Introduction
Tau is a microtubule-associated protein involved in regulating axon integrity and function [
25,
76]. Notably, tau aggregation and deposition are hallmarks of Alzheimer’s disease (AD) as well as numerous other tauopathies [
14,
31,
36,
44,
53]. Pathological modifications and aggregation of tau associate with cognitive decline [
23,
29,
30]. The deposition of amyloid-β (Aβ) in plaques represents the other hallmark AD pathology and likely contributes to neurotoxicity in AD [
45,
53]. Alterations in synapse morphology, synapse loss, and axon degeneration occur early in the progression of AD [
8,
23,
52,
75]. This has led to the hypothesis of a “dying-back” pattern of degeneration, where axon degeneration precedes loss of cell bodies [
42]. The hypothesis that tau pathology begins in the axonal compartment and appears in the somatodendritic compartment afterwards is often suggested, however, a direct analysis of axon enriched layers in the hippocampal formation has not previously been conducted [
29,
46]. Moreover, the amyloid cascade hypothesis suggests that Aβ pathology precedes and induces the accumulation of tau pathology presumably starting in axonal target regions of neurons affected by tau [
34,
35,
69]. Animal models of tauopathy also support the contention that tau deposition occurs in the synaptic and axonal compartments prior to the somatodendritic compartment [
4,
22]. Observations in human tissue describing the appearance of extensive neuropil threads (NTs) before detection of neurofibrillary tangles (NFTs) in neuronal cell bodies within a given neuroanatomical region support this hypothesis, though neuropil threads can represent both axons and dendrites [
29,
66‐
68,
71,
75]. None of these studies assessed the terminal fields and somata of specific pathways in the earliest stages of tau deposition precluding a clear determination of whether axonal pathology precedes cell body pathology in cells.
During the course of tau deposition in disease the tau proteins undergo fairly well-characterized changes including both the morphology of and the post-translational modifications to the tau proteins within the pathology [
5,
11,
18,
33,
56]. Importantly, there are specific changes that occur during the earliest detectable deposition of tau inclusions in human brains. These are often referred to as pretangle markers for their ability to recognize tau pathology prior to its maturation and coalescence into compact tau inclusions [
5,
12,
13,
15,
18]. For example, AT8, a triple phosphoepitope including phospho-S199/S202/T205, appears early in diffuse granular pretangle inclusions in neurons [
9,
32]. Additionally, conformational display of an N-terminal region known as the phosphatase-activating domain (PAD), a change recognized by the TNT2 antibody, occurs in pretangle neurons in AD and several tauopathies [
18,
19]. Importantly, AT8 causes an extension of the N-terminus of tau away from the microtubule binding repeats [
38], and both AT8 and exposure of PAD are linked to a specific mechanism of tau toxicity involving impaired axonal function (i.e. axonal transport inhibition) [
40,
41,
48]. Though these markers are modifications of tau that appear in the earliest detectable tau inclusions, it remained unclear whether these pathogenic forms of tau first appeared in axons before progressing to the neuronal cell bodies in humans.
The hippocampal formation comprises the entorhinal cortex (EC), dentate gyrus (DG), hippocampus proper (subdivided into CA1, CA2, CA3, and hilus), subiculum, presubiculum, and parasubiculum [
2,
3]. The EC receives neocortical input and projects through the angular bundle and perforant path to terminate in the molecular layer of the DG [
1,
3]. The DG granule cell projections, known as mossy fibers, terminate in the CA3 stratum lucidum layer (Str. Luc.). Next, CA3 pyramidal cell projections, known as Schaffer collaterals, terminate in the CA1 stratum radiatum layer (Str. Rad.). A majority of the CA1 pyramidal cell projections terminate in the subiculum, however, some also project back to the EC. Finally, most subiculum projections pass back through the angular bundle to the EC to complete the circuit [
2,
3]. The well-defined intrahippocampal circuitry and relatively distinct strata provide an ideal structure to analyze the compartmental progressive deposition of pathological tau within discrete neuronal pathways in post-mortem human tissue.
Tau pathology in the form of NFTs and NTs follows a regional progression in severity, that was described by Braak and Braak in the early 1990’s (i.e. Braak staging) [
12,
14]. Braak staging was originally developed based on silver staining of tau pathology and later adapted to AT8+ pathology, and the primary focus was on the emergence and distribution of NFTs and NTs, and specifically excluded the tau pathology within neuritic plaques (NPs) [
12,
14]. Deposition of tau inclusions begins in the transentorhinal and EC at Braak stages I-II, which is not associated with cognitive decline [
30,
50] or NFT pathology in the hippocampal pyramidal cells [
12,
14]. The limbic stages (Braak III-IV) display spread of tau pathology into the hippocampal formation, initially including primarily the CA1 region, but it is not until stages V-VI that the entire hippocampal formation is affected [
12,
14]. Cognitive decline occurs at these stages and patients may display criteria for mild cognitive impairment (MCI), a prodromal stage of AD [
5,
28,
61,
62,
74]. Finally, the isocortical stage (Braak V-VI) displays extensive tau pathology throughout the hippocampal formation and subdivisions of the cerebral cortex. In the presence of threshold densities and distributions of neuritic and amyloid plaques, Braak stages III and IV are termed “intermediate” Alzheimer’s disease neuropathologic change (ADNC) while Braak stages V and VI are designated as “high” ADNC [
57]. Although a “high” level of ADNC does not predict, with 100% certainty, the presence of dementia, it has been accepted by an international panel of neuropathologists [
57] that “intermediate” and “high” ADNC are sufficient to cause cognitive impairment and data suggests the probability of Braak V or VI subjects having dementia is > 95% [
64]. In addition, recent neuropathological examination suggested that brains containing AD-like NFT pathology, but without detectable Aβ pathology represent a condition termed primary age-related tauopathy (PART) [
21]. PART is typically associated with cognitive status ranging from no impairment to MCI.
The current study addresses the hypothesis that tau pathology first deposits in the axonal compartment of neurons prior to its appearance in the somata. Using post-mortem human hippocampal sections from non-demented (ND) controls and MCI cases, the extent of local axonal and somatodendritic tau pathology (AT8 phosphorylated and PAD exposed), as well as local Aβ pathology in the CA3-Schaffer collateral and DG-mossy fiber pathways was measured. We found AT8 phosphorylation and PAD exposure occurs in the axon compartment of affected neurons even in the absence of observable cell body pathology. Additionally, these tau pathological modifications were observed in the absence of amyloid plaques. Interestingly, separation of cases into PART and non-PART groups revealed that PART cases have significantly less AT8 and TNT2 pathology in these two intrahippocampal pathways. Overall, our results support the hypothesis that tau pathology may begin in the axonal compartment and is observed independently of Aβ plaque deposition.
Discussion
A long-held hypothesis in the field is that pathological tau deposition in disease begins in axons and progressively shifts retrogradely to the somatodendritic compartment [
6,
14], but to our knowledge, no evidence from relatively discrete intrahippocampal neuronal pathways was available previously. The present study characterized the localization of mossy fiber and Schaffer collateral pathway tau pathology in a cohort of ND and MCI human cases. Our focus on relatively discrete neuronal pathways within in the hippocampus provides an opportunity to dissect the spatial changes that occur in the axonal and cell body compartments of neurons with some degree of specificity. Moreover, the use of Braak I-III cases with a range of local pathology load was instrumental in uncovering the first detectable deposition of tau inclusions in cell bodies the hippocampal pathways analyzed, which is reflected in the lack of local pathology and sparseness of pathology in several cases. Two early pathological markers in tauopathies, AT8 phosphorylation [
9,
12,
32] and PAD exposure (TNT2) [
18,
19], appeared in the axonal compartment of these pathways, even in cases where no cell body pathology was observed. Additionally, there was a strong correlation between local axonal pathological staining and the local number of stained cell bodies. Local AT8 and TNT2 positive neurite pathology also correlated well with overall indices of AD-related pathology such as Braak staging, global tangle density and NIA-Reagan level (particularly in the Schaffer collateral pathway). It is noteworthy that other hippocampal formation pathways such as the CA1 projections and EC-perforant pathway were not usable for the purposes of this study because both cell body and axonal pathology already existed in all cases. Taken together, these results support the hypothesis that tau pathology, at least in the studied pathways, is first observed in the axonal compartment and subsequently progresses into the somatodendritic compartment.
The purpose of this study was specifically to capture the earliest possible signs of pathological tau deposition within well-defined circuits to establish whether there is a difference in the temporal appearance of tau pathology in the axonal or cell body compartments of affected neurons. While the results indicate that PAD exposed (TNT2) and AT8 pathological changes occur early in disease progression, in post-mortem human tissue studies we cannot rule out the possibility that the tau deposition here is independent of a progressive condition that would have definitively converted to AD. Indeed, cohorts of MCI patients from previous studies clearly indicate that some patients do not ultimately convert to AD [
62]. More recently, AD-like limbic tau pathology in the absence of Aβ pathology was termed primary age-related tauopathy (PART) [
21], a condition typically associated with no to mild impairments. Here, we assessed the local accumulation of AT8 and TNT2 pathology in the DG-mossy fiber and CA3-Schaffer collateral pathways of the hippocampus. When cases were separated into potential PART (i.e. no Aβ patholoy) or non-PART (i.e. with Aβ pathology) groups the majority of measures found significantly more tau pathology in non-PART cases. While these findings could suggest Aβ exacerbates the deposition of these tau pathologies in these specific intrahippocampal pathways it is important to note that the non-PART group contained substantially more Braak stage III cases compared to the PART group. The only additional notable distinction between PART and non-PART groups was the lack of AT8+ DG neurons in PART cases suggesting DG neurons may be relatively spared from tau deposition in the early stages of PART progression.
Similar to synaptic loss [
8,
23], the extent of total tau burden within the temporal lobe and hippocampus in neuroimaging and neuropathological studies shows a high degree of correlation with cognitive decline [
30,
51,
70]. Interestingly, MMSE scores did not significantly correlate with any of the measures of local AT8+ and TNT2+ pathology in the DG-mossy fiber or CA3-Schaffer collateral pathways. It is likely that the extent of pathology in these discrete hippocampal strata represents the earliest possible stage of detection and thus below the threshold for an association with overt functional impairment, which is relatively mild in this cohort of cases. Case selection for this study specifically excluded those with dementia, limiting the range of MMSE scores available for correlation. The more extensive tau burden within temporal lobe and medial temporal lobe structures that contain a larger tau burden (e.g. the entorhinal cortex-perforant pathway) are likely better indices of cognitive decline in these early prodromal AD stages. This aligns with findings that other temporal lobe and medial temporal lobes are more severely affected when compared to CA3 and dentate neurons [
27].
Synaptic loss and axon dysfunction begin in the prodromal stages of AD (i.e. MCI) and continue as AD progresses and these neuropathological events also occur in other tauopathies [
8,
23,
66‐
68]. The high level of colocalization between AT8 and TNT2 with the axon-specific antibody SMI-312, but not the dendritic marker MAP2, further supports our conclusion that we evaluated predominantly axonal pathology. Additionally, several tau-positive neurites were not clearly colocalized with either SMI-312 or MAP2 making their origin difficult to determine. This could be the result of technical problems such as poor multi-labeling in the relatively small neurites containing densely packed tau inclusions. Alternatively, the loss of SMI-312 or MAP2 could be due to biological processes including degradation of cytoskeletal components such as neurofilaments that occurs during axonal degeneration [
16,
17]. Though these issues preclude definitive identification of all tau-positive neurites in the tissue our data confirm that many of the neurites are axonal in origin (as indicated by the SMI-312 marker) within these specific hippocampal regions.
The presence of both TNT2 and AT8 positive tau pathologies in axons early in disease has important functional implications. The N-terminal PAD domain of tau (i.e. amino acids 2–18) was identified as a biologically active motif that when aberrantly exposed inhibits anterograde fast axonal transport in squid axoplasm [
41,
48]. Additionally, the underlying molecular pathway of the axonal dysfunction was PAD-mediated activation of a protein phosphatase 1-glycogen synthase kinase-3β (PP1-GSK3β) signaling cascade [
40,
42,
58,
59]. Our observation that AT8 pseudophosphorylation structurally exposes PAD aligns with prior structural studies using FRET assays [
38], and we previously demonstrated that AT8 tau is toxic to axonal transport [
41]. Thus, the appearance of PAD exposed tau and AT8 tau first in the axonal compartment of affected neurons before the observation of somatodendritic pathologies and the emergence of clinical symptoms further supports the hypothesis that the tau-mediated activation of the PP1-GSK3β signaling cascade may represent a relevant mechanism of neurodegeneration in tauopathies [
42,
60]. Additionally, PAD exposure is a common occurrence across a range of tauopathies beyond AD, such as frontotemporal lobar degeneration [
19]. Several pathological modifications of tau can contribute to PAD exposure, including phosphorylation [
38,
73], oligomerization, and aggregation [
20,
39]. Therefore, our observations that pathological modifications of tau previously shown to cause axonal dysfunction suggests that this may be one of the early tau-based mechanisms of degeneration in AD.
Our data now provide human tissue-based evidence supporting the hypothesis that tau deposition can occur in axons prior to the somata, but the nature of human tissue studies does not clarify whether the axonal tau pathologies are mobile and traverse retrogradely to the somata or are generated locally in each compartment. Originally, tau was thought to be an axon-specific protein [
65], however, follow-up studies clearly found that tau is present throughout the neuron and only somewhat enriched in axons [
10,
49]. Under disease conditions, redistribution of tau from the axon to the somatodendritic compartment is thought to be an important early event in pathogenesis. Additionally, phosphorylation of tau at residues known to alter tau’s conformation in disease are localized to the somatodendritic compartment, including AT8 [
15,
38,
43]. The issue of whether tau pathologies traverse through neurons in any direction is likely going to be difficult to determine from human studies, however, a number of model systems suggest that pathological tau species are mobile within neurons. For example, a study investigating tau localization in neurons found that P301L mutant tau redistributed from the axonal compartment to the somatodendritic compartment in both transgenic mice expressing P301L tau and wild-type rat hippocampal primary neurons [
37]. In addition, rat primary hippocampal neuron cultures exposed to Aβ oligomers showed an increased redistribution of tau to the somatodendritic compartment, and subsequent analyses found an increase in phosphorylation, including at the AT8 epitope, of this somatodendritic tau compared to untreated cultures [
78], indicating the distribution of pathological tau can change in neurons. Although it is of interest to know whether these events occur in humans, independent of knowing the mechanisms of tau spread, the current work places forms of tau that are toxic to axonal function in the axons of neurons prior to the cell body in the early stages of tau deposition.
Notably, we observed local AT8+ and TNT2+ tau pathology independent of MOAB2+ Aβ pathology. The amyloid cascade hypothesis suggests that amyloid pathology occurs first in AD, and that tau pathology is a downstream consequence of Aβ pathology [
35]. We observed evidence to the contrary; AT8 and TNT2 tau pathology occurred in the hippocampus in the absence of Aβ plaque pathology. Our data clearly indicate that overt Aβ pathology is not present despite the presence of pathological tau accumulation in the axon that was quite robust in some cases, thereby demonstrating a disconnection between Aβ plaque deposits and the emergence of tau inclusions. While it is possible that sections adjacent to those analyzed in this study could contain Aβ pathology in these regions, the lack of pathology in 55–84% of cases and the agreement of our results with the known spatiotemporal distribution of amyloid pathology (i.e. occurs first in neocortical regions) and tau pathology (i.e. occurs much later in neocortical areas) suggests the disconnect between tau and amyloid pathology is unlikely a result of our sampling [
12,
14,
72]. The containment of the axonal compartment and cell bodies within discrete hippocampal strata provides a relatively clear demonstration that amyloid pathology in the terminal regions of tau-affected neurons does not necessarily occur in the pathways/regions assessed. Our findings align with previous findings by Braak and colleagues (1994) that very little to no Aβ pathology was present despite the presence of AT8 tau pathology within the transentorhinal/EC regions of Braak stage 0-III cases [
12]. More recently, Lace et al. demonstrated that Aβ pathology was not correlated to AT8 tau pathology in the entorhinal cortex in a cohort of 93 cases ranging from nondemented to AD and Braak stages I-VI [
47]. Additionally, we found that 29.5% of our cases displayed tau pathology without Aβ pathology and Lace et al. found this in 20% of their cases [
47].
Amyloid plaques are thought to precipitate via a sequential process of going from monomeric proteins, to oligomeric species and then fibrillar forms [
63]. The MOAB-2 antibody is a pan-Aβ specific antibody that reacts with monomeric, oligomeric and fibrillar forms of Aβ [
77], suggesting that the tau pathologies observed in these pathways were not associated with pre-fibrillar or fibrillar forms of Aβ pathology. Again, we cannot completely rule out the possibility that some species of Aβ were not effectively detected or do not exist in other hippocampal regions or these specific pathways that contribute to tau accumulation without examining the entire region of interest. Additional studies assessing the time course of the spatiotemporal distribution of various Aβ and tau species may further clarify the relationship between these two hallmark AD pathologies. However, the data presented here directly challenge the proposition that aggregated Aβ triggers tau pathology in the Schaffer collateral and mossy fiber pathways within the hippocampus.
Overall, this study provides strong evidence that two early tau pathological markers, AT8 phosphorylation and PAD exposure, appear first in axons followed by deposition of cell body pathology. To our knowledge, this is the first study to systematically investigate two well-defined pathways within the hippocampus to differentiate between axon enriched strata and the corresponding cell bodies. Furthermore, visualizing these pathologically relevant modifications of tau in ND and MCI human tissue cases provides valuable insight into early tau pathology in the human condition before the onset of AD. Importantly, both AT8 and PAD exposed forms of tau are linked to mechanisms of toxicity in axonal functions such as axonal transport. Coupling these results with the lack of observable Aβ pathology suggests amyloid is not causing the early tau changes in these hippocampal pathways. This raises questions of whether the amyloid cascade is a viable hypothesis to explain the complex nature of AD etiology, but perhaps later in the disease process tau and Aβ pathologies work together to enhance ongoing cell dysfunction and degeneration once the pathologies overlap or interact. Nonetheless, our findings suggest that early axonal tau pathologies may trigger degenerative events in the hippocampal circuitry prior to overt cognitive decline, and more longitudinally focused future studies specifically on connecting the earliest forms of tau pathology in axons and clinical decline are needed.