Introduction
The aggregation and propagation of hyperphosphorylated tau is a key feature of Alzheimer’s disease (AD) pathogenesis, and the deposition of these species into neurofibrillary tangles is one of the defining hallmarks observed at neuropathological examination of AD [
1]. Altered tau phosphorylation and secretion can be detected in cerebrospinal fluid (CSF), through measurement of tau phosphorylated at certain amino acid residues, with threonine 181 (p-tau181) being most commonly used, and total mid-region-containing species of tau (t-tau) [
2]. Recent research suggest that tau phosphorylated at other sites, such as threonine 217 (p-tau217), are comparable [
3,
4] and, in some studies, more accurate in detecting these alterations [
5]. Biomarker modeling studies suggest that the abnormal release of these species occurs early in the disease process—before neurodegeneration and symptomatic disease [
6]. Instead, the alterations in soluble tau co-occur with early accumulation of amyloid-β (Aβ) [
3] and before tau positron emission tomography (PET) becomes abnormal [
7,
8]. It is only rather late in the disease process, close to clinical disease onset, that significant tau aggregation can be seen by tau PET outside medial temporal brain regions [
9]. Taken together, this may at least partly explain the fact that current fluid biomarkers of tau pathology present relatively strong associations with amyloid PET but more modest associations with tau PET. Furthermore, changes in the proteolytic processing of tau is likely important for its tendency to aggregate [
10]. In line with this, tau exists as fragments [
11‐
13], rather than as an intact protein, both in tangles and in the CSF. The importance of tau proteolysis for the propensity of the peptides to become hyperphosphorylated and subsequently aggregate was demonstrated in a study where asparagine endopeptidase (AEP, also called legumain) was shown to cleave tau at amino acid 368 (N368) in an age-dependent manner [
14]. This fragment—which lacks the C-terminal tail of tau—resulted in tau aggregation and phosphorylation as well as neurodegeneration and was present in neurofibrillary tangles in AD brains [
14]. To test the hypothesis that C-terminal tau in CSF reflects tangle pathology, we developed an assay targeting tau368. We showed that it decreased across the AD continuum and correlates with tau PET signal, especially when used in a ratio with t-tau [
15]. Thus, in this study, we aimed to characterize how tau368/t-tau reflects pathological neurofibrillary tau deposition as determined by PET in a larger cohort, to investigate how it relates to cognitive function, and how it can be used in differential diagnostics. Furthermore, we explored if these features are different when comparing to fluid biomarkers that are currently used to index tau pathology (t-tau, p-tau217 and p-tau181).
Discussion
In this study, we show that tau368/t-tau is associated with AD pathophysiological and clinical features in the symptomatic phase of the disease, not entirely captured using established CSF biomarkers indicative of tau pathology. We demonstrate that by normalizing tau368 to t-tau, there is a stronger relationship with uptake of a second-generation tau PET tracer in individuals with symptomatic AD as compared to using either of the biomarkers in isolation, especially in limbic and isocortical brain regions. In contrast, in the group of CU individuals (encompassing young, CU– and CU+), no significant associations were observed between tau368/t-tau and tau PET (all p > 0.05), while several significant associations were seen for p-tau181, p-tau217, and t-tau, further suggesting that these biomarkers capture different aspects of tau pathology.
Further, we found that the tau368/t-tau ratio better reflects the cognitive performance, and thereby clinical disease severity in patients with symptomatic AD, as compared to p-tau181, p-tau217, and t-tau. Finally, we saw that younger individuals with symptomatic AD had higher tau PET SUVR in limbic and isocortical brain regions as well as lower tau368/t-tau.
As previously described, accumulating evidence supported both by clinical observational studies [
3], as well as experimental studies on humans [
26] and animals [
27] suggest that increasing concentrations of CSF p-tau forms and t-tau associate with the emergence of Aβ pathology [
4]. Used in conjunction with the CSF Aβ42/Aβ40 ratio, they offer excellent diagnostic accuracy and are widely used in clinical settings [
28]. However, as previously mentioned, their correlation with aggregated tau, as indexed in vivo using tau PET, is confounded by their collinearity with amyloid [
29]. In this study, we confirm and extend our previous finding that tau368/t-tau is altered in AD and that it correlates with tau PET in patients with symptomatic AD [
15] in brain regions associated with more advanced tau pathology (corresponding to neuropathological Braak stages III–IV and V–VI), while this association was less clear for p-tau181, p-tau217, and t-tau. The associations with tau PET that are seen for CSF p-tau and t-tau in this, and many other studies [
4,
5], seem to be driven by individuals early in the AD continuum, and as the disease progresses, the variability becomes greater [
4]. This is also consistent with studies showing that p-tau plateaus at later stages of the disease [
30]. Of note, however, tau368/t-tau was not associated with tracer SUVR in the transentorhinal brain regions, which may be explained by the fact that tau accumulation in these stages are less prominent with disease progression, as compared to brain regions affected with more advanced disease, as indicated in a previous study [
20].
The non-linear relationship between tau PET tracer uptake and tau368/t-tau found in a previous study [
15], and also observed here, supports that there is a variable processing of tau in healthy subjects, which is then distorted towards secreting a higher relative abundance of N- to mid-terminal tau, being later shifted to produce more C-terminal fragments (which are more prone for aggregation) as AD progresses. In addition, the correlation found with total cognition exclusively for CSF tau368/t-tau ratio may suggest that it may be reflective of clinical disease stage, rather than state, which is likely to be better captured with CSF p-tau. In agreement, tau PET has been shown to better reflect cognitive performance as compared to CSF p-tau or t-tau CSF biomarkers, as it measures tau aggregates, likely being a downstream event of soluble tau secretion [
31‐
33]. Thus, biomarkers of soluble tau release are generally considered to poorly reflect cross-sectional degree of cognitive impairment [
34]. In addition, previous studies have indicated that decreased CSF tau368/t-tau increases concordance between CSF and PET status (more cases labeled as both CSF-positive and tau PET-positive) [
35] and that it may correlate more strongly with tau PET tracer uptake as compared to p-tau181 and t-tau [
15]. This may be due p-tau and t-tau rather capturing the rate with which soluble tau species are secreted at a certain time-point. As an analogy, the change in Aβ homeostasis, which occurs in the AD pathogenesis, is better accounted for when the concentration of Aβ42 is normalized to Aβ40, thus accounting for interindividual differences in Aβ production and clearance [
28]. In a similar manner, normalizing CSF tau368 to t-tau may lead to a better marker of tau aggregates, by correcting for shifting proteolytic processing and secretion of tau, with N- to mid-terminal tau truncated fragments being released into the CSF, and that more C-terminal species containing the aggregation-prone microtubule-binding region are retained in the core of tangles [
36], and thus becoming relatively scarcer in AD. However, the concept of targeting alterations in tau processing as a fluid biomarker may not only have implications for AD, as a recent study showed that tau peptides in the microtubule-binding region (MTBR-tau
275 and MTBR-tau
282) when used in a ratio with a t-tau-like peptide were able to discriminate individuals with corticobasal degeneration (CBD), certain frontotemporal lobar degeneration (FTLD)-
MAPT variants and AD, when compared with other tauopathies and clinical mimics [
37]. The opportunity of more accurately stage tau pathology both in AD and other tauopathies in an affordable manner, both for clinical management and for clinical trial enrollment, as demonstrated in the donanemab phase 2 trial in AD, would be of great use. In that trial, only patients with tau pathology which could be considered as moderate as indexed by tau PET SUVR in an AD-related topographic distribution were included [
38].
Further, as mentioned, younger individuals with symptomatic AD had higher tau PET load in brain regions being affected later in the disease process as well as lower tau368/t-tau, whereas this was not the case for p-tau181, p-tau217, and t-tau. Previous studies have found that neocortical tau PET tracer uptake is higher in younger individuals with AD [
39,
40], further highlighting that tau368/t-tau is captures an aspect of tau pathology not well-reflected using only core CSF biomarkers. This possibly reflects the more aggressive disease course which has been observed in younger patients, as suggested by previous studies [
41,
42]. As we believe that this finding was reflecting a pathophysiological feature of AD in younger individuals, we did not age-adjust our analyses.
There are limitations to this study. First, a relatively low number of participants with symptomatic AD were included, precluding firm conclusions regarding the changes of tau368/t-tau in relation to tau pathology across brain regions, as detailed in Fig.
3. Further, although [
18F]MK-6240 is a highly sensitive in vivo marker of tau aggregates in general [
9], it is also a measure of tau tangles that would best quantified using neuropathology. Also, the cross-sectional nature of this study prevents us from studying the disease progression of individual patients. A future longitudinal study would be more appropriate to model tau368/t-tau in relation to clinical disease progression (e.g., cognitive worsening and rate of tau accumulation).
To conclude, we found that tau368/T-tau captures aspects of tau pathology in the symptomatic stages of AD, not accurately reflected by the core CSF biomarkers. This could potentially influence how tau pathology is indexed using fluid biomarkers both in clinical settings, as well as in clinical trials.
Publisher’s Note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.