Soluble pre-fibrillar tau and β-amyloid species emerge in early human Alzheimer’s disease and track disease progression and cognitive decline
- First Online:
- Cite this article as:
- Koss, D.J., Jones, G., Cranston, A. et al. Acta Neuropathol (2016) 132: 875. doi:10.1007/s00401-016-1632-3
- 2k Downloads
Post-mortem investigations of human Alzheimer’s disease (AD) have largely failed to provide unequivocal evidence in support of the original amyloid cascade hypothesis, which postulated deposition of β-amyloid (Aβ) aggregates to be the cause of a demented state as well as inductive to tau neurofibrillary tangles (NFTs). Conflicting evidence suggests, however, that Aβ plaques and NFTs, albeit to a lesser extent, are present in a substantial subset of non-demented individuals. Hence, a range of soluble tau and Aβ species has more recently been implicated as the disease-relevant toxic entities. Despite the incorporation of soluble proteins into a revised amyloid cascade hypothesis, a detailed characterization of these species in the context of human AD onset, progression and cognitive decline has been lacking. Here, lateral temporal lobe samples (Brodmann area 21) of 46 human cases were profiled via tau and Aβ Western blot and native state dot blot protocols. Elevations in phospho-tau (antibodies: CP13, AT8 and PHF-1), pathological tau conformations (MC-1) and oligomeric tau (TOC1) agreed with medical diagnosis (non-AD cf. AD) and Braak stage classification (low, intermediate and high), alongside elevations in soluble Aβ species (MOAB-2 and pyro-glu Aβ) and a decline in levels of the amyloid precursor protein. Strong correlations were observed between individual Braak stages and multiple cognitive measures with all tau markers as well as total soluble Aβ. In contrast to previous reports, SDS-stable Aβ oligomers (*56) were not found to be reliable for all classifications and appeared likely to be a technical artefact. Critically, the robust predictive value of total soluble Aβ was dependent on native state quantification. Elevations in tau and Aβ within soluble fractions (Braak stage 2–3 cf. 0) were evident earlier than previously established in fibril-focused disease progression scales. Together, these data provide strong evidence that soluble forms of tau and Aβ co-localise early in AD and are closely linked to disease progression and cognitive decline.
KeywordsAlzheimer’s disease Amyloid Tau Cognitive decline Dementia Pathology
The original ‘amyloid hypothesis’ postulated that aggregated β-amyloid (Aβ) drives pathological mechanisms, results in the hyperphosphorylation of the microtubule-associated protein tau, and causes neurodegeneration and cognitive decline . Accordingly, much of the early work focused on the deposition of Aβ as senile plaques (SPs) and their relationship to the emergence of tau pathology. This work was supported by genetic mutations identified in familial cases of AD (fAD), and has been greatly influential for both experimental research and therapeutic endeavours. Mutations in the amyloid precursor protein (APP) and the presenilin (PSEN) genes clearly demonstrated that dysfunctions in APP processing are causative of dementia in these cases. Yet, the contribution of specific cleavage products and protein aggregates to AD onset and progression as well as cognitive decline remains highly debated.
A prominent criticism of the amyloid hypothesis has been the lacking association of total plaque load with cognitive status, which is in contrast to the more robust and graded correlation of tau pathology to neuronal loss and symptomatic presentation. Indeed, the highly reproducible regional progression of tau containing neurofibrillary tangles (NFTs) and neuropil threads (NTs) provides the basis for neuropathological severity grading (Braak staging) . Braak staging remains the most accepted post-mortem method of classifying AD progression, although amyloid pathology in the form of SPs and neuritic plaques must also be present to confer a true diagnosis of AD. Overall, NFTs are generally accepted as better indicators of mental impairment, clinical AD symptoms and neurodegeneration over plaque load [26, 53]. Both NFTs and NTs first emerge almost exclusively within the entorhinal cortex (EC) and hippocampal formation (transentorhinal stage, Braak stage 1–2), this pathology alone is associated with little or no cognitive impairment. Only when surrounding cortical and subcortical structures of the limbic system are affected (limbic stage, Braak stage 3–4) do individuals become symptomatic, and definitive symptoms develop fully only once this neuropathology engulfs the neocortex (neocortical stage, Braak stage 5–6).
For the conclusive neuropathological diagnosis of AD, an evaluation of Aβ depositions via complementary neurohistopathological schemes, such as the five SP-based Thal (Aβ) phases and CERAD (Consortium to Establish a Registry for Alzheimer’s Disease) staging must be made . However, when considered independently, plaque deposition is less systematic cf. tau pathology and fails to correlate sufficiently with cognitive impairment . Relative to NFTs, SP deposition follows a different spatial progression and is first observed within the neocortex . Plaque numbers become more widespread reaching the EC and hippocampus long after the emergence of NFTs and NTs, with tau and amyloid pathologies only overlapping in advanced stages . Such a spatial divide between the two pathological hallmarks makes it difficult to reconcile the proposed mechanistic and causative link between Aβ and tau, and cannot easily be incorporated into a common pathological scheme.
Despite this scenario, extensive research has focused on targeting amyloid-related pathologies for diagnostic purposes, e.g. based on PET amyloid imaging . More recently, tau-based imaging markers have emerged with promising initial results . Equally, biomarker research in cerebrospinal fluid (CSF) is on the hunt for multifactorial biomarkers, though assays with high specificity and sensitivity remain elusive. For example, a recent multi-centre study reported an age-independent occurrence rate of ~22 % of healthy controls (n = 1233) in which CSF Aβ levels were low enough to be classified as pathological, yet were negative for neurodegeneration and cognitively intact .
Beyond aspects related to disease diagnosis, neither Aβ plaques nor tau NFTs are likely the principle causative factor, as both of these forms of pathology are detected with considerable frequency in a subset of the non-demented population, referred to as non-demented high pathology controls (HPCs) . Moreover, neurons can likely survive with NFT inclusions for several decades  and a separation of NFT formation and cognitive decline has been demonstrated in animal models . Recent adaptations of the amyloid cascade hypothesis detail critical interactions with tau pathology, mechanisms of early synaptic loss and highlight central roles of soluble pre-fibril Aβ and tau species .
This shift away from insoluble pathology to potentially earlier emerging soluble Aβ and tau species required advanced biochemical techniques to establish an understanding for the progression of pre-fibrillar pathology. Significantly, a staging of Aβ pathology in line with disease progression has been proposed based on the abundance of low molecular weight Aβ species in cellular fractions, resolved by SDS-PAGE, and the degree to which these species demonstrate biochemical modifications . The proposed stages correlate with established neuropathological grading scales, such as Braak and CERAD scores, to some degree, suggesting the potential for a stereotypical evolution of molecular pathology. However, associations with cognitive readouts and specific Braak stages were not included in this study. Therefore, a detailed evaluation of potentially pathology relevant soluble tau and Aβ species and their respective association with cognitive decline and disease progression within human cases identified as early and late-stage AD is still lacking.
Here, lateral temporal lobe tissue lysates (Brodmann area 21) from human AD cases and non-AD controls were quantified for phosphorylated, conformationally altered and oligomeric tau species alongside amyloidogenic changes in APP metabolism and levels of total, monomeric, oligomeric and pyro-glutamate (pyro-glu) modified Aβ species. Markers were assessed relative to several classifications, i.e. based on clinical diagnosis, grouped Braak stages (Br 0–2 cf. 3–4 cf. 5–6) as well as individual Braak stage, and correlated with each other as well as with cognitive decline. Our data demonstrate the earlier coincidence of soluble tau and Aβ and support the potential pathological role of both proteins early in the disease process of AD.
Demographic data of human cohort and categories for data pooling
Age range (years)
Mean age (years)
PMI range (h)
Mean PMI (h)
86.04 ± 1.4
44.91 ± 5.2
6.25 ± 0.09
83.58 ± 1.15
46.58 ± 4.9
6.34 ± 0.08
Grouped Braak stages
85.28 ± 1.88
40.67 ± 6.2
6.23 ± 0.08
85.86 ± 1.45
51.54 ± 6.9
6.22 ± 0.16
83.86 ± 1.39
46.00 ± 5.7
6.38 ± 0.08
76.67 ± 1.33
30 ± 13.45
6.05 ± 0.05
87 ± 1.96
42.8 ± 7
6.32 ± 0.09
87.56 ± 1.71
53.4 ± 9.5
6.2 ± 0.26
82.8 ± 2.22
48.2 ± 10.6
6.24 ± 0.19
83.8 ± 1.2
50.5 ± 10.1
6.28 ± 0.09
83.89 ± 2.12
42.6 ± 6.9
6.5 ± 0.14
To serve as positive and negative immunoblot controls, murine forebrain samples from several transgenic strains were used (12-month old wild-type mice (C57/BL6) and 12-month-old bigenic fAD mice, as previously described ). Brain lysates from 12-month-old BACE1−/− mice were a generous gift from Prof Michael Ashford (see ). Mice were housed and sacrificed in accordance with UK Home Office regulations, University European Directive on the Protection of Animals used for Scientific Purposes (2010/63/EU) and the Animal (Scientific Procedures) Act 1986.
Immunoblot quantification of AD markers
Quantification of Aβ and tau markers was based as far as possible on near-native state preparations to limit the degree of sample manipulation required. Therefore, most markers were assessed by dot blots, with the exception of Western blots used for 6E10 and BACE1 antibodies to ascertain detection of single and relevant protein species. The oligomeric tau antibody TOC1 was also validated in Western blots to confirm detection of the previously reported oligomeric species, before samples were fully characterised via dot blots.
Brain lysate preparation
100 mg of frozen cortical tissue was homogenised in ~1:10 (w/v) Igepal (Sigma, Dorset, UK) based lysis buffer (in mM: 20 HEPES, 150 NaCl, 0.1 EDTA, 1 % Igepal: pH = 7.6). All buffers were supplemented with complete protease inhibitors (Roche) and PhosStop tablets (Roche). The use of the non-ionic, non-denaturing Igepal, which is chemically indistinguishable from the widely used ‘Nonidet P-40’, ensured adequate lysis of plasma, endoplasmic and Golgi but not nuclear membranes, and prevented aggressive solubilisation of large aggregates . Following manual homogenisation, samples were spun (13,000g, 4 °C, 20 min), supernatants separated from pellets, aliquoted and stored at −80 °C. For the determination of aggregated non-soluble protein pathology, Igepal derived pellets were re-suspended in excess Igepal buffer (1 ml) homogenised via repeated aspiration with a 1 ml pipette tip, briefly vortexed and spun (13,000g, 4 °C, 20 min). Following removal of the supernatants, the process was repeated. Resulting pellets were subsequently re-suspended in 1:1 (w/v) 70 % formic acid, incubated overnight at 4 °C with continuous agitation before a final spin (18,000g, 4 °C, 20 min) to yield the collected supernatant, which was stored at −80 °C until use.
SDS-PAGE electrophoresis and protein transfer
Soluble lysates were adjusted for protein concentration (3 µg/µl) as per bicinchoninic acid colorimetric protein assay (BCA, Sigma). For standard denaturing SDS-PAGE electrophoresis, samples were mixed with lithium dodecyl sulphate (LDS, Nupage, Thermo Fisher Scientific, Paisley, UK) and dithiothreitol (15 mM DTT, Sigma) before being boiled for 10 min at 70 °C. For the identification of oligomeric tau species, samples were treated as above, except without DTT or boiling prior to gel loading. Insoluble samples were first mixed with 4 volumes of neutralising buffer (2 M Tris + 2 M NaH2PO4) before the addition of LDS and DTT.
Samples (30 µg/lane for soluble and 10 µl/lane for neutralised insoluble samples) were separated on 4–12 % Bis–Tris gels (Nupage, Thermo Fisher) under variable conditions dependent on protein target. For Aβ detection, proteins were separated for 35 min at 200 V constant voltage in MES buffer, transferred to 0.2 µm nitrocellulose membranes via IBlot (Nupage, Thermo-Fisher), and boiled for 3 min in 0.01 M Dulbecco’s Ca2+ and Mg2+ free phosphatase buffered saline (VWR International, Leicestershire, UK). For tau blots, electrophoresis was conducted for 45 min at 200 V in MOPS buffer and transferred to 0.45 µm nitrocellulose membranes via wet transfer conditions (1 h, 25 V constant).
Solubility and protein extractions
The specific extraction of non-fibrillar proteins from human tissue post-mortem was confirmed in a series of comparative Western blots (Supplementary Figure 1). Total tau, visualised by the HT-7 antibody, demonstrated the selective isolation of tau within well-defined protein bands following Igepal lysis. In contrast, subsequent treatments of the remaining insoluble pellet with 70 % formic acid presented a smear of HT-7 immunoreactivity. This was only evident in AD cases and is indicative of highly post-translationally modified and aggregated tau species as found in paired helical filaments and NFTs. Similarly, limited solubilisation of amyloid species was observed via Igepal lysis, when compared to that isolated via formic acid from the pellet (see Supplementary Figure 1).
Native state immunoblots
To ensure the preservation of potentially heat- and conformation-dependent epitopes, dot blots were conducted for a number of Aβ and tau pathological markers. Igepal soluble lysates were adjusted to 2 µg/µl with distilled H2O (concentration determined with BCA assay, as above) and directly dotted onto 0.2 µM nitrocellulose membranes (5 µl/samples, 10 µg/dot), which were allowed to dry prior to further processing. In a subset of experiments, designed to determine the heat-sensitivity of specific epitopes, lysates were heated (70 °C for 10 min) prior to being dotted.
Blocking and antibody detection
All immunoblots were washed in 0.05 % Tween 20 (Sigma) containing Tris-buffered saline (TBST, in mM; 50 Trizma base, 150 NaCl, pH = 7.6) and blocked for 1 h at RT in 5 % milk powder containing TBST prior to overnight incubations at 4 °C with primary antibodies (see Supplementary Table 2) under continuous agitation. Antigen–antibody coupling was visualised using appropriate secondary antibodies conjugated to horseradish peroxidase (goat anti-mouse IgG or IgM, goat anti-rabbit IgG; 1:5000, Merk Millipore, Watford, UK) and enhanced chemiluminescence (1.25 mM Luminol, 30 µM coumaric acid, 0.015 % H2O2). All steps were followed by multiple washes in TBST. Processed blots were subsequently stained for measurements of total protein using Coomassie total protein stain (as described previously ). Images were captured via Vilber-Fusion-SL camera (Vilber, Eberhardzell, Germany) at 8-bit for illustration and 16-bit for analysis.
To ensure specificity of the detected signal, appropriate secondary antibody control dot blots were conducted, in which the primary antibody was excluded. For all cases, no detectable immunoreactive signal was observed for either IgG or IgM secondary antibody isotype in images captured following exposure times of up to 3 min, which was the maximal exposure time required for signal detection in this study (data not shown).
Immunoreactivity was quantified from 16-bit digitised images using ImageJ (Ver 1.47, NIH, USA) based on area under curve (AUC) measures normalised to total protein load established via AUCs for Coomassie processed blots. Immunoblot intensity data were normalised within blots using total protein adjusted values and expressed relative to appropriate control groups (outlined below), before being pooled across blots.
Post-mortem confirmed clinical diagnosis, essentially equating to non-AD cases (=Braak (Br) 0–3) and AD (=Br 4–6).
Grouped Braak stages: low, intermediate and high severity of Braak pathology (Br 0–2, Br 3–4 and Br 5–6, respectively).
Individual Braak score: cases listed per individual stages and normalised to a single Braak score (Br 2).
Statistical analysis was performed using Prism (V.6, GraphPad). Data were subject to Shapiro–Wilk test for normal distribution prior to statistical analysis of significance between groups. Individual comparisons were analysed using either a Student’s two tailed t test or Mann–Whitney if determined non-parametric. Multiple group comparisons were established via 1-way analysis of variance (ANOVA) or the non-parametric Kruskal–Wallis ANOVA, if reported as significant, selected pair comparisons were conducted via post hoc Bonferroni or Dunn’s tests, respectively. Analysis of covariance (ANCOVA) was employed to control for PMI and tissue pH (Minitab 17). Correlation analysis was conducted using Spearman’s ranks correlation, which does not assume normality of data sets. Having established either positive or inverse correlation of markers in relation to Braak stages, individual Braak stages were probed for significant deviation from Braak 0 via a one-tailed t test. For all tests, p < 0.05 was taken as significant with subsequent levels of statistical reliability reported for p < 0.01, p < 0.001 and p < 0.0001.
Demographic data and cognitive scores
Quantification of tau pathology
Tau phosphorylation was further investigated using two other commonly employed phospho-tau antibodies, PHF-1 (Fig. 2b, p-ser396/ser404) and CP13 (Fig. 2c, p-ser202). As for AT8, PHF-1 and CP13 reactivity was increased in AD compared to non-AD cases (Fig. 2bii + cii, 9- and 13-fold increase for PHF-1 and CP13, respectively, p < 0.0001 for both). Both PHF-1 and CP13 levels also increased with pathology severity (Fig. 2biii + ciii, p < 0.0001), reporting elevations within intermediate and high Braak stages compared to low pathology cases. Also, strong correlations with individual Braak stages were observed with PHF-1 and CP13 (Fig. 2biv + civ, r = 0.80, p < 0.0001 for PHF-1 and r = 0.77, p < 0.0001 for CP13). For the three phospho-epitopes, CP13 changes were detected earliest (Br 0 cf. Br 2: p < 0.05) and PHF-1 latest (Br 0 cf. Br 5, p < 0.05). Phospho-tau dot blots were validated for specificity by means of correlation of immunoreactivity with traditional Western blots in a subset of cases for PHF-1 (r = 0.79, p < 0.0001, n = 18, data not shown). A significant effect of Braak stage (F(5,80) = 10.47, p < 0.0001) and epitope (F(2,80) = 8.61, p < 0.001) and an interaction (F(10,40) = 2.43, p < 0.05) was reported when comparing the phospho-epitope specific antibodies, likely due to the variable magnitude of hyperphosphorylation and differential antibody affinities apparent between Braak stages 2–4. Nevertheless, further analysis demonstrated a robust effect of subject matching, indicating those cases with high AT8 staining also demonstrated high levels of CP13 and PHF-1 staining. This is perhaps unsurprising given the similarity of dot blot immunostaining between epitopes (see example blots in Fig. 2).
Conformational and oligomeric tau
Although hyperphosphorylation is the most extensively studied aspect of tau pathology, several additional protein modifications offer further key parameters in disease pathology. Conformational changes in the natively unfolded structure of the tau protein, leading to an interaction of the N-terminal domain with the microtubule binding domain (MTBD), may be closely modulated by phosphorylation status .
In addition to the emergence of conformational tau pathology, formation of oligomeric tau species has been proposed to correlate with behavioural deficits in animal models  and was also previously reported to be elevated in human AD cases [40, 64, 79]. TOC1 binds to a conformation-dependent epitope preferentially exposed upon oligomerisation (aa209–224) , and here recognised a single band when tested in Western blot applications under non-denaturing conditions (no DTT or boiling, ~180 kDa; Fig. 3c i, for full blot see Supplementary Figure 2). This band was previously established via SELDI-TOF MS as a tau dimer . TOC1 immunoreactivity was characterised in all cases via native state dot blots (Fig. 3cii) and robustly increased for AD compared to non-AD cases (Fig. 3ciii, p < 0.0001), significantly tracked across grouped Braak stages (Fig. 3civ, p < 0.0001) and correlated with individual Braak stage (Fig. 3cv, r = 0.59, p < 0.0001). Relative to Braak stage 0 a significant elevation in reactivity emerged at Braak stage 5 (p < 0.05), although in a subset of cases a strong signal prior to this was apparent (Br 4 cf. Br 0, p = 0.07).
Correlations of tau biomarkers
Correlations of tau pathology markers
APP processing and pathology
The reduction of fAPP with diagnosis and disease severity may suggest a facilitation of APP cleavage via amyloidogenic secretases in AD cases. Several groups have previously reported a disease-dependent increase in BACE1 and thus a promotion of Aβ production pathways [24, 34]. However, despite the decrease of fAPP, BACE1 expression levels were unaltered in any of the analytical stratifications (Fig. 4b; p > 0.05 for all).
Next, cases were further probed for post-translational modification of Aβ species with pyro-glu . In Western blots, a single immunoreactive band was identified migrating ~12 kDa, which in a subset of samples approached significant elevations based on diagnosis and when samples grouped in low, intermediate and high Braak stages (p = 0.07 and p = 0.08, respectively), but did correlate with individual Braak stages (r = 0.32, p < 0.05; data not shown). When all cases were characterised in non-denaturing immuno-dot blot conditions, pyro-glu Aβ immunoreactivity was enhanced based on diagnosis criteria (Fig. 7b i, p < 0.05), increased in line with grouped Braak stages (Fig. 7b ii, p < 0.05) and correlated with individual Braak stage (Fig. 7b iii, r = 0.32, p < 0.05). Post hoc analysis indicated that significance was largely driven by late-stage pathological changes (Br 5–6 cf. Br ≤ 2, p < 0.05), yet no individual Braak stage was significantly elevated from Br 0 cases. As can be seen in scatter plots in Fig. 7b, significance was primarily driven by two highly reactive AD cases, however, further investigation (Grubb’s outlier test) did not report these as significant (p > 0.05).
Correlation between Aβ-related biomarkers
Correlations between amyloid pathology markers
Tau and Aβ: mutual correlations
Correlations between tau and amyloid markers
Pathological correlates of Braak staging and cognition
Correlation of tau and amyloid markers with Braak stage and cognitive scores
Quantification of soluble, abnormally phosphorylated tau detected in brain lysates was highly predictive of the clinical diagnosis and rose with the severity of neuropathological classifications, regardless of the phospho-epitope. In agreement with the stereotypical progression of NFTs over the course of AD, phosphorylation at the AT8 epitope was first significantly elevated at Braak stage 4 [2, 15]. Classically based on the detection of NT and NFTs via argyrophilic stains, modern Braak staging utilises AT8 and thus includes the additional detection of pre-tangle tau. The accumulation of soluble, phosphorylated tau is evident within the temporal cortex and precedes mature tau lesions in many low Braak stage (I–III) cases . Here, tau species detected in our soluble fraction most likely reflect changes within pre-tangle neurons rather than NF and NFTs, and already appeared at Braak stage 2 (CP13). All three phospho-tau antibodies strongly correlated with Braak stages, in close agreement with previous reports . This was nevertheless surprising given the well-documented post-mortem reduction in phosphorylation , largely due to the energy-independent activity of phosphatases. Although samples used in the current study had longer PMIs compared to those from some previous publications, no impact of PMI as covariate or correlation of any marker with PMI was observed. Moreover, no evidence for protein degradation was detected in either blotting techniques. Ultimately, the long PMIs may have resulted in an under-estimation of pathological tau phosphorylation but strong correlations of phospho-tau markers with other pathological markers as well as cognition provide evidence for reliable disease tracking.
The rise of tau phosphorylation at Braak stage 2 was mirrored by the conformation-specific antibody Alz-50, although as previously discussed non-specific FAC1 binding may be a confounding factor here . The improved MC-1 antibody [37, 81] demonstrated a superior signal associated with diagnosis, Braak staging and cognitive decline. However, MC-1 reactive tau was not elevated significantly until Braak stage 4, at which point both CP13 and AT8 phosphorylation markers were also raised. Tau’s ability to adopt folded MC-1-reactive conformation may be increased in situ following phosphorylation at multiple sites as has been observed in vitro . Hence, changes in phosphorylation may indeed be required before conformational changes can occur .
Previous studies have shown a robust increase in TOC1 reactivity in AD compared to control samples [40, 63], but have not assessed the association of TOC1 with disease progression and cognitive impairment scores. Thus, an important conclusion from our data is that oligomeric tau appears to be a strong predictor of cognitive scores, clinical diagnosis and neuropathological severity. Recent experimental studies have indeed linked low-weight oligomeric tau rather than fibrillary species to e.g. increased neuronal toxicity, impaired axonal transport, inhibition of synaptic plasticity, synaptic loss, mitochondrial impairment and memory deficits [22, 46, 47, 63, 64]. Overall, our data lend direct support to the patho-physiological relevance of non-filamentous, oligomeric tau. Nevertheless, strong correlations with cognitive decline were also observed for conformational and phospho-tau as previously reported in humans [25, 30] and animal models [43, 44, 83]. Our data expand on the suggested disruptive role of monomeric phospho-tau , and suggest oligomerization caused by pathological modification of monomeric tau to play a key role. Factors causing the formation of oligomeric tau species remain largely undefined, but native tau does not appear to readily self-aggregate under near-physiological conditions [20, 42]. Not surprisingly, several post-translational modifications facilitate tau aggregation, and our current work implies facilitation of oligomerisation via multiple phospho-epitopes, consistent with the extensive co-localization between TOC1 and phospho-ser422 tau, an early pre-tangle phospho-epitope , in the EC and cholinergic basal forebrain . Conversely, tau oligomers do not readily co-localise with markers of more mature fibrillary tau species [63, 79]. Taken together, the coincident appearance of phospho-epitopes and oligomerization is supportive of the suggestion that abnormal phosphorylation may facilitate tau oligomerization  thus generating pathological tau entities early in the disease process. Nevertheless, future studies are required to better define which phospho-epitopes are directly involved in this process in situ.
With respect to measures of cognitive ability, all markers except Alz-50 provided robust associations. As we focused here on tau species that are soluble in mild detergent, non-denaturing conditions, our findings strongly suggest that soluble tau is sufficient and relevant for cognitive impairment, and adds support to the proposition that soluble pathological species are more relevant to disease progression as opposed to the NFT and plaque inclusions . Correlations obtained here were clearly stronger than previously reported for NT and NFTs in AD and MCI cases , despite the limited number of cases and varying origin.
APP processing and soluble Aβ species
Early stage alterations in Aβ markers were twofold: a diagnosis-specific, progressive decline in fAPP and a corresponding increase in MOAB-2 reactive Aβ species. The progressive decline of fAPP is in contrast to many other studies, which found either no change [8, 60] or an increase [3, 82]. There are several possible explanations for this inconsistency, but perhaps the most pertinent is the composition of sample groups. None of the previous investigations had quantified APP levels according to individual Braak staging, and here the greatest loss of APP was seen between Braak stage 0 and 2, two stages often pooled into a non-AD control groups. Thus, the early loss of total APP levels may have been overlooked.
The observed reduction in APP may be a consequence of transcriptional downregulation , however, the strong negative correlation between total APP and soluble Aβ species suggests increased amyloidogenic APP cleavage. Whilst no change in the expression levels of BACE1 was observed (in contrast to the investigations of others [24, 34]), we cannot rule out the possibility of increased BACE1 activity in AD [1, 24, 72].
Regarding Aβ species, monomeric, pyro-glu modified and total soluble levels were elevated according to diagnosis and Braak stage severity. These forms of Aβ also robustly correlated with individual Braak stages and both monomeric and soluble levels of Aβ further correlated, albeit to varying degrees, with measurements of cognition. Our data also suggest that detection of monomeric Aβ via Western blots is not always possible, particularly in non-AD cases, but also in a number of AD cases. The apparently low levels of soluble monomeric Aβ agree with recent findings which indicate a potential over-estimation of respective Aβ species following aggressive tissue homogenisation via sonification . Genuine soluble, monomeric Aβ species may in some cases be below the detection limit of conventional Western blotting, even in cases with Braak stage 5 pathology . Detection of Aβ may be further complicated, as even common antibodies such as 6E10 and 4G8 may only bind sub-species of Aβ, dependent on conformation . Regardless of the analytical parameters, monomeric Aβ was elevated in agreement with both diagnosis and Braak stage, closely matching data from others . Yet, a significant increase occurred only in advanced pathological cases and did not follow a progressive scale, in agreement with previous observations .
Our data strongly suggest that chemical and physical parameters of standard Western Blot techniques are not suitable for the reliable detection of native Aβ species. Although numerous potential SDS-stable Aβ oligomers were visible following the overexposure of 6E10 blots, it is far from clear if any of these entities are native to the brain and/or disease relevant. Specifically, this relates to the oligomeric species known as *56, a dodecamer of Aβ, which has been identified in several animal models  including our own , and found to decrease as a function of disease . It has been suggested to be an indicator of a prodromal time window, triggering degenerative cascades before sequestration into increasingly insoluble plaque deposits. Here, such an inverse relationship between disease severity and this oligomeric Aβ species also seemed apparent, declining from early Braak stage cases to high pathology cases. However, the utility of such SDS-stable oligomeric species must be questioned when considering that the presence of SDS, even at low concentrations (0.2 %) can greatly enhanced the abundance of dimeric  trimeric  and higher, multimeric Aβ species [5, 31, 32, 66]. Furthermore, and in line with the impact of heat and reducing agents such DTT to modulate amyloidogenic β-sheet aggregation [13, 45, 69], we demonstrate that the relative abundance of *56 was strongly reduced and any correlation with disease progression lost in the absence of such treatments. This suggests that some of the previous data may be due to technical artefacts. The observation of highest levels of *56 in Braak stage 0 cases is also somewhat at odds with the hypothesis of *56 being pathologically causative, as one would anticipate lowest levels in healthy adults and thus at least an initial elevation between what is considered a pathology-free status (Br 0) and those cases assumed to be at risk of developing AD (Br 2 and 3).
Contrary to SDS-PAGE resolved oligomers and monomers, dot blot detection of soluble unmodified oligomeric Aβ allowed consistent detection and demonstrated a robust disease-specific increase and a progressive correlation with Braak stages and cognitive decline. Critically, this relationship was completely abolished following heating and induced a similar inverse relationship with disease progression, as observed for *56. Together, our data strongly suggest that artificial modification of Aβ species has hampered many previous studies, via promotion of soluble Aβ species into oligomeric configurations which may mask epitopes or destroy binding sites. The heat-sensitivity likely explains the reported failure of MOAB-2 immunoreactivity to detect AD cases from pre-heated basal forebrain lysate following dot blot analysis by others  and thus highlights the need to carefully review both sample preparation and subsequent methods for Aβ quantification.
Even though our approach does not allow the detection of particular Aβ species, MOAB-2 does not cross-react with APP or other metabolites, and it binds a C-terminal epitope with stronger affinity to more toxic Aβ1–42 than to Aβ1–40, independent of aggregation state ). The increase in total soluble levels of Aβ with disease diagnosis is in line with ELISA-based measurements from cortical tissue [19, 50]. Compared to measurements of insoluble Aβ  and plaque load [6, 59, 62], MOAB-2 reactivity provides much better correlation with cognitive decline (ranging from 0.46 up to 0.71) as well as improved correlation with disease progression as established with the biochemical staging of Aβ . Similarly, the immunoreactive levels of soluble Aβ correlated with all tau markers tested (see below). In comparison, strong associations were not observed for pyro-glu Aβ: Although selective for diagnosis, elevations were only apparent between low and high Braak stage cases. Furthermore, pyro-glu Aβ only weakly correlated with individual Braak stages and failed to correlate with any measure of cognition. This is largely at odds with previous reports of disease specificity  and toxicity , although in agreement with previous work where a correlation was established between levels of pyro-glu-modified Aβ and tau phosphorylation .
Early co-localisation of soluble tau and amyloid pathology
It is fairly well established that Aβ plaques precede the emergence of NFTs in the neocortex, while NFTs are formed earlier in limbic areas . However, our findings clearly indicate that initial emergence of tau pathology coincides with increased soluble Aβ in Brodmann area 21 as early as Braak stages 2–3. This pathology therefore precedes the emergence of mature NFTs, classically determined at Braak stage 4 in this brain region. Thus, our data suggest a closer overlap for soluble Aβ and tau species in the temporal lobe, contrasting with the spatial separation of the two toxic pathways based on the mature fibrillary forms of these proteins. Thus, at least in Brodmann area 21, the early co-localization of soluble tau and Aβ pathology implies a close spatiotemporal relationship between these two pathologies.
In summary, our findings are supportive of the coincident emergence of disease-relevant, soluble, pre-fibrillar forms of both Aβ and tau during the earliest stages of AD. Potentially toxic protein species identified here strongly correlate with Braak staging and cognitive decline. Causal and functional links between Aβ and tau pathways will remain a hotly debated topic, but here we provide evidence that both pre-fibrillar Aβ and tau appear as likely contributors to the pathogenesis of AD. Late-stage insoluble protein aggregates may have very little relevance to the pathogenic mechanisms of disease, and therefore targeting pre-fibrillar tau and/or Aβ species will be a more effective therapeutic strategy for AD.
We would like to gratefully acknowledge all donors and their families for the tissue provided for this study. Human tissue samples were supplied by the Brains for Dementia Research programme, jointly funded by Alzheimer’s Research UK, the Alzheimer’s Society and the Medical Research Council, and sourced from the MRC London Neurodegenerative Diseases Brain Bank, the Manchester Brain Bank, the South West Dementia Brain Bank (SWDBB), the Newcastle Brain Tissue Resource and the Oxford Brain Bank. The Newcastle Brain Tissue Resource and Oxford Brain Bank are also supported by the National Institute for Health Research (NIHR) Units. The South West Dementia Brain Bank (SWDBB) receives additional support from BRACE (Bristol Research into Alzheimer’s and Care of the Elderly). Alz-50, CP13, MC-1 and PHF-1 antibodies were gifted from Dr. Peter Davies and brain lystates from BACE1−/− mice were obtained from Prof Mike Ashford. The work presented here was funded by Alzheimer’s Research UK (Grant refs: ARUK-PPG2014A-21 and ARUK-NSG2015-1 to BP and DK and NIH/NIA grants NIH/NINDS R01 NS082730 and R01 AG044372 to NK).
Open AccessThis article is distributed under the terms of the Creative Commons Attribution 4.0 International License (http://creativecommons.org/licenses/by/4.0/), which permits unrestricted use, distribution, and reproduction in any medium, provided you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons license, and indicate if changes were made.