|Home | About | Journals | Submit | Contact Us | Français|
Advances in our understanding of tau-mediated neurodegeneration in Alzheimer's disease (AD) are moving this disease pathway to center stage for the development of biomarkers and disease modifying drug discovery efforts. Immunoassays were developed detecting total (ttau) and tau phosphorylated at specific epitopes (p-tauX) in cerebrospinal fluid (CSF), methods to analyse tau in blood are at the experimental beginning. Clinical research consistently demonstrated CSF t- and p-tau increased in AD compared to controls. Measuring these tau species proved informative for classifying AD from relevant differential diagnoses. Tau phosphorylated at threonine 231 (p-tau231) differentiated between AD and frontotemporal dementia, tau phosphorylated at serine 181 (p-tau181) enhanced classification between AD and dementia with Lewy bodies. T- and p-tau are considered “core” AD biomarkers that have been successfully validated by controlled large-scale multi-center studies. Tau biomarkers are implemented in clinical trials to reflect biological activity, mechanisms of action of compounds, support enrichment of target populations, provide endpoints for proof-of-concept and confirmatory trials on disease modification. World-wide quality control initiatives are underway to set required methodological and protocol standards. Discussions with regulatory authorities gain momentum defining the role of tau biomarkers for trial designs and how they may be further qualified for surrogate marker status.
The “Biomarker Working Group of the National Institute on Aging” (Frank et al., 2003) made first critical suggestions regarding implementation of Alzheimer's disease (AD) biomarker candidates for clinical trials and recommended a specified panel of “core”, feasible” biomarker candidates for validation within the worldwide Alzheimer's Disease Neuroimaging Initiative (ADNI). Therefore, in this review and update on tau-related biomarkers we focus on established, pre-selected and validated (by international consensus processes) “core feasible” biomarkers, which have been evaluated in several studies by independent international research groups (with the greatest available level of evidence), and give a practical guide to their implementation in clinical routine and their potential role in clinical trials. Biomarkers may have a key role in AD trials in the baseline evaluation of patients eligible for the trial and as diagnostic markers to enrich the patient sample with pure AD cases (Blennow and Hampel, 2003; Carrillo et al., 2009; Hampel et al., 2008). First international multi-center reliability studies using core, feasible biomarker candidates from cerebrospinal fluid (CSF) have been successfully concluded (Buerger et al., 2009; Shaw et al., 2009). Among other current uses of tau-related core feasible biomarkers in such trials is for patient stratification and enrichment (Hampel and Broich, 2009).
AD is a progressive neurodegenerative disorder that causes dementia in approximately 10% of individuals older than 65 years (Blennow et al., 2006). One of its typical brain lesions is neurofibrillary tangles (NFTs) that consist of hyperphosphorylated forms of the microtubule associated protein tau that is assembled into paired helical filaments or PHFs (Grundke-Iqbal et al., 1986; Kosik et al., 1986; Lee et al., 1991; Nukina and Ihara, 1986; Wood et al., 1986). Tau expression is high in non-myelinated cortical axons, especially in the regions of the brain that are involved in memory consolidation such as the limbic cortex including the hippocampus (Trojanowski et al., 1989). Hyperphosphorylation of tau causes the protein to detach from the microtubules, thereby destabilizing microtubules and compromising axonal transport (Bramblett et al., 1993; Ishihara et al., 1999). While tau phosphorylation promotes axonal and synaptic plasticity in the developing brain (Lovestone and Reynolds, 1997), it is pathological in the adult brain and specifically related to a group of disorders referred to as tauopathies, which includes AD and some forms of frontotemporal dementia (FTD) (Ballatore et al., 2007).
Tau proteins may be considered promising candidate biomarkers for Alzheimer-type axonal degeneration and NFT formation. However, molecular characterization of (CSF) tau presents an analytical challenge for several reasons. One is the high heterogeneity of the protein: in the adult human brain there are six different tau isoforms produced from a single gene by alternative mRNA splicing (Figure 1). This heterogeneity is compounded by extensive post-translational modifications, including phosphorylation, glycosylation and oxidation of the protein (Hernandez and Avila, 2007). Of a potential 79 serine and threonine phosphorylation sites in the longest isoform, 39 different sites have been verified (Hanger et al., 2007). It is not clear if tau phosphorylation plays a critical role in regulating the propensity of the protein to aggregate, nor is it known if the hyperphosphorylation of tau (triggered by increased rate of phosphorylation and/or decreased rate of dephosphorylation) and tangle formation are a cause or a consequence of AD (Ballatore et al., 2007). However, most studies, both in animal models (Gotz et al., 2001; Lewis et al., 2001) and those in longitudinally followed elderly humans (Gustafson et al., 2007; Stomrud et al., 2007), suggest that tau pathology may be downstream of the amyloidogenic cascade in AD, but it is clear that tau pathology alone causes neurodegeneration as exemplified by familial and sporadic tauopathies (Ballatore et al., 2007).
Another technical challenge is the low concentration of tau in CSF, ranging from approximately 300 ng/L in healthy individuals to 900 ng/L in AD patients (Blennow and Hampel, 2003). Considering that this quantity is distributed over many different modified forms and six splice variants, the amount available for analysis of each molecular species falls close to the detection limit of most assays. Nevertheless, tau was recently immunoprecipitated from human CSF and characterized by mass spectrometry (Portelius et al., 2008), showing that CSF tau indeed is amenable to detailed molecular characterization. Pilot data using immunoprecipitation and western blot recently revealed a tau isoform pattern specific to progressive supranuclear palsy (Borroni et al., 2009; Borroni et al., 2008), suggesting that certain tau isoforms may reflect disease-specific neurodegenerative processes. These studies, however, warrant replication.
Enzyme-linked immunosorbent assay (ELISA) has been used extensively to determine CSF tau concentrations in clinical samples. Initial studies utilized antibodies insensitive to the modification status of the protein, thereby measuring the total tau (t-tau) protein concentration (Figure 1). The first study in which t-tau could be successfully analyzed in CSF was published in 1995 and showed that the t-tau concentration was significantly elevated in AD patients compared with controls as well as in patients with other neurodegenerative disorders (Arai et al., 1995; Blennow et al., 1995). Of the more than 50 studies conducted on AD patients and controls to date, almost all have shown an increase in t-tau in AD patients by approximately 300% with a sensitivity and specificity of 80–90% (Blennow and Hampel, 2003; Blennow and Zetterberg, 2009; Hampel et al., 2008; Shaw et al., 2009) (Table 1).
As described by Buerger et al. (Buerger nee Buch et al., 1999) t-tau has a statistically significantly greater discriminative power in the young (<70y/o), compared to the old-old (>70y/o) study population of healthy age-matched control subjects, AD cases and comparisons with major-depression. Therefore, the effect of age on t-tau concentrations and proposed age-related cut-offs (Buerger nee Buch et al., 1999) should be taken into account, particularly in studies investigating t-tau as a single diagnostic marker, however, this does not seem to contribute additional diagnostic value in current multi-center validation trials using the combined set of the three core feasible CSF biomarkers.
When comparing the CSF t-tau concentration of AD patients with that of other neurodegenerative diseases, such as FTD or vascular dementia (VaD), the specificity drops to approximately 50–60%, rendering t-tau of limited use as a diagnostic marker for distinguishing AD from other dementing illnesses (Blennow and Hampel, 2003; Shaw et al., 2007). In fact, t-tau seems to be a general marker of damage to cortical axons, a view substantiated by results from studies of stroke, brain trauma and Creutzfeldt-Jakob disease (Blennow et al., 2005; Hesse et al., 2001; Ost et al., 2006; Otto et al., 2002; Riemenschneider et al., 2003; Zetterberg et al., 2006).
By using antibodies recognizing specific phosphorylated motifs in the tau amino acid sequence (p-tau), some p-tau isoforms (mainly p-tau181, −199 and −231) were found that appeared to be more characteristic to AD. P-tau231 and p-tau181 can be used to distinguish AD from control groups and even from FTD, dementia with Lewy bodies (DLB), VaD and major depression (Bian et al., 2008; Buerger et al., 2003; Grossman et al., 2005; Hampel et al., 2004a; Hampel and Teipel, 2004; Vanmechelen et al., 2000). For review of t- and p-tau as biomarkers for FTD see Hampel and Teipel (2004). While other p-tau species also have been measured such as p-tau199, p-tau199+202, as well as p-tau396+404, most studies have focused on p-tau231 and p-tau181 (for review see Blennow and Hampel, 2003).
CSF p-tau levels correlate with cognitive decline in patients with mild cognitive impairment (MCI) (Buerger et al., 2002a) and with neocortical NFT-pathology in AD (Buerger et al., 2006). Furthermore, both t-tau and p-tau predict rate of cognitive decline in different stages of AD (Blom et al., 2009; Buerger et al., 2005; Samgard et al., 2009) and concentration of p-tau231 declined longitudinally from mild to moderate AD (Hampel et al., 2001) and correlated significantly at baseline with rate of hippocampal atrophy in mild to moderate AD, acting as an indicator of structural disease progression (Hampel et al., 2005). In a recent European multi-center-study CSF p-tau reliably predicted AD in subjects with MCI with high accuracy (80%) as a single biomarker in a relatively short but clinically relevant observation interval of 1.5 years (Ewers et al., 2007). Finally, numerous studies have shown that a combination of the three core feasible biomarker candidates t-tau, p-tau and the 42 amino acid form of amyloid beta (Aβ42) can be used with optimized accuracy to detect incipient AD in subjects with MCI with positive and negative predictive values of >80% (Hansson et al., 2006; Herukka et al., 2005; Mattsson et al., 2009; Zetterberg et al., 2003) (Table 2), and addition of Apolipoprotein E (APOE) gentotype adds further to these measures as shown in a recent study from ADNI (Shaw et al., 2009). These results imply that CSF t-tau and p-tau are useful both as diagnostic markers for AD and as markers of disease intensity. Using the monoclonal antibodies (MAbs) AT180 and AT270 which recognize p-tau181+231, respectively, on human tau (Goedert et al., 1994), an ELISA method was developed to measure p-tau181+231 (Blennow et al., 1995). A marked increase was found in AD compared with controls and patients with other neurodegenerative disorders (Blennow et al., 1995).
As reviewed in Brunden et al. (Brunden et al., 2009), tau is normally phosphorylated at multiple serine and threonine residues, and tau hyperphosphorylation reduces microtubule binding and may enhance aggregation. Therefore, it is possible that changes in protein kinases and/or phosphatases could enhance tau phosphorylation. A number of kinases and phosphatases have been implicated as contributing to tau hyperphosphorylation including the kinases glycogen syntase kinase 3 (GSK-3), cyclin-dependent kinase 5 (CDK-5) and microtubule-affinity regulating kinase (MARK) while protein phosphatase 2A (PP-2A) appears to be the most plausible phosphatase involved in this abnormal posttranslational modification of tau (Matsuo et al., 1994). However, additional post-translational modifications may also contribute to tau pathology and dysfunction. For example, tau undergoes a specific type of serine/threonine O-glycosylation and these modifications can reduce the extent of tau phosphorylation. Thus, a decrease in tau O-glycosylation could result in increased hyperphosphorylation. Tau can also be tyrosine phosphorylated, sumoylated and nitrated, but it is not fully understood what effects these modifications have on tau.
The implementation of biochemical markers in trials on potential disease modifying compounds is still in its infancy. Phosphorylation of tau is currently hypothesized to be a downstream product of Aβ toxicity, and an increase in t- and/or p-tau levels in CSF is thought to reflect neuronal cell death with release of tau-related proteins into the extracellular CSF compartment. The temporal relationship between impacting Aβ production or aggregation and tau phosphorylation and neuronal cell death has not been defined, and the expected effect on tau of anti-Aβ therapies is uncertain. Tau-related therapies might inhibit new paired helical filament and tangle formation or the molecular sequence of a specific tau hyperphosphorylation pattern in AD without impacting existing levels of these measurable proteins in CSF. A small number of patients included in the phase IIa AN1792 active vaccination trial who had lumbar punctures and tau measures evidenced a mean significant decrease in CSF t-tau concentrations (Gilman et al., 2005) in vaccinated AD cases versus placebo-treated controls, however, the anti-Aβ passive immunization trial AAB-001 showed no statistically significant effect on CSF tau concentrations (Grundman and Black, 2008). In a recent placebo-controlled randomized single-blinded multicenter study lithium (Hampel et al., 2009b) was tested for the treatment of AD. As the primary outcome variables in vivo CSF and plasma biomarkers were used (t-tau, p-tau, Aβ42, GSK-3). Over the observation period of 10 weeks, lithium-treated patients did not show a significant change in CSF or plasma markers or neuropsychological performance when compared to placebo treated patients at two timepoints. This disappointing clinical and biomarker outcome of this proof-of-principle trial led to the final termination of another planned large National Institutes of Health (NIH) Alzheimer's Disease Cooperative Study (ADCS) lithium trial program in the US, therefore sacrificing a potentially effective treatment approach in the light of conceptual design problems regarding the implementation and value of biomarkers as outcomes in a particular trial. In conclusion, it may have been premature to pull lithium from trial because of a lack of change in tau biomarkers since lithium targets GSK-3 which also modulates Aβ. In order to reduce interpretation caveats at late stages of phase 2 and 3 trials, biomarker candidates for the various roles and functions in trials should be best co-developed with drugs as an integrated part of the trial program and best driven through preclinical to clinical stages.
Currently, the relationship between the diminished CSF Aβ or t- and p-tau concentrations and clinical outcome in clinical trials is still unknown. The degree of pathological changes in the levels of CSF Aβ, t- and p-tau has not been credibly linked to a specific degree of improvement on clinical outcomes such as the Alzheimer's Disease Assessment Scale-cognitive subscale (ADAS-cog), the Mini-Mental State Examination (MMSE), or the Clinical Dementia Rating Scale (CDR). Reducing the levels of Aβ, t- and p-tau would indicate that a biological target has been hit. This might be sufficient in some cases to support advancing clinical trials (go- no go decision to proceed to next phase) of potentially disease modifying compounds. In pivotal trials they might play a supporting role to clinical outcomes. Although decreasing Aβ, t- and p-tau is the primary objective of anti-amyloid agents such as secretase inhibitors, changing CSF measures should be viewed as an intermediary goal; avoiding disease progression and irreversible neuronal death is the main objective of treatment (Table 3). The quantitative linkage between turnover, release, expression and concentration of core CSF measures and cell death is not yet established in humans, however, measures of t- and ptau might most likely indirectly reflect neuronal cell death. More than in trials on potential disease modifiers, future prevention trials might depend entirely on biomarker outcomes. Therefore, validating core biomarkers, such as CSF t- and p-tau for use in primary prevention trials is clearly another key research priority.
To date, based on accumulated evidence from the rapidly progressing international controlled multi-site biomarker validation field (Ewers et al., 2007; Hansson et al., 2006; Mattsson et al., 2009; Vemuri et al., 2009a, b; Visser et al., 2009), tau-related biomarker information can be suggested for support recruitment and inclusion criteria, regarding enhanced diagnostic accuracy between AD and age-matched controls, for predicting AD in subjects with MCI, predicting time of cognitive decline and/or conversion to AD, enrichment of target populations for proof-of-concept trials with reduced sample variance, lower sample size and trial cost, may serve as a additional safety measures, and in the continued absence of a validated surrogate marker, may substantiate the claim for disease modification as secondary outcomes in addition to the demonstration of efficacy in the required clinical co-primary outcomes cognition and function.
To this end, further multi-cite and assay quality control issues (see ongoing international QC-program supported by the Alzheimer Association) need to be solved and in close collaboration with regulators qualification and validation of CSF tau biomarkers is urgently required for effective use in trials (for conceptual review see Cummings, 2009; Frank et al., 2003; Hampel and Broich, 2009; Hampel et al., 2009a) (for detailed review and discussion of other studies with preliminary findings that assessed tau biomarkers in AD treatment trials where there was either a treatment effect that tested for the possible value of biomarkers as indices of (a) neurodegeneration and possible reversal see Gilman et al., 2005 or where there was a (b) specific biochemical effect see Lannfelt et al., 2008).
There are very little data on tau levels in peripheral bodily fluids so much more research is needed to confirm and extend these studies as it is not entirely clear how increased brain levels of tau in AD might be reflected in elevated levels of tau in blood. However, the continuous production of CSF requires that it exit the subarachnoid space surrounding the brain and presumably, as CSF is drained through the subarachnoid granulations into the venous circulation, products released from the brain into the CSF could be conveyed to blood when CSF enters the venous circulation. However, there might be peripheral sources of tau that could account for tau being present in the blood such as when minor peripheral nerve injuries occur that release tau from injured nerves into soft tissues so it can be absorbed by the local vasculature. Thus, while the initial reports on the detection of tau in blood are promising, further research is needed in order to exploit these findings for the development of a validated assay to measure tau in blood.
The CSF is in direct contact with the brain interstitial fluid and thus probably provides a more accurate measure than peripheral blood of tau metabolism. However, despite the low frequency of post-lumbar puncture headache (Blennow et al., 1993; Peskind et al., 2005), spinal tap is commonly regarded as an invasive and time-consuming procedure that is difficult to implement in large studies. Although certain positron emission tomography techniques seem to visualize NFTs directly in living patients (Small et al., 2006), the procedure is very expensive and has limited availability. Hence, determination of tau proteins in peripheral blood would be of great value, especially in large clinical studies. In spite of considerable efforts, however, reliable methods to determine tau protein levels in serum or plasma in patients with neurodegenerative diseases are still lacking. One pilot study reported detectable tau immune reactivity in plasma without any clear relation to dementia diagnoses (Ingelson et al., 1999). Another study reported transient increases in serum tau protein levels measured by conventional ELISA in a subset of stroke patients (Bitsch et al., 2002). These data are promising but highlights the need for improved assays with lower detection limit and higher specificity in order to make them useful in the evaluation of patients with neurodegenerative diseases, in which the release of tau proteins to the peripheral circulation most likely is much lower than in acute stroke. For comprehensive review of blood-based biomarker candidates in AD see Schneider et al., 2009; and Ewers et al., 2009.
Based on the foregoing, it is clear that measures of t-tau and p-tau in CSF are informative for the diagnosis of AD and the distinction of AD from other neurodegenerative disorders including FTD, but it is equally evident that no single biomarker in CSF or neuroimaging modality will be sufficient for the early reliable diagnosis and prediction of AD. Indeed, biomarkers have a number of critical applications beyond diagnosis of the disease state including identifying those at greatest risk for AD among cognitively normal individuals and those with MCI as well as for use as indicators of disease progression and response to disease modifying therapies. Thus, even at this advanced stage in the quest for “ideal” core AD biomarkers, the use of algorithms that combine measures of CSF t-tau, p-tau, Aβ42, the APOE genotype as well as validated and suitable neuroimaging modalities (such as MRI-based hippocampal or mediotemporal lobe indices) (Chou et al., 2009; Schuff et al., 2009; Vemuri et al., 2009a, b) appear to be the most informative for the applications discussed here. Indeed, as disease modifying therapies that target Aβ42 and tau move dynamically forward in ongoing clinical trials, it will be increasingly relevant to have at hand these and other biomarkers that will accelerate the pace of finding interventions to prevent, arrest or reverse AD (Brunden et al., 2009; Hampel et al., 2009c). For comprehensive reviews on biomarkers in AD reflecting both microvascular alternations and amyloidogenic cascade related molecular mechanisms in AD see Hampel et al. (2009c) and Ewers et al. (2009).
This work was supported by a Science Foundation Ireland (SFI) investigator program award 08/IN.1/B1846 to HH, by grants from the NIH (AG-10124 and AG- 24904) to JQT and LMS, and by grants from the Swedish Research Council (2006-6227 and 2006-2740) to HZ and KB, Alzheimer's Association (NIRG-08-90356) to HZ and KB, and the Royal Swedish Academy of Sciences (FOTKAW09-039-A) to HZ.
Publisher's Disclaimer: This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final citable form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.