Introduction
Alzheimer’s Disease (AD) is the leading cause of age-related dementia, affecting over 5 million people in the United States alone [
5]. AD pathology is characterized by two primary lesions: extracellular amyloid plaques and intraneuronal tau-laden neurofibrillary tangles. The mechanisms that drive AD remain unclear, but the ‘amyloid cascade hypothesis’ first proposed by Hardy and Higgins posits that beta-amyloid (Aβ) accumulation is the initiating factor in AD pathogenesis [
8,
31,
33]. Increased deposition of Aβ, in turn, is thought to promote the hyperphosphorylation of tau leading to neurofibrillary tangle (NFT) formation [
2]. Together, Aβ and tau pathologies disrupt critical biological functions such as axonal transport and synaptic function and promote neuroinflammation, ultimately leading to widespread synaptic and neuronal loss [
65,
83].
The role of neuroinflammation in the development and progression of AD has been studied for several decades [
26,
27,
34]. However, the recent identification of AD risk polymorphisms in several microglial-enriched genes such as TREM2, MS4A, and CD33, has intensified this area of research [
9,
15,
25,
28,
32,
42,
51,
86]. Microglia serve as one of the brain’s primary mechanisms of Aβ clearance, but also play critical roles in neuronal homeostasis and synaptic plasticity [
63,
69,
76]. In response to Aβ, microglia increase their expression of pro-inflammatory cytokines, which has been shown to promote tau hyperphosphorylation and NFT pathology [
10,
49] and contribute to synaptic and neuronal dysfunction [
3,
30,
69].
To better understand the role of microglia in the interactions between Aβ and tau pathologies we crossed two transgenic AD models, 5xfAD and Thy-Tau22 mice, to create a novel bigenic line termed ‘T5x’ mice. 5xfAD mice express three APP mutations (Swedish, Florida, London) and two PS1 mutations (M146L, L286V) that are co-inherited and driven under control of the neuronal-specific Thy1.2 promoter [
61]. 5xfAD exhibit intraneuronal Aβ accumulation beginning at 1.5 months, amyloid plaque deposition and gliosis starting at 2 months, synaptic loss by 6 months, and neuronal death beginning at 9 months of age [
22]. Amyloid deposition in these mice is most prominent in the hippocampus, subiculum, deep cortical layers, and the basal lateral amygdala. Increased neuroinflammation and microglial activation have also been shown to play an essential role in mediating disease progression in this model [
43]. Thus, 5xfAD mice provide an aggressive amyloidogenic model that exhibit robust AD-associated plaque pathology and microgliosis. In comparison, Thy-Tau22 mice express human four repeat tau with two mutations (G272V, P301S) driven under the Thy1.2 promoter and progressively develop hippocampal hyperphosphorylated tau, neurofibrillary tangles, and gliosis [
72]. Studies examining the effects of exercise, caffeine and A
2A receptor modulation in Thy-Tau22 mice suggest important roles for microglia and neuroinflammatory responses in the accumulation of tau pathology [
7,
44].
Interactions between Aβ and tau in mouse models were first reported in two seminal studies in 2001 that clearly demonstrated that Aβ accumulation could accelerate the development of tau pathology [
24,
47]. Subsequently, the 3xTg-AD model was generated and extended our understanding of the influence of Aβ on tau and was used to explore the role of microglial inflammation in this process [
39,
62]. Since then several other bigenic models have been created and studies have continued to investigate the effects of amyloid on tau and suggested that the amyloid cascade follows a unidirectional pathway [
18,
35,
71]. However, many of these models exhibit far less pathology than occurs in human AD cases and thus more complex interactions that might occur over decades in the human brain or at later stages of advanced disease may not be faithfully recapitulated in many of these models [
85]. Furthermore, the potential role of inflammation in these interactions between Aβ and tau remains greatly understudied.
In the present study, we combined a model of Aβ accumulation (5xfAD mice) with a progressive model of neurofibrillary tangle pathology (Thy-Tau22 mice). The resulting ‘T5x’ mice were generated to provide insight into the later stages of disease progression akin to that observed in clinically-diagnosed AD patients and the interactions and consequences of advanced amyloid pathology on tau and visa-versa. By comparing T5x bigenic mice to their single transgenic littermates we have uncovered potential new roles for tau in the modulation of Aβ and neuroinflammatory response. Most notably, we identify and examine how amyloid and tau synergize to alter microglial activity and promote Aβ clearance. Thus, these data provide additional insight into the relationship between AD pathology and neuroinflammatory response and suggest that tau can exhibit reciprocal interactions with amyloid.
Material and Methods
Generation of T5x mice
Thy-Tau22 mice express human 4 repeat tau with two frontotemporal dementia-associated point mutations (G272V and P301S) under control of the neuronal driven promoter Thy1.2 and are maintained on a C57Bl6/J background [
72]. The 5xfAD mice used in this study are also maintained on a congenic C57Bl6/J background and were purchased from the Mutant Mouse Resource and Research Center (MMRRC, stock# 034848-JAX). The 5xfAD model co-expresses human amyloid precursor protein (APP695) carrying the Swedish, Florida, and London mutations and a human presenilin-1 (PS1) transgene carrying the M146L and L286V mutations under the Thy-1 promoter. Both APP and PS1 transgenes are co-integrated and thus co-inherited. Heterozygous Thy-Tau22 and 5xfAD mice were crossed to create Thy-Tau22-5xfAD (T5x) mice, as well as Thy-Tau22, 5xfAD, and WT littermates that were genotyped via PCR amplification of human tau, PS1, and APP transgenes. The number of mice from each sex-balanced genotype was: WT (
n = 8), 5x (
n = 11), Tau (
n = 14), and T5x (
n = 9). All mice were maintained on a purebred C57Bl/6 background and group housed (2–4 mice/cage) on a 12 h/12 h light/dark cycle with access to food and water ad libitum. All animal procedures were performed in strict accordance to the National Institutes of Health and University of California Institutional Animal Care and Use Committee.
Tissue preparation and neuropathological analysis
Following behavioral testing, all mice were sedated with euthasol and sacrificed by cardiac perfusion with 0.1 M phosphate buffered saline (PBS). Brains were removed and one hemisphere was snap frozen on dry ice while the other hemisphere was postfixed in 4 % paraformaldehyde for 48 h then stored in PBS + 0.05 % sodium azide. Fixed half-brains were placed in 30 % sucrose for at least 48 h before being cut in the coronal plane (40 μm sections) using a freezing sliding microtome.
Immunohistological staining
Brain sections were rinsed in PBS three times and blocked in PBS + 0.05 % Triton-X with 5 % donkey or goat serum for one hour. Primary antibodies used included: 6E10 (Covance; 1:500) and 82E1 (ABL America; 1:500) against Aβ, total Tau (human + mouse; Dako; 1:1000), human tau (HT7, ThermoFisher; 1:1000), phosho-tau epitopes AT8 (Pierce; 1:500) and PHF-1, and conformational tau epitope MC-1 (generously provided by Peter Davies; 1:1000). Analysis of gliosis, phagocytosis, and dendritic architecture utilized CD68 (Abcam; 1:200), IBA1 (Wako; 1:1000), GFAP (Abcam; 1:1000), Beta3Tubullin (Covance; 1:1000). Sections were incubated in primary antibodies at 4° Celsius overnight. Sections were then washed three times with PBS and incubated with appropriate Alexa fluor-conjugated secondary antibodies at room temperature for one hour. Sections were then rinsed three additional times, mounted on slides and coverslipped with Fluoromount-G with DAPI.
Biochemical analysis
Hippocampus and cortex was microdissected from frozen brains and processed to collect both soluble and insoluble extracts. Briefly, microdissected tissue was homogenized in TPER (ThermoFisher) and centrifuged at 12,000 RPM for 15 min. Supernatant was collected as the soluble fraction and the pellet was treated with 70 % formic acid and spun down at 25,000 rpm. The resulting supernatant was collected as the insoluble fraction. Insoluble protein samples were neutralized for Western blotting and further precipitated with trichloroacetic acid (TCA) when probing for insoluble tau. Protein samples were denatured at 95 °C for 15 min before being loaded onto 4-20 % TGX precast polyacrylamide gels (Bio-rad). Antibodies used for western blotting include: HT7 (1:1000), PS199 (Abcam; 1:1000), PS202 (Abcam; 1:1000), AT100 (ThermoFisher; 1:1000), AT270 (ThermoFisher; 1:1000), PHF1 (1:1000), 6E10 (1:1000), GFAP (1:1000). Mesoscale Discovery immunoassay kits (Mesoscale Diagnostics) were used for cytokine (K15048G) and Aβ38, 40, and 42 (K15199E) quantification of cortical samples following the manufacturers protocols. The proinflammatory MSD was able to detect levels of that were within the standard curve, whereas brain levels of IFN-γ, IL-12p70, CXCL1 and IL-4 were below the threshold of detection.
Confocal microscopy and quantification
Equivalent brain sections were picked and immunofluorescent sections were imaged using Olympus FX1200 confocal microscope. Amyloid plaque burden identified by 82E1 were visualized through a Z-stack image taken through the entire depth of the section at 1 μm intervals. Confocal files were then rendered in 3D and analyzed by a blinded observer using the volume function of IMARIS software (Bitplane). Microglia and astrocyte quantification by confocal microscopy was also analyzed using IMARIS. Microglia number was quantified using IMARIS spot function and process morphology was measured using filament length and branching functions. Astrocyte IMARIS quantification was performed using the volumetric function. Microglia phagocytosis of Aβ was quantified using a combined immunofluorescent staining of IBA1, 82E1, and CD68. High magnification Z-stack images were taken of randomly selected plaques while being blinded to IBA1 microglial staining. Quantification of internalized Aβ was done according to previously described protocols [
40,
54]. To account for varying total microglia numbers across images, the internalized Aβ index was normalized to the number of microglia per image. Quantification of neuronal degeneration was performed in equivalent hippocampal brain sections using immunohistochemical labeling of Beta-III Tubulin (B3T). Four randomly selected square sub-areas were selected in pyramidal, radiatum, or molecular layers of hippocampus CA1 and quantified by optical density using ImageJ.
Nanostring analysis
RNA was isolated from microdissected hippocampi of WT, Thy-Tau22, 5xfAD, and T5x mice using RNeasy Plus Universal Mini Kit (Qiagen). RNA samples were run a custom Nanostring panel (Nanostring Technologies) examining mouse AD-linked genes. To evaluate mouse amyloidogenesis, we compared levels of murine RNA for genes APP, BACE1, BACE2, ADAM10, PSEN1, and PSEN2.
Statistical analysis
Statistical analysis was performed using StatView software (SAS Institute Inc.). Statistical comparisons between multiple groups were performed using ANOVA followed by Fischer’s PLSD post-hoc tests. Error bars represent the standard error of the mean. Groups were considered statistically significant when *p < 0.05 for both ANOVA and posthoc analysis.
Discussion
Although the notion that amyloid increases tau pathology is well established, very few studies have in contrast examined whether a reciprocal relationship between tau and the development of amyloid pathology might exist. Beginning in 2001, two seminal studies demonstrated that Aβ could exacerbate the development of tau pathology in transgenic AD tau models [
24,
47]. Subsequently, a similar causal relationship between Aβ and tau pathology was established in several other AD models [
13,
17,
39,
62]. Even in the absence of the expression of a human tau transgene, many AD transgenic mice exhibit some degree of Aβ-induced tau hyperphosphorylation, a prerequisite for the development of neurofibrillary tangles. Likewise, in humans, triplication of APP, either in some family pedigrees or in trisomy 21, leads to elevated levels of Aβ and the development of tau pathology [
53,
79]. Independent of Aβ, alterations in Presenilin-1 have also been shown to increase tau hyperphosphorylation in transgenic mice expressing mutant tau [
14,
78]. Thus, it is quite likely that the presence of a mutant PS1 transgene in 5xfAD influences tau phosphorylation both directly and indirectly by enhancing Aβ42 generation.
As expected, the accumulation of Aβ in T5x mice leads to a dramatic increase in tau hyperphosphorylation and increased accumulation of neurofibrillary tangle pathology within the hippocampus and neocortex. Yet surprisingly, T5x mice also exhibited a 50 % reduction in amyloid plaque burden and insoluble Aβ species versus 5xfAD littermates. Although the expression of both the APP and tau transgenes are driven by the Thy1 promoter, the expression of APP in T5x mice was unchanged and therefore, decreased Aβ levels were not the result of altered transgene expression. Furthermore, levels of soluble Aβ and APP processing enzymes were unaffected; suggesting that the observed decreases in plaques and insoluble Aβ was not due to effects on APP processing.
However, the combination of amyloid and tau pathology did produce a dramatic effect on neuroinflammatory processes, included increased microgliosis and astrogliosis in T5x versus 5xfAD or Thy-Tau22 littermates. Therefore, we examined whether changes in specific cytokines could explain the increase in neuroinflammation. Our data show that the combined presence of amyloid and tau pathology exacerbates the levels of both TNF-α and IL-2 (Fig.
4). In contrast, the observed decreases in IL-1β, IL-5, and IL-6 in T5x mice versus 5xfAD littermates may reflect the predominant influence of amyloid on the release and maturation of certain pro-inflammatory cytokines. Although IL-1β has been shown to increase tau phosphorylation, amyloid has also been reported to promote astrocytic and microglial release of IL-1β [
20,
36,
40,
74]. Thus, while one may expect elevated IL-1β in T5x mice in association with increased tau hyperphosphorylation [
49], the influence of reduced amyloid burden on IL1β induction may take precedence over effects caused by changes in tau. Therefore, it may be the case that IL-1β, IL-5 and IL-6 are less elevated in T5x than 5xfAD mice as a consequence, rather than a cause of reduced amyloid burden. However, future studies that examine the progression of pathology and inflammation in T5x across multiple ages will likely help to resolve these questions.
TNF-α, a pro-inflammatory cytokine that is released by activated microglia was one of the most upregulated cytokines observed in T5x mice [
41,
55]. Although TNF-α does not directly increase microglial proliferation, TNF-α promotes astrocyte proliferation and GM-CSF release, which in turn, can stimulate microglial proliferation [
6,
45,
73]. Taken together, elevated astrocyte numbers and a ~5-fold increase in TNF-α in T5x mice relative to 5xfAD and Thy-Tau22 littermates suggests that the combined stress of amyloid and tau pathology work synergistically to promote both astrocytic and microglial proliferation. Furthermore, the significant decrease in IL-10 expression in T5x mice suggested that microglial phagocytosis of Aβ may also be altered as recent studies have shown that deletion of IL-10 can enhance microglial Aβ phagocytosis [
29]. Our finding that T5x mice exhibit a similar decrease in IL-10 relative to Thy-Tau22 and 5xFAD littermates suggest that decreased IL-10 is a result of the combined stress of tau and amyloid pathology. Subsequent analysis confirmed that T5x microglia indeed have increased Aβ phagocytosis, providing further support that tau pathology contributes to microglial regulation of amyloid. It should be noted that while our results support strong interactions between tau pathology and microglial activation, the exact temporal relationship between the two is still to be determined. Additional longitudinal studies will be needed to determine whether accumulation of tau pathology first promotes microglial activation or whether changes in microglia precede and drive increases in tau hyperphosphorylation.
While cytokines are often determined by the extent of AD pathology, they can also control the progression of the disease through the modulation of microglial response. The influence of microglia in AD has received growing attention with the recent discovery of genetic risk polymorphisms in several microglial-enriched genes [
15]. Some of these risk factor genes such as CD33 have themselves been implicated in microglial Aβ phagocytosis [
25,
81]. Thus, the role of microglia in limiting Aβ pathology continues to garner new attention. Microglia have long been implicated in the modulation of tau hyperphosphorylation and misfolding through cytokine-mediated neuroinflammatory activation of tau kinases [
49,
84]. More recently, studies have shown that the misfolding of tau can promote microglial activation [
88], and tau oligomers and fibrils have been found to interact directly with microglia [
56]. In addition, microglia have also recently been implicated in the propagation of tau within the brain [
4]. The fact that microglia exhibit the capacity to directly phagocytose the same tau oligomers that can promote their activation, highlights the possibility that tau pathology is not only exacerbated by microglial activation, but likely in turn modulates microglial responses within the brain [
52]. We reasoned that changes in cytokine levels observed in T5x mice would affect microgliosis and our current findings suggest that tau and amyloid pathology in T5x mice synergistically recruit a robust microglial response. Significantly increased activated amoeboid microglia in T5x mice versus Thy-Tau22 littermates suggests that amyloid pathology is the strongest determinant of microglial activation. Amyloid holds an advantage over predominately intracellular tau aggregates in its potential influence over microglial activation as evident of amyloid soluble oligomers released from abundant extracellular plaques [
55,
70,
77]. However, our findings of increased microglial activation, phagocytosis, and numbers in T5x mice, suggest that tau pathology can also contribute to or further modulate microglia activation state.
The remaining question is exactly how development of tau pathology influences the neuroinflammatory response. One possibility is that amyloid driven tau hyperphosphorylation and subsequent release of intracellular factors and tau oligomers from degenerating neurons leads to altered microglial activation states, potentially increasing Aβ phagocytosis and thereby providing a somewhat beneficial response. Previous studies have reported the activation of microglia by the release of apoptotic factors and intracellular molecules such as ADP or ATP [
21,
37,
58]. When we examined T5x mice for dendritic degeneration, we observed a reduction in hippocampal Beta-III tubulin staining in T5x mice compared to transgenic and WT littermates (Additional file
6: Figure S6). Notably, hippocampal regions displaying the significant reduction in Beta-III tubulin were accompanied by the previously described rod microglia (Fig.
4 g-j) [
80]. Although not extensively characterized, rod microglia have been found in human AD cases and exhibit abnormally frequent interactions with synaptic clefts [
46,
59,
82]. T5x rod-like microglia are similar in morphology to recently described rod-microglia which appear highly activated and respond to diffuse neuronal injury [
87]. The association between reduced β3-Tubullin density and rod microglia presence therefore suggests that degenerating dendrites and subsequent release of intracellular factors such as ADP may contribute to the altered microglia activation state in T5x mice. However, further studies are clearly needed to elucidate the exact mechanisms by which tau synergizes with Aβ to modulate microglial activation.
It is important to acknowledge that our findings both corroborate and differ from other recently examined amyloid-tau crosses. We report a significant reduction in amyloid plaque burden and insoluble Aβ in T5x mice relative to 5xfAD littermates that appears to be mediated via increased microglial-mediated Aβ clearance. Using a similar model, one group has shown that 5xfAD crossed with Tg30 tau mice likewise exhibit decreased Aβ pathology, although the potential role of microglia in this finding was not examined [
35]. In contrast, two other studies crossed 5xfAD mice with other tau models and reported no change in Aβ plaques by optical densitometry, although only 3 or 4 mice per group were compared versus group sizes of 9–10 mice for the current Aβ ELISA analysis [
75]. Yet, one of these studies nevertheless showed a non-significant 45 % reduction in Aβ plaque load at 3 months of age. It is therefore quite possible that the use of 3D plaque volume quantification, ELISA, or a larger sample size would have revealed a similar significant reduction in Aβ to that observed in the current study. In another study, Tg2576 APP mice were crossed with mutant tau VLW mice. Surprisingly, this study reported a significant increase in amyloid deposition in bigenic mice [
67]. Likewise, a recent study that examined a cross between rTgTauEC and APP/PS1 mice also reported increased amyloid burden [
66].
These varying reports suggest that perhaps the magnitude of tau pathology and/or microgliosis can influence the effects on Aβ pathology. In support of this notion, we find that T5x mice exhibit a dramatic increase in microglial number, shift in morphology, and change in cytokines that is not apparent in Thy-Tau22 littermates with lower levels of tau pathology (Figs.
4 and
5). Interestingly, in other studies that crossed different tau models with 5xfAD mice [
35,
75], there were either no significant changes or a similar decrease in amyloid pathology. In contrast, one study reported an increase in amyloid burden when Tg2576 mice were crossed with VLW tau mice [
67]. 5xfAD mice, which carry three APP and two PS1 mutations, generate considerably more amyloid pathology than Tg2576 mice and the presence of a PS1 transgene may further promote tau hyperphosphorylation both directly [
14,
78] and indirectly by enhancing Aβ42 generation. Thus, differences in levels of tau, the type and magnitude of amyloid pathology, and their corresponding effects on microglial activation state might explain these discrepancies. For example, it is quite possible that the more aggregation prone Aβ42 leads to a greater increase in tau pathology and a differential activation of microglia that combine to promote microglial phagocytosis. Another possible explanation for these contrasting findings is the potential impact of background strain on pathology. In the current study we crossed congenic C57/Bl6 5xfAD mice with congenic C57/Bl6 Thy-Tau22 mice, thus background strain was identical between all four genotypes examined. In some of the contrasting studies, the cross-bred mice exhibited differential mixed background strains that include Bl6, CH3, and FVB [
66] and thus variability in background genetics even between littermates could influence the development of pathology. Although our current data clearly suggests that tau pathology can reduce the accumulation of Aβ, these prior reports that in many cases corroborate our findings but in others contrast our findings clearly suggest that this remains a highly complex question that will need additional examination. For example, it is quite possible that changes in APP transport or alterations in autophagy also play a role in this process and thus studies that further examine these additional potential mechanisms are also needed.
Competing interests
The authors declare that they have no competing interest.
Authors’ contributions
WC, WWP, EMA, and M B-J. designed research; WC, AL, and MB-J performed research; WC, AL, and MB-J. analyzed data; LB and DB provided the Thy-Tau22 mice; WC, WWP, EMA, and MB-J wrote the paper. All authors read and approved the final manuscript.