Epidemiological pathology of Tau in the ageing brain: application of staging for neuropil threads (BrainNet Europe protocol) to the MRC cognitive function and ageing brain study

Tau pathology associated with brain ageing appears to form a continuum with that associated with AD. There is a hierarchy of progression that differs from that of amyloid deposition, as defined in staging schemes such as that of Thal et al. [34]. Indeed, NFT formation may precede Aβ-amyloid deposition, with occasional NFTs appearing at quite young ages [6]. Tau aggregation in mesial temporal structures without significant Aβ encountered in older brains has given rise to the term “primary age-related tauopathy” (PART) [7], but this may reflect early tau deposition preceding Aβ-stages, and thus be part of the ageing-AD spectrum [8]. The lack of a discontinuity between age-related tauopathy and the tauopathy described in AD is consistent with subclinical tau and amyloid representing early stages of AD, with increasing likelihood of dementia with increasing burdens of pathology. It is therefore important to understand the different forms of tau pathology in brain ageing, their pathogenesis and relationship to dementia.

Much of the work on tau pathology and pathogenesis has focused on the neurofibrillary tangle, located within the neuronal soma. However, the NFT pathology seen in AD, some tauopathies, and ageing is accompanied by tau pathology within the neuritic compartment in the form of NTs. Given that tau modification affects its ability to regulate microtubule function [26], neuritic-compartment tau pathology may be an important contributor to dysfunction of axons and dendrites, and failure of synaptic support. The contribution of neuritic tauopathy to cognitive function is therefore an important question. The strong correlation observed in this population-representative cohort between BNE NT stage and Braak NFT stage suggests that NT formation in general parallels that of NFT formation, and supports the objective of the BrainNet Europe approach to the simplification of the staging process. The fact that both NTs and NFTs are so overlapped may suggest that their development is synchronous, but our data, which is cross-sectional, do not allow us to determine this definitively. The relationship of tau to dementia is thus to neuronal tau in general, not specifically to either NFTs or NTs. It would be further of interest to determine if NTs and NFTs are occurring within the same populations of neurones. Like NFT staging, we found that NT progression is highly stereotypic so that only ~1 % of cases deviated from the BNE NT staging scheme, and this was due to isolated foci of neocortical NT formation. The BNE NT staging scheme therefore appears to be reliable and relatively easy to apply.

BNE NT stage and Braak NFT stage both increase with age among the participants without dementia, consistent with the known age effects on prevalence of tau pathology over the age-course [6], but this age-effect was not seen within the individuals who had dementia, who had a tendency to be at the higher stages. This age-effect in non-demented individuals was not seen with TSA.

Using a logistic regression model, inclusion of BNE NT stage adds additional information about the dementia status compared with age alone. Although BNE NT stage assigns some cases to a different stage compared with Braak NFT protocol (Table 1), both schemes have a similar effect in dementia models, so that the addition of BNE NT stage does not provide additional information beyond Braak NFT staging. A common assumption in dementia neuropathology regards Braak NFT stages V and VI as being highly predictive of dementia. Comparison of NT and NFT stages show that 10 cases with Braak NFT stages V or VI have a BNE NT stage of IV or less, similarly 5 cases of NT stage V or VI were scored Braak NFT stage IV or less. For either protocol a score of stages V or VI is highly predictive of dementia; only 1 in 20 cases for either protocol would be incorrectly assigned to a dementia diagnosis on this metric. However, as previously reported in the CFAS cohort, between 1 in 3 (BrainNet Europe) and 1 in 4 (Braak) elderly individuals with stage 0-II tauopathy have dementia at death, and approximately 60 % have dementia with stage III-IV pathology using either protocol [22, 25]. Thus, as with NFTs, there is considerable overlap in burdens of NT pathology between those with dementia and those without in this population sample.

A spectrum of astroglial tau-pathology is increasingly recognised as a common feature in brain ageing, distinct from the forms of astrocytic tau, such as astrocytic plaques, tufted astrocytes and globular inclusions that have diagnostic significance for specific tauopathies [15]. There has been limited evaluation of these forms of age-related astroglial tau and a recent consensus paper has suggested a more harmonised approach to its assessment under the umbrella term of ageing-related tau astrogliopathy (ARTAG) [16]. Astroglial tau pathology is common in this aged cohort, most commonly TSA which, as one of the major constituent subtypes of ARTAG, are suggested to be an age-related tauopathy rather than disease-specific. Located particularly in subpial, subependymal and perivascular locations, these have been described in mesial temporal structures especially in ageing brains [19, 29]. Recognised best with AT8-immunohistochemistry, they appear to express 4R-tau and can be variable detected with antibodies to p62. Variable staining with Gallyas silver preparations suggests that some of the tau is fibrillary. Tufted astrocytes, which are a feature of progressive supranuclear palsy, were identified in 10 % of our cases, suggesting that they can be seen in ageing brain. However, they were present as isolated cells, and the morphology of many of these showed more finely granular processes, suggesting that they are more akin to the “astrocytes with finely granular tau immunoreactivity in processes” which have been described in mesial temporal structures in the elderly [15, 17]. We now show that these isolated forms of astrocyte tau pathology are not confined to mesial temporal structures in the ageing brain, whilst their significant association with TSA supports the concept that are part of the constellation of age-related astroglial tauopathy.

We demonstrated that the frequency of TSA increases with ageing in this cohort of older people, among whom more that 75 % were 85 years and over at death. There was, however, only a positive trend with age when participants with dementia and without were considered together, and this had a low odds ratio. Other studies suggest that they are uncommon in younger brains, not included in our study [29]. Thus, whilst age may be an important influence on their pathogenesis, as for neuronal tau, ageing is only a marginal effect within an already aged-population. A fuller study of the relationship to age will require a cohort with a wider ageing spectrum.

Whilst TSA are particularly prevalent in mesial temporal areas, we further show that they develop around the brainstem, which is an important site in the early development of AD pathology [33], and less frequently in neocortical areas. Although not formally assessed in this study, they are also very common in a subpial location beneath the substantia innominata and in the semilunar and ambient gyri. Thus they are particular prevalent at basal surfaces of the brain. These are also areas with early NT and NFT formation. Our previous study, restricted to the mesial temporal area, did not demonstrate a relationship between TSA and Braak NFT stage [19]. However, this study, which takes account of three brain areas, does demonstrate weak correlation of TSA with both BNE NT stage and Braak NFT stage, suggesting that TSA formation may be loosely associated with neuronally-derived tau pathology.

TSA were significantly associated with tau in subpial neurites, right at the brain surface. These, along with subependymal tau, are in neurites in close proximity to CSF. Indeed some tau appears to be present within ependyma. The cellular location of this staining remains to be determined, but neuronal processes are described that contact ventricular CSF [35]. It could be hypothesised that this represents a disposal pathway for aggregated tau. The presence of corpora amylacea in this zone might also support this as a disposal route. An alternative possibility is that tau at this location might represent an early site of induction in response to CSF factors (Personal Communication, Udo Rub, Frankfurt, Germany). In either case, the presence of tau at this location is very common and may be important in the pathogenesis of age and AD-associated tau. Astrocytes have been previously shown to take up Aβ-aggregates, contributing to plaque progression and moving these aggregates towards the pial surface. The association of TSA with subpial neuritic tau raises the possibility that they result from astrocytic uptake of aggregated tau along this disposal pathway. However, in situ induction of tau in astrocytes still needs to be formally excluded.

There are a number of limitations to this study. CFAS is a population-based study. This has advantages for this type of analysis, where an unbiased approach is ideal, but it also imposes limitations compared to clinicopathologically defined case–control cohorts, to which the population-based approach is complementary (reviewed in [32]). For example, dementia in CFAS is defined by a dichotomizing, algorithmic approach. This approach is reliable compared to clinical assessment, is ideal for population studies, and has provided high quality epidemiological data on dementia. [13, 21, 36]. It does, however, lack the more textured approach of detailed neuropsychological assessment that can be employed in well-defined clinic-derived populations, thereby limiting assessment of the relationship of grading schemes to specific aspects of cognitive impairment in CFAS.

The areas chosen were primarily to assess the BNE NT stage, as defined by the BNE protocol, but this may constitute limited sampling for comprehensive assessment of prevalence of tau cytopathologies, including astroglial tau. This is particularly relevant for TSA, which can be focal and which are also seen in subpial basal forebrain areas not assessed in this sampling protocol. Thus TSA prevalence may be underestimated. Semi-quantification of TSA in this study was also limited to presence in one, two or three areas. Assessment of further areas and more quantitative approaches may produce a more graded scaling, allowing more powerful exploration of relationships. Sampling guidelines currently being suggested are likely to have a valuable impact here [16]. Whilst this is a sizeable cohort, the low odds ratios in some analyses (especially for TSA), suggest that larger numbers may better establish some relationships. Furthermore, the CFAS cohort is an aged cohort; assessment in cohorts with a wider age spectrum are required to better elucidate broader ageing relationships. Lastly, although the AT8 antibody is widely used in research and diagnosis, this is a tau prevalence study based on a single antibody recognising a specific tau modification.