuracy was retained for discriminating MCI from all categories. Mielke et al[37] recently analyzed plasma sphingolipid changes among autopsy-confirmed AD, Lewy Body Dementia (DLB) and control subjects. The authors found significant plasma ceramide alterations and monohexosylceramide alterations between dementia cases (AD and DLB) and controls suggesting that these biomarkers may have utility in identifying possible AD and/or DLB pathology. O’Bryant and colleagues cross-validated a serum-based algorithm for discriminating AD from controls across an independent platform, animal model and brain tissue and demonstrated preliminary data for the algorithm in discriminating AD from PD[38]. More recently, that group[5] created the locked-down referent cohort for an AD blood screen intended for primary care use and demonstrated excellent positive and negative predictive values when compared to screening tests. In the long-term, it is likely that the most viable and applicable COU for blood-based biomarkers within the “diagnostic” realm is to serve as the first-step in a multi-stage diagnostic process where CSF and PET amyloid and tau imaging will serve as the final diagnostics of presence of AD pathology[5]. Given the cost of PET and CSF methods relative to blood-based methods, the availability of a blood-based tool in primary care settings that is utilized to determine who does and does not undergo PET and CSF exams has a viable cost and patient acceptability strategies, which are also the strategies followed in the cancer arena (i.e., PET scans are not first-line diagnostics[39]).

1.4 Blood Biomarkers of Amyloid Pathology

Another COU with high potential to aid in clinical trials is the identification of blood-based biomarkers that can identify those individuals with high (or low) likelihood of being amyloid positive[40–42]. Westwood and colleagues[40] recently examined proteomic markers among longitudinal plasma samples collected over a 12-year period among non-demented individuals with [11C]PiB PET scans available. In this study, seven plasma proteins (including A2M, Apo-A1, and multiple complement proteins) were significantly associated with amyloid burden. In a small-scale pilot study, Kaneko[42] examined 40 PiB positive individuals (controls, MCI, AD) along with 22 PiB negative individuals (controls) and found that plasma amyloid proteins (Aβ40, Aβ42) and Aβapproximate peptides (AβAPs; APP669-71) were significantly correlated with amyloid PET positivity with a sensitivity and specificity of 0.93 and 0.96, respectively. In a larger analysis of 273 participants of the AIBL study, Burnham et al[43] identified a plasma-based nine-analyte signature that yielded a sensitivity and specificity of 0.80 and 0.82, respectively. Saykin and colleagues conducted a pilot study among 96 participants of the ADNI study and found a significant relationship between plasma amyloid and [11C]PiB uptake among APOE non-carriers[44]. None of these studies reflect the population that would be needed to support this COU. Though still very early in discovery phases, this COU has tremendous potential for influencing the design of clinical trials targeting amyloid.

The vast majority of the blood-based biomarker work described above (across potential COUs) remains in early stage discovery with only a few instances where multiple replication steps have been undertaken. If these discovery findings are to become clinically meaningful, a great deal of work must be completed. Besides the requisite cross-validation and longitudinal understanding, there is the significant need to fully comprehend the impact of other factors (including pre-analytical) on disorders and diseases and associated interventions on the levels of these blood-based biomarkers. For instance"hallmark” AD biomarkers have been shown to change in association with factors such as depression, cardiac arrest, head injury as well as hematological and cancer interventions[45, 46]. Additionally, the stratification of populations by genetics (ApoE4 genotype) or concomitant/comorbid diseases must be considered. These important considerations need to be addressed prior to considering moving from initial discovery towards the clinic. This process of going from discovery to clinic is best undertaken as a partnership between academia and industry/biotech in order to most effectively leverage the differing skillsets of those within each organizational structure and to appropriately consider the long-term plan and study designs. In the next section, we provide an updated model for advancing biomarker discovery through the stages of development towards clinical implementation.

2.1 Biomarker Development Concepts of Relevance and Available Resources

There are several relevant resources that can assist in the process of establishing a biomarker discovery program that has the goal of translating these discoveries to clinic.

2.2 Defining the Context of Use (COU)

While the definition of the COU is outlined above, this point warrants additional consideration. The COU sets the entire stage of science for any new putative biomarker and the importance of this step cannot be overstated as it is largely ignored in academic discovery science. While Industry-lead work has less room for basic biology discovery, this is one of the primary objectives of academic work. However, there are many interesting, novel and potentially useful discoveries that have little chance of reaching the clinic or impact patient care. Within the development of the COU, conceptualization and inception phases should include the following considerations: scalability, manufacturability, compatibility with traditional large-scale methods, intellectual property (IP), and regulatory pathway[47]. When using a blood-based biomarker for detecting AD in primary care clinics, each of these points will be considered individually:

1. Scalability – If the COU for the blood based biomarker is defined as “a detection tool for primary care clinics to determine which patients should or should not be referred for additional cost-intensive and invasive procedures”, how does scalability become a consideration? First, there are currently over 40 million Americans age 65 and above and this segment of the population will grow dramatically in the near future[59]. If a novel biomarker is to fit the CMS-regulated annual wellness visit (AWV), which is a current need based on the 2015 report of the Gerontological Society of America[60], it must be available to all primary care clinics. This translates to a biomarker platform that can potentially be performed on over 10,000 patients daily. Therefore, is this biomarker (or biomarker assay) sufficiently scalable to be offered to everyone in need? If the biomarker requires specialized equipment, cumbersome pre-analytic procedures or even a single reference lab, this biomarker will not meet the scalability needs of primary care providers.

2. Manufacturability – Academic investigators excel in creating new and novel procedures that surpass currently available methods. However, it is important to consider whether the product components can be produced at a level that will meet the 10,000 patients per day scale? If this is a new and novel platform, how easy is it utilized and produced? If this is a new biomarker (or set of biomarkers) that leverages existing platforms (e.g. ELISA)? Can the antibodies and reagents meet CLIA and good manufacturing practices (cGMP) and be produced large-scale reliably? What is the long-term availability of antibodies given, for instance, the inherent difficulties with long-term availability with polyclonal antibodies? If not, a potentially scalable biomarker without available manufacturing components will have to meet that latter need prior to being considered for movement towards clinic.

3. Compatibility with traditional large-scale methods – A blood-based biomarker serving the first-step in the AD assessment process for primary care providers offers advantages to increasing appropriate access to invasive and costly methods for confirmatory diagnostics (as well as therapies); however, how does this biomarker make it to primary care providers? There is an existing large-scale (global) network of companies specifically designed to collect and analyze blood samples at a scale far beyond 10,000 patients per day. However, can this biomarker fit into that infrastructure? Can this biomarker work with the existing global network of companies already providing primary care clinics with daily blood work results? If not, this barrier must be considered before the path to clinic can be realized.

4. IP Considerations – One phrase commonly heard in public-private-partnership meetings is “academics discover things and industry brings things to patients.” A new and novel biomarker that meets all of the outlined needs above that does not have an adequately structured IP strategy has a significantly reduced chance of reaching patients because there will be no financial incentive to capture an industry partner.

5. Regulatory Considerations - Considerations regarding regulatory issues early in the process also help appropriately design the studies, without which much of the data produced in the academic laboratory will likely be rendered useless when the regulatory path is realized.

Together, these points provide a contextual pathway to advance academic discovery toward clinical utilization. As can be seen in Figure 1, the standard inherent approach in academic research does not consider these points, which explains a tremendous amount of the failure to replicate academic findings in industry settings. In fact, the “reproducibility crisis” likely has little to do with the soundness of the academic research, but more to do with the context within which the work was conducted. Rather than the traditional “hand-off” model of scientific discovery findings to industry laboratories, a partnership that leverages the strengths of academic centers, pharmaceutical, diagnostic and biotechnology partners at the outset can greatly expedite moving new and better tools into the clinic in a manageable timeframe. Figure 2 proposes a new collaborative public-private partnership model that begins at the conceptualization of the biomarker itself. While industry’s primary expertise is not basic discovery, academic research aiming at discovery that is, from the outset, put into one of the above frameworks toward clinical implementation, will have a far greater chance of success. Additionally, this model allows for the “fail fast, fail forward” industry mindset. That is, one can identify flaws early in the process, learn from those issues and continue moving forward either with this asset or a new one. The new model combines the strengths of both academia and industry partners from the very outset and incorporates the IOM framework. From Figure 2, one can see the gradual shift in leadership from full academic lead on the left to full industry lead on the right. The “hand-off”, as traditionally conceptualized, would best be considered at the shift from Level 2 to Level 3 work. The four levels of the research are as follows:

2.3 STAGE 1 – Define Context of Use; Academic Lead

Defining the COU sets the stage for the entire program of research. Without a clearly defined COU, the research risks being unfocused without a clear path toward patient care. Clearly define the optimum and minimum performance criteria (e.g. sensitivity, specificity, positive and negative predictive value) and standards within the specific COU should be established. Bidirectional discussions from the outset between all partners is essential to fully describe and characterize the COU in order to ensure that the team is on the same page and remains focused towards a common goal. Additionally, early stage discussions should clearly articulate the considerations for the funding pathway based on existing models (e.g. Industry/State funded centers, Sponsored Research, corporate mini-labs[48]). Academic Roles: Innovation here is crucial. Academic partners are responsible for identifying and discovering new biomarkers that may have potential for marketability. The current model incorporating industry and other partners from the outset allows for (a) rapid communication and consideration of novel ideas and findings through a lense of their marketability, which also leads to (b) discussions of new and novel COUs for the biomarkers, (c) potential generation of improved methods and technologies that offer significant advantages over available biomarkers with similar COUs (keeping in mind that it is exceedingly difficult to “beat good enough” in the marketplace), and (d) identification of the infrastructure upon which to build the collaborative program of research. Industry Roles: Evaluate the COU within the competitive landscape, market value and opportunity, and focus on relevant endpoints and understanding of the regulatory pathway for assay and clinical validation as well as the potential for approval. Go-no-Go: Failure to identify a novel and useful COU that has a readily identifiable market potential.

Discovery and Test Validation Phase – Academic Lead