Somatodendritic accumulation of Tau in Alzheimer's disease is promoted by Fyn‐mediated local protein translation

Abstract

The cause of protein accumulation in neurodegenerative disease is incompletely understood. In Alzheimer's disease (AD), the axonally enriched protein Tau forms hyperphosphorylated aggregates in the somatodendritic domain. Consequently, a process of subcellular relocalization driven by Tau phosphorylation and detachment from microtubules has been proposed. Here, we reveal an alternative mechanism of de novo protein synthesis of Tau and its hyperphosphorylation in the somatodendritic domain, induced by oligomeric amyloid‐β (Aβ) and mediated by the kinase Fyn that activates the ERK/S6 signaling pathway. Activation of this pathway is demonstrated in a range of cellular systems, and in vivo in brains from Aβ‐depositing, Aβ‐injected, and Fyn‐overexpressing mice with Tau accumulation. Both pharmacological inhibition and genetic deletion of Fyn abolish the Aβ‐induced Tau overexpression via ERK/S6 suppression. Together, these findings present a more cogent mechanism of Tau aggregation in disease. They identify a prominent role for neuronal Fyn in integrating signal transduction pathways that lead to the somatodendritic accumulation of Tau in AD.

Introduction

Accumulation of proteins in an aggregated form is a unifying feature of neurodegenerative disorders. However, what causes particular signature proteins to accumulate in diseased brains is only incompletely understood, particularly as these proteins tend to aggregate in compartments that differ from those to which they are predominantly localized under physiological conditions. In Alzheimer's disease (AD), the two hallmark lesions are amyloid plaques that are formed from amyloid‐β (Aβ) peptide and neurofibrillary tangles that are composed of the microtubule‐associated protein Tau. A large number of studies have addressed the link between these lesions, placing Aβ upstream of Tau in a patho‐cascade (Bloom, 2014; Frandemiche et al, 2014; Jin et al, 2011; Pooler et al, 2015). Here, oligomeric Aβ (Aβo) is believed to be a major neurotoxic species, causing synaptic dysfunction and neurodegeneration by a so far incompletely understood mechanism (Lesne et al, 2006; Shankar et al, 2008; Walsh et al, 2002).

The kinase Fyn, a member of the Src family of non‐receptor tyrosine kinases (SFKs), has emerged as a key mediator of the toxic effects of Aβ due to its crucial role in regulating synaptic functions. Fyn is widely expressed in the hippocampus, integrating various signaling pathways and having a critical modulatory effect on synaptic protein trafficking (Ohnishi et al, 2011), as well as synaptic plasticity and learning (Trepanier et al, 2012). When overexpressed in human Aβ‐forming amyloid precursor protein (APP) transgenic mouse models, Fyn accelerates synaptic and cognitive impairment (Chin et al, 2005), whereas synaptic degeneration and memory loss are rescued when Fyn is either depleted or its activity is suppressed in an APP mutant background (Chin et al, 2004; Kaufman et al, 2015). A role for Fyn in mediating Aβo toxicity is further supported by studies in cultured neurons (Larson et al, 2012; Um et al, 2013, 2012). Fyn has also been shown to be targeted to dendritic spines by Tau, a process that is facilitated by the interaction of Tau with the SH3 domain of Fyn, and of tyrosine‐phosphorylated Tau with the SH2 domain of Fyn (Ittner et al, 2010; Lee et al, 2004; Usardi et al, 2011). Therefore, both Aβo toxicity and Tau pathology involve Fyn kinase.

Early in development, Tau is distributed throughout the neuron; with maturation, however, it becomes enriched in the axon (Kosik & Finch, 1987). In AD and related tauopathies, Tau accumulates in both the soma and the dendrites in a hyperphosphorylated form (Wang et al, 2013). It is generally assumed that hyperphosphorylated Tau in the axon detaches from the microtubules and passes through the axon initial segment, which serves as a diffusion barrier for physiologically phosphorylated Tau, before accumulating in the cell body and dendrites, a process that is partly mediated by Aβ (Li et al, 2011; Sohn et al, 2016; Zempel et al, 2013). However, whether Aβ employs a mechanism other than relocalization of Tau to account for the massive accumulation of Tau in the somatodendritic compartment remains unclear. Here, we report an additional, more cogent mechanism that involves local Aβ‐mediated protein translation of Tau in the somatodendritic domain. More specifically, we show that this activation occurs through Fyn, the serine/threonine‐directed kinase ERK and the ribosomal protein S6, and is associated with increased phosphorylation of Tau at multiple residues. Together, these findings reveal de novo protein synthesis of Tau in the somatodendritic compartment, mediated by the Fyn/ERK/S6 signaling pathway, as a novel pathomechanism in AD.

Results

Fyn massively boosts exogenous Tau expression via protein translation

We and others have shown that Fyn and Tau not only interact, but that Tau is also a substrate of Fyn, with Y18 being the primary phosphorylation site (Ittner et al, 2010; Lee et al, 2004; Usardi et al, 2011). To investigate this cross talk in more detail, we co‐transfected tagged Tau and Fyn expression vectors into HEK293T cells. Tau levels were massively induced by Fyn, a stimulatory effect that was virtually absent when GFP and Tau were co‐expressed. Similarly, Fyn levels were increased in the presence of Tau, whereas GFP levels were not affected regardless of whether Tau or Fyn was co‐expressed (Fig 1A). This was further confirmed by labeling Tau with an amino‐terminal V5‐ or GFP‐tag (Fig EV1A–D). We observed massively pronounced Tau overexpression (up to 43.6‐fold) in Tau and Fyn co‐transfected HEK293T cells (Fig 1B and C). In accordance with the fact that Tau accumulation is always accompanied by increased phosphorylation in tauopathies, we also observed hyperphosphorylation at serine and threonine residues, including the AD epitopes AT180 (T231/S235), 12E8 (S262/S356), and pS422‐Tau (S422) (up to 68.4‐fold), in addition to Y18 phosphorylation (Fig 1B and C). Tau expression was sharply induced by increasing the dose of Fyn when cells were co‐transfected with constant amounts of Tau plasmid (Fig EV1E and F). Interestingly, no such effects were apparent when Fyn was kept constant and the dose of Tau plasmids was increased (Fig EV1G and H). The use of a kinase dead (FynKD, K299M mutation) and a constitutively active mutant (FynCA, Y531F mutation) demonstrated that Tau expression was regulated by Fyn activity (Fig EV1I and J). Together, these data demonstrate that the Fyn‐mediated increase in Tau levels is Fyn dose‐ and activity‐dependent.

Figure 1. Fyn boosts exogenous Tau translation via ERK/S6 signaling.

1. Representative Western blots of extracts from HEK293T cells transfected with GFP+Tau, GFP+Fyn and Tau+Fyn expression constructs. C‐terminally V5‐tagged Tau and Myc‐tagged Fyn validated with corresponding anti‐tag antibodies.
2. Western blotting reveals massively increased levels of total and phosphorylated Tau induced by Fyn. pY18‐Tau: Y18; AT180: T231/S235; 12E8: S262/S356; pS422‐Tau: S422.
3. Quantification of (B) (mean ± s.e.m., n = 3,3; two‐tailed t‐test, pY18‐Tau, t ₍₄₎ = 6.218, **P = 0.0034; AT180, t ₍₄₎ = 19.74, ****P < 0.0001; 12E8, t ₍₄₎ = 11.76; ***P = 0.0003; pS422‐Tau, t ₍₄₎ = 5.39, **P = 0.0057; Tau‐V5, t ₍₄₎ = 6.418, **P = 0.003).
4. Activation of MAP kinases (p38, JNK1/2, and ERK1/2) and protein translational regulators (mTOR, p70S6K, and S6) as shown by Western blotting.
5. Quantification of (D): mean ± s.e.m.; one‐way ANOVA, Sidak's post hoc test, p‐p38/p38, n = 3,3,3,3, F _(3,8) = 5.948, Tau+Fyn versus Fyn, *P = 0.0317, Tau+Fyn versus Tau, *P = 0.0246; p‐JNK/JNK, n = 3,3,3,3, F _(3,8) = 101.2, Tau+Fyn versus Fyn, ****P < 0.0001, Tau+Fyn versus Tau, ****P < 0.0001; p‐ERK/ERK, n = 3,3,3,3, F _(3,8) = 30.2, Tau+Fyn versus Fyn, ***P = 0.0002, Tau+Fyn versus Tau, ***P = 0.0006; p‐mTOR/mTOR, n = 3,3,3,3, F _(3,8) = 0.1717, P = 0.9125; p70S6K/actin, n = 6,6,6,6, F _(3,20) = 5.763, Tau+Fyn versus Tau, **P = 0.0037; p‐S6/actin, n = 6,6,6,6, F _(3,20) = 43.66, Tau+Fyn versus Fyn, ****P < 0.0001, Tau+Fyn versus Tau, ****P < 0.0001.
6. Co‐transfection of Tau with various Fyn mutants shows that Fyn kinase activity is essential for phospho‐activation of ERK1/2 and S6. FynKD (kinase‐dead mutant); FynWT (wild‐type); FynCA (constitutively active).
7. Quantification of (F): mean ± s.e.m., n = 3 per group. One‐way ANOVA, Sidak's post hoc test; p‐ERK/ERK, F _(2,6) = 40.77, Tau+FynWT versus Tau+FynKD, *P = 0.049, Tau+FynWT versus Tau+FynCA, **P = 0.0021; pS6/S6, F _(2,6) = 25.64, Tau+FynWT versus Tau+FynKD, *P = 0.0134, Tau+FynWT versus Tau+FynCA, *P = 0.0425.
8. Quantification of immunoblots (Fig EV3A) for effects of phorbol 12‐myristate 13‐acetate (PMA) on Tau expression and the ERK/S6 signaling pathway. Cells were treated with PMA (100 nM) or DMSO for 24 h (mean ± s.e.m., n = 4 per group, two‐tailed t‐test, DakoTau/actin, t ₍₆₎ = 16.6, ****P < 0.0001; p‐ERK/ERK, t ₍₆₎ = 10.2, ****P < 0.0001; p‐S6/S6, t ₍₆₎ = 7.594, ***P = 0.0003).
9. Quantification of immunoblots (Fig EV3D) showing the effects of U0126 on Tau expression and the ERK/S6 signaling pathway. Cells were treated with U0126 (20 μM) or DMSO for 24 h (mean ± s.e.m., n = 4 per group; Tau‐V5/actin, one‐way ANOVA, Dunnett's multiple comparisons, F _(2,9) = 128.1; Tau+DMSO versus Tau+Fyn+DMSO, ****P = 0.0001; Tau+Fyn+DMSO versus Tau+Fyn+U0126, ****P = 0.0001; p‐ERK/ERK, two‐tailed t‐test, Tau+DMSO versus Tau+Fyn+DMSO, t ₍₆₎ = 5.050, **P = 0.0023; p‐S6/S6, two‐tailed t‐test, Tau+DMSO versus Tau+Fyn+DMSO, t ₍₆₎ = 7.439, ***P = 0.0003).
10. Quantification of immunoblots (Fig EV3E) evaluating the Tau‐Fyn‐ERK immuno‐complex using immunoprecipitation (IP) (mean ± s.e.m., two‐tailed t‐test; Tau bound to ERK1/2, n = 3 per group, t ₍₄₎ = 6.372, **P = 0.0031; Fyn bound to ERK1/2, n = 3 per group, t ₍₄₎ = 3.835, *P = 0.0185).

Data information: *P < 0.05, **P < 0.01, ***P < 0.001, ****P < 0.0001.Source data are available online for this figure.

Figure EV1. Fyn triggers Tau expression in HEK293T cells in a dose‐ and activity‐dependent manner.

• A
  Western blot analysis indicates that Fyn boosts the expression of exogenous Tau that has been tagged N‐terminally with V5 in HEK293T cells.
• B
  Quantification of Western blots in (A) (mean ± s.e.m.; n = 3 per group; two‐tailed t‐test, t ₍₄₎ = 4.123, *P = 0.0146).
• C
  Western blots of lysates from HEK293T cells transfected with GFP‐tagged Tau only or co‐transfected with Myc‐tagged Fyn using anti‐GFP and anti‐Myc antibodies.
• D
  Fluorescence and bright‐field images of cells from (C). Scale bar: 200 μm.
• E, F
  Representative Western blots showing the dose effect of Fyn on Tau expression in HEK293T cells (E), quantified in (F) (mean ± s.e.m., n = 3 per group; one‐way ANOVA, Dunnett's post hoc test compared to the 6:0 control group, F _(4,10) = 22.27; 6:2 versus 6:0, ***P =0.0002, 6:3 versus 6:0, ***P = 0.0002, 6:6 versus 6:0, **P = 0.0017).
• G
  A range of Tau doses fails to alter Fyn protein levels.
• H
  Quantification of Western blots in (G) (mean ± s.e.m., n = 3 per group, one‐way ANOVA, Dunnett's multiple comparisons relative to the 6:1 control group, F _(3,8) = 3.391, P = 0.0743; n.s., not significant).
• I
  Representative Western blots indicating protein levels in HEK293T cells when Tau was co‐transfected with constructs encoding kinase dead (FynKD), wild‐type (FynWT), or constitutively active Fyn (FynCA).
• J
  Quantification of Western blots in (I) (mean ± s.e.m., n = 6 per group. One‐way ANOVA, Sidak's test compared to the Tau+FynWT group, F _(2,15) = 35.15, Tau+FynWT versus Tau+FynKD, **P = 0.0010, Tau+FynWT versus Tau+FynCA, **P = 0.0025).

Data information: *P < 0.05, **P < 0.01, ***P < 0.001.

To determine how Fyn was regulating the overexpression of Tau, we next analyzed transcript levels, including GFP as a control. GFP mRNA levels were 1.1‐fold higher, and MAPT (Tau‐encoding gene) levels 1.6‐fold higher in the presence of Fyn (Fig EV2A), suggesting that the > 43.6‐fold increase observed in Tau protein levels (Fig 1B and C) was unlikely due to an effect of Fyn on transcription. To address this more directly, we used the transcriptional inhibitor actinomycin D and the protein translation inhibitor cycloheximide. As expected, Fyn‐regulated Tau overexpression was more strongly suppressed by cycloheximide than by actinomycin D, suggesting that Fyn induces Tau translation rather than transcription (Fig EV2B and C).

Figure EV2. Fyn boosts Tau expression in HEK293T cells via ERK/S6‐mediated protein translation.

1. GFP and Tau constructs were transfected with or without Fyn in HEK293T cells, followed by determination of GFP and MAPT transcript levels using quantitative real‐time PCR. Relative transcript levels are shown as fold changes compared to the GFP or Tau only group, respectively, after normalizing separately to Actin levels (mean ± s.e.m., two‐tailed t‐test; GFP, n = 3,3, t ₍₄₎ = 3.026, *P = 0.0389; MAPT, n = 6,6, t ₍₁₀₎  = 4.043, **P = 0.0023).
2. 12 h after Tau was transfected alone or with Fyn, HEK293T cells were challenged with or without two concentrations (in μg/ml) of actinomycin D (ActD) and cycloheximide (CHX), respectively, for another 12 h, before collection for analysis. *, Unspecific band.
3. Quantification of (B). Mean ± s.e.m.; n = 4 per group; one‐way repeated measures ANOVA, Sidak's multiple comparisons, F _(4,12) =29.62, control versus ActD 0.1, ***P = 0.0001; control versus ActD 1, ***P = 0.0003; control versus CHX10, ****P < 0.0001; control versus CHX100, ****P < 0.0001; ActD1 versus CHX100, *P = 0.0466.
4. Immunoblots for phosphorylated forms of a range of eukaryotic translation‐related proteins and components of the translational machinery, including Akt, eEF2 (eukaryotic elongation factor 2), eIF2α (eukaryotic translation initiation factor 2α), eIF4E, and 4EBP1 (eIF4E binding protein 1) were evaluated in transfected HEK293T cells.
5. Quantification of (D) suggests no significant difference (mean ± s.e.m., n = 6 per group, one‐way ANOVA; p‐Akt, F _(3,20) = 0.8977, P = 0.4596; p‐eEF2, F _(3,20) = 2.46, P = 0.0924; p‐eIF2α, F _(3,20) = 0.9528, P = 0.4341; p‐eIF4E, F _(3,20) = 0.6305, P = 0.6039; p‐4EBP1, F _(3,20) = 0.1484, P = 0.2491).
6. Western blots showing dose‐dependent phosphorylation of ERK1/2 and S6 by Fyn in Tau‐expressing HEK293T cells.
7. Quantification of results in (F); mean ± s.e.m., n = 3 per group; one‐way ANOVA, Dunnett's comparisons relative to the 6:0 control group, p‐ERK/ERK, F _(2,6) = 40.77, 6:3 versus 6:0, *P = 0.0255, 6:6 versus 6:0, **P = 0.0069; p‐S6/S6, F _(2,6) = 25.64, 6:3 versus 6:0, **P = 0.0010, 6:6 versus 6:0, **P = 0.0019.

Data information: *P < 0.05, **P < 0.01, ***P < 0.001, ****P < 0.0001.

Specific activation of the Fyn/ERK/S6/Tau signaling cascade

We were next interested in dissecting the signaling pathway whereby Fyn exerts its effects on Tau protein levels by testing the MAP kinase (MAPK) family of upstream translational regulators that includes the kinases p38, JNK1/2, and ERK1/2. Together with ribosomal protein S6, these kinases have been implicated in the general regulation of protein synthesis (Kelleher et al, 2004; Pende et al, 2004), with evidence suggesting that S6 activation occurs downstream of MAPKs and requires Fyn activity (Salmond et al, 2009). We determined both the total and activated, that is, phosphorylated, levels of the three kinases and found that all showed pronounced activation in Tau/Fyn co‐transfected cells, but not untransfected or single‐transfected cells (Fig 1D and E). We also determined the phosphorylation levels of additional regulators of the translational machinery, including the serine/threonine kinase Akt, mTOR (mammalian target of rapamycin), eEF2 (eukaryotic translation elongation factor 2), eIF2α (eukaryotic translation initiation factor 2), eIF4E, 4EBP1 (a repressor of protein synthesis initiation factor 4E), p70S6K (p70S6 kinase), and S6. Remarkably, of these, only phosphorylation of S6 (S235/S236) and its upstream kinase p70S6K were significantly elevated (Figs 1D and E, and EV2D and E). Activation of S6 was also confirmed by finding increased phosphorylation of another activation epitope, S240/S244 (data not shown). Interestingly, activation of the ERK/S6 cascade required co‐transfection of Tau, that is, the cascade was not activated by Fyn alone. Tau may therefore serve as a scaffold for the signal cascade facilitating its translation.

The above results suggested that the Fyn‐mediated increase in Tau involved a pronounced activation of S6 that was independent of mTOR, but occurred in the presence of extensively activated MAPKs. Therefore, considering the previously reported evidence of an mTOR‐independent Ras/ERK/S6 signaling pathway (Roux et al, 2007), we hypothesized a role for Fyn in increasing Tau protein levels by activating the ERK/S6 signaling cascade. We confirmed this role by first using an increasing dose of Fyn that caused massively elevated levels of activated, phosphorylated ERK1/2 and S6, with no changes observed for total levels (Fig EV2F and G). Inclusion of FynKD and FynCA further confirmed a role for Fyn kinase activity in the activation of the signaling cascade (Fig 1F and G).

Next, we determined whether activation of ERK led to an increase in Tau levels (thus bypassing the need for Fyn activation) and whether inhibiting ERK (in the presence of Fyn) blocked or reduced the Fyn‐mediated increases in Tau levels, respectively. Indeed, we found that using the putative MAPK activator phorbol 12‐myristate 13‐acetate (PMA) (Figs 1H and EV3A) or a constitutively active form of ERK (ERK‐MEK‐LA; Fig EV3B and C) activated the ERK/S6 signaling pathway and caused increased Tau, whereas the MAP kinase inhibitor U0126 blocked this pathway, greatly reducing the boosting effect of Fyn (Figs 1I and EV3D). Interestingly, the possible formation of a Tau‐Fyn‐ERK complex was uncovered using a co‐immunoprecipitation assay (Figs 1J and EV3E).

Figure EV3. Effects on Tau expression by manipulating components of the ERK/S6 signaling cascade and formation of a Tau‐Fyn‐ERK complex.

1. Evaluation of the effects of phorbol 12‐myristate 13‐acetate (PMA) on Tau expression and the ERK/S6 signaling pathway in Tau and Tau+Fyn transfected cells. Cells were treated with PMA (100 nM) or DMSO for 24 h.
2. Representative immunoblots showing the effects of overexpression of constitutively active ERK (ERK‐MEK‐LA) on Tau expression. Exo, exogenous MEK‐fused ERK.
3. Quantification of Tau expression levels in (B) (mean ± s.e.m., two‐tailed t‐test; n= 3 per group, t ₍₄₎ = 9.13, ***P = 0.0008).
4. Evaluation of the effects of U0126 on Tau expression and the ERK/S6 signaling pathway in Tau and Tau+Fyn transfected cells. Cells were treated with U0126 (20 μM) or DMSO for 24 h.
5. Evaluation of the Tau‐Fyn‐ERK immuno‐complex in Tau+Fyn transfected cells determined by immunoprecipitation and Western blots.

Data information: ***P < 0.001.

New synthesis of endogenous Tau is induced by Aβo in neurons

We next sought to determine whether activation of the Fyn/ERK/S6/Tau pathway also occurs in neurons, in which Aβo is thought to trigger the relocalization of hyperphosphorylated Tau from the axon to the somatodendritic domain, where Tau is found to accumulate in AD and related tauopathies (Li et al, 2011; Zempel & Mandelkow, 2014). To determine whether endogenous Tau is induced by Aβo, we treated primary hippocampal cultures with increasing doses of this peptide, validating the preparation by Western blotting (Appendix Fig S1A) and transmission electron microscopy (Appendix Fig S1B). We performed a high‐speed centrifugation of the day 1 (D1) preparation (Appendix Fig S1A) to remove fibrils and used the supernatant for treatment. Tau levels were increased (2.2‐fold) for the higher, 3 μM concentration of Aβo, compared to a 60% reduction for MAP2, another member of the microtubule‐associated protein family that, different from Tau, is exclusively localized to dendrites (Fig 2A and B). Different from Aβo, a preparation of scrambled Aβ peptide did not cause increased Tau expression (Appendix Fig S1C and D). Immunofluorescence staining for MAP2 revealed degenerated neurites after incubation with Aβo for 20 min (Appendix Fig S1E and F). However, consistent with a previous study using the same Aβo concentration (Um et al, 2012), the treatment did not cause significant apoptosis (Appendix Fig S1G and H).

Figure 2. Aβo induces endogenous Tau translation.

1. Immunoblots of primary neurons treated with Aβo for 20 min.
2. Quantification of (A) (mean ± s.e.m., n = 4 per group; one‐way ANOVA, Tukey's post hoc test, Tau5, F _(2,9) = 5.374, *P = 0.0379; MAP2, F _(2,9) = 20.83, control versus 3 μM Aβo, ***P = 0.0008, 1 μM versus 3 μM Aβo, **P = 0.0014; AT180, F _(2,9) = 11.35, control versus 1 μM Aβo, *P = 0.0403, control versus 3 μM Aβo, **P = 0.0034; 12E8, F _(2,9) = 5.351, control versus 3 μM Aβo, *P = 0.0297; pS422‐Tau, F _(2,9) = 6.744, control versus 3 μM Aβo, *P = 0.0159).
3. RT–PCR analysis of Mapt transcript levels in Aβo‐treated neurons using two primer pairs (pair #1, amino‐terminus; pair #2, carboxy‐terminus) (mean ± s.e.m., n = 3 per group), one‐way ANOVA, pair #1, F _(2,6) = 3.207, P = 0.1129; pair #2, F _(2,6) = 0.003922, P = 0.9961.
4. Immunoblots for Tau and MAP2 levels in neurons that were treated with 100 μg/ml cycloheximide (CHX) for 0.5 h before subjecting them to 0 or 3 μM Aβo for 20 min.
5. Quantification of (D) (mean ± s.e.m.; two‐tailed t‐test; Tau5, n = 6,6, t ₍₁₀₎ = 2.593, *P = 0.0268; MAP2, n = 3,3, t ₍₄₎ = 7.929, **P = 0.0014; Tau5 in CHX, n = 3,3, t ₍₄₎ = 0.5115, P = 0.6359; MAP2 in CHX, n = 3,3, t ₍₄₎ = 2.778, *P = 0.0499).

Data information: *P < 0.05, **P < 0.01, ***P < 0.001. Source data are available online for this figure.

We then questioned whether Aβo‐mediated Tau overexpression occurs concomitantly with hyperphosphorylation and found that Aβo increased AT180, 12E8 and S422 phosphorylation in neurons (Fig 2A and B). We also again determined Mapt levels, using two different primer pairs against the amino‐ and carboxy‐terminus of the Mapt sequence shared among all transcript variants, which revealed no significant changes in transcript levels (Fig 2C). Tau overexpression was almost completely blocked by cycloheximide, whereas Aβo‐induced MAP2 reduction was not affected (Fig 2D and E), implicating protein translation of Tau in the Aβo‐induced effect.

Aβo induces Tau synthesis specifically in the somatodendritic domain

To visualize specific new protein synthesis, we introduced a recently developed technique, Puro‐PLA, that couples labeling with puromycin (Puro), which is incorporated into the nascent polypeptide chain causing termination, and the proximity ligation assay (PLA; Tom Dieck et al, 2015; Fig 3A). We treated neurons with Aβo (Fig 3B) and paired DakoTau with an antibody against Puro (DakoTau‐Puro pair), which revealed pronounced new protein synthesis of Tau in a pattern that was highly similar to that obtained for MAP2, but only partially overlapped with the axonal mask for total Tau staining, thereby suggesting that Aβo‐mediated Tau synthesis was mainly occurring in the somatodendritic domain (Fig 3C). As a positive control, we paired the rabbit anti‐Tau with a mouse anti‐Tau antibody, which gave a strong PLA signal after Aβo treatment, unlike the negative control (DakoTau/histone 3 pair; Appendix Fig S2A). Pretreatment with anisomycin blocked protein synthesis and hence abrogated the effect of Aβo on the Puro‐DakoTau‐PLA signal (Fig 3D). A similar somatodendritic localization of the PLA signal to that obtained with the Puro‐DakoTau pair was demonstrated by pairing a second Tau‐specific antibody (anti‐carboxy‐terminal Tau; CTau) with Puro (data not shown), and with a Puro‐actin antibody pair (Appendix Fig S2B). The PLA signal in Aβo‐treated neurons, using the Puro‐DakoTau and Puro‐CTau antibody pairs, was increased ~5.4‐ and ~3.0‐fold, respectively, whereas the Puro‐actin signal remained unchanged after Aβo treatment, indicating specificity in the Tau synthesis induced by Aβo (Fig 3E). Similar findings were obtained for Puro‐DakoTau‐PLA when the order of Puro‐tagging was reversed, such that Aβo treatment occurred prior to Puro‐tagging (Appendix Fig S2C and D).

Figure 3. Aβo triggers Tau synthesis in the somatodendritic domain.

1. Scheme of tracking the de novo synthesis of a protein of interest (POI). The Puro (puromycin)‐labeled POI is recognized by both an anti‐Puro (red Y) and an anti‐POI specific antibody (blue Y). PLA (proximity ligation assay) detection is achieved when PLA^plus and PLA^minus oligonucleotides (orange and green squiggles) coupled to respective secondary antibodies (gray Y and black Y) are close enough to be ligated and amplified as rolling circles which are fluorescently tagged (red bars).
2. Experimental setup. Aβo (3 μM); Aniso, anisomycin (40 μM). The arrow indicates a washing step that removes excess Puro from the incubation medium.
3. Counterstaining of Puro‐PLA‐labeled Tau with MAP2 and total Tau in Aβo‐treated neurons. Arrows indicate axons (Tau‐positive and MAP2‐negative) that are also PLA‐negative.
4. Representative Puro‐DakoTau‐PLA images. MAP2, green; PLA, red; DAPI‐labeled nuclei, blue.
5. Quantification of Puro‐PLA images for three antibody combinations (including anti‐carboxy‐terminal Tau, CTau) (mean ± s.e.m.; Puro‐DakoTau, n = 23,23,17 images; Puro‐CTau, n = 42,44,35 images; Puro‐Actin, n = 18,18,18 images per group; images were obtained from 3 to 4 independent experiments; nonparametric Dunn's multiple comparison to Aβo group, ***P < 0.001, ****P < 0.0001, n.s., not significant).
6. Puro‐DakoTau‐PLA fluorescence images of neurons cultured in microfluidic chambers counterstained for MAP2 and total Tau. Cells were seeded on one side of the chamber (chamber 1) with differentiating axons extending neurites through the microgrooves (450 μm) to chamber 2. The indicated chamber was incubated with 1 μM puromycin for 5 min prior to washing and fixation. Microfluidic flow from the treated to the second chamber was prevented by maintaining a smaller volume of medium in the treated side. DAPI‐labeled nuclei (blue).

Data information: Scale bar in (C, D): 50 μm, (F): 100 μm.

Given that Tau protein in healthy mature neurons is enriched in axons (Kosik & Finch, 1987), we questioned where the Tau‐encoding Mapt mRNA was localized. Therefore, we performed a fluorescence in situ hybridization (FISH) assay to probe for Mapt transcripts. Inspection of cultured neurons on coverslips suggested a prominent localization of Mapt transcripts, as indicated by the probe signal, in the soma as well as in neurites, regardless of whether they were MAP2‐positive (dendrites) or MAP2‐negative (axons; Fig EV4A).

Figure EV4. Predominant presence of endogenous Tau transcripts in the somatodendritic compartment and localization of Fyn protein.

1. Fluorescence in situ hybridization (FISH) assay suggesting the presence of endogenous Tau‐encoding mRNA in both the somatodendritic and axonal compartments. Mapt RNA probes were prelabeled with CAL Fluor Red‐590 (in red) with MAP2 counterstained in green (DAPI‐labeled nuclei in blue). Red arrowheads denote the MAP2‐negative neurites that were positive for the probes.
2. Typical immunostaining for Tau (green) and MAP2 (red) of neurons cultured in the microfluidic chamber device (Fig 3F), with the dashed line marking the maximal length that dendrites can assume (DAPI‐labeled nuclei in blue).
3. FISH assay of wild‐type neurons cultured in chamber slides. Red arrowheads denote MAP2‐negative neurites that are positive for the probe. MAP2 (green); CFR 590‐labeled Mapt probes (red); DAPI‐labeled nuclei (blue).
4. Representative immunofluorescence staining for Fyn, Tau, and MAP2 in wild‐type (WT) and Fyn‐knockout (FynKO) neurons (Fyn, red; Tau, green; MAP2, blue).
5. Representative immunofluorescence images of WT neurons cultured in microfluidic chambers and stained for Fyn, Tau, and MAP2. The dashed line indicates where dendrites end and beyond which axons extend. Fyn (red); Tau (green); MAP2 (blue).

Data information: Scale bars: 50 μm.

As an additional experimental system to address subcellular localization, we used a microfluidic device that is ideally suited to discriminate axons (Tau‐positive) and dendrites (MAP2‐positive) based on their length (Fig EV4B). The FISH analysis in these chambers showed a prominent localization of Mapt transcripts in grooves with both dendrites and axons (Fig EV4C), which is supported by previous findings of abundant MAPT mRNAs in the somatodendritic compartment of neurons from human brain (Kosik et al, 1989). To further establish the source of newly synthesized Tau, we combined the microfluidic chambers with the Puro‐PLA reaction. By treating the two isolated chambers of the device separately with Puro followed by PLA labeling, we again discovered a strong Tau synthesis signal in cell bodies and proximal dendrites (chamber 1), but not in axons (chamber 2) (Fig 3F).

Taken together, these findings demonstrate that, whereas Mapt transcripts are ubiquitously present, Aβo induces the local translation of Tau in the somatodendritic domain.

Aβo causes somatodendritic activation of Fyn/ERK/S6 signaling

Numerous studies have highlighted a role for Fyn in mediating Aβ toxicity (Larson et al, 2012; Um et al, 2013, 2012). Fyn has also been implicated in local protein translation in oligodendrocytes (Wake et al, 2011; White et al, 2008). We therefore aimed to determine whether Fyn has a role in the Aβo‐induced Tau translation in neurons.

Fyn staining in both Tau‐ and MAP2‐positive neurites of neurons cultured on coverslips suggested ubiquitous localization in dendrites and axons (Fig EV4D); this was confirmed using microfluidic chambers (Fig EV4E). Using an antibody (p‐SFK) against phosphorylated Y416, an epitope shared between SFK members (Um et al, 2012), it has been shown that, unlike SFKs such as Src, Yes or Lyn, Fyn is specifically activated by Aβo (Um et al, 2013). Consistent with this, Western blotting suggested elevated p‐SFK levels in Aβo‐treated neurons, whereas total Fyn levels did not differ between groups (Fig 4A and B).

Figure 4. Aβo activates somatodendritic Fyn/ERK/S6 signaling.

1. Immunoblots for the indicated antibodies for extracts from primary neurons incubated with Aβo for 20 min.
2. Quantification of (A) (mean ± s.e.m.; one‐way ANOVA, Tukey's post hoc test; p‐mTOR/mTOR, n = 4,4,4, F _(2,9) = 1.84, P = 0.2138; p‐SFK/Fyn, n = 6,6,6, F _(2,15) = 6.674, control versus 3 μM Aβo, **P = 0.007; p‐ERK/ERK, n = 4,4,4, F _(2,9) = 26.98, control versus 1 μM Aβo, ***P = 0.0004, control versus 3 μM Aβo, ***P = 0.0004; p‐S6/S6, n = 4,4,4, F _(2,9) = 22.82, control versus 1 μM Aβo, ***P = 0.0003, control versus 3 μM Aβo, **P = 0.0033).
3. Immunofluorescence images of non‐treated primary neurons stained for Fyn, p‐SFK and MAP2. Arrows indicate axonal signals (MAP2‐negative), and arrowheads indicate spine‐like structures emanating from MAP2‐positive dendrites (p‐SFK, green; Fyn, red; MAP2, blue).
4. Quantification of the fluorescence intensity of images (Appendix Fig S3C) labeling with the indicated antibodies in Aβo‐treated (1 μM) neurons. (mean ± s.e.m.; p‐SFK, n = 26,27 images, Mann–Whitney test; p‐ERK, n = 30,30 images, Mann–Whitney test; p‐S6, n = 26,27 images, two‐tailed t‐test, t ₍₅₁₎ = 6.886; all images obtained from three independent experiments; ****P < 0.0001).
5. Immunofluorescence images showing localization of both newly synthesized Tau and p‐S6 in the somatodendritic compartment. Neurons were labeled with Puro and treated with Aβo as shown in Fig 3B. Puro‐DakoTau‐PLA was performed with subsequent counterstaining of p‐S6 and MAP2. The lower row shows close‐ups of the inset.

Data information: Scale bar: (C, E): 20 μm. **P < 0.01, ***P < 0.001, ****P < 0.0001. Source data are available online for this figure.

Given that Fyn causes massively increased Tau levels through activation of the ERK/S6 pathway in HEK293T cells, we sought to determine whether Aβo‐induced Fyn activation in neurons also activates this pathway, leading to local Tau synthesis. Not surprisingly, increased levels of phosphorylated ERK1/2 and S6 were observed in neurons at two Aβo doses (1 and 3 μM), whereas mTOR phosphorylation was not induced (Fig 4A and B). We also observed a greater than 19‐fold increase in phospho‐p38 levels and a trend toward increased phosphorylation of the p54 JNK isoform (Appendix Fig S3A and B).

Interestingly, under basal condition, immunofluorescence staining revealed that a higher p‐SFK immunoreactivity was confined to the MAP2‐positive somatodendritic compartment (excluding axons and spines), compared to total Fyn immunoreactivity (Fig 4C). Neuronal activation of the Fyn/ERK/S6 signaling pathway by Aβo was also confirmed by p‐SFK, p‐ERK, and p‐S6 immunostaining (Fig 4D and Appendix Fig S3C). Co‐staining of p‐ERK and p‐S6 with both MAP2 and Tau indicated their localization to the somatodendritic domain (Appendix Fig S4A and B) and, in particular, revealed an additional localization of p‐S6 at synapses as indicated by PSD‐95 co‐staining (Appendix Fig S4C). The distinctive pattern of p‐S6 and S6 immunoreactivity suggests a specific activation of the ribosomal machinery in individual neurons (Appendix Fig S4D). To further assess the association of the translational machinery and Tau synthesis, we co‐labeled p‐S6 with newly synthesized Tau using the Puro‐PLA reaction and found that both the PLA and the p‐S6 signal localized to the somatodendritic compartment (Fig 4E). Together, these data demonstrate that Aβo activates Fyn/ERK/S6 in the somatodendritic compartment, and suggest that Fyn might serve as a direct link between Aβo and Tau induction.

Aβ burden causes Tau accumulation via Fyn/ERK/S6 activation in vivo

To determine the in vivo relevance of the Fyn/ERK/S6 pathway, we explored Aβ‐depositing APP23 transgenic mice that overexpress the human amyloid precursor protein (APP) carrying a familial AD mutation (Sturchler‐Pierrat et al, 1997). Immunofluorescence confirmed that, compared to age‐matched wild‐type brains, old (24–28 months old) APP23 brains presented with a pronounced Aβ burden throughout the cortical (data not shown) and hippocampal areas (CA3) (Fig 5A), and in particular, that p‐S6 immunoreactivity was elevated, with a typical cytosolic localization (Fig 5A).

Figure 5. Aβ burden causes Tau accumulation and Fyn/ERK/S6 activation in vivo .

1. Representative immunofluorescence images showing increased amyloid burden and levels of p‐S6 in brain sections from 25‐ to 28‐month‐old APP23 animals, as shown for the CA3 region, with a larger magnification of the dashed region shown on the right; 24‐month‐old wild‐type mice were included as a control. DAPI was used to label nuclei (blue).
2. Scheme of synaptic fractionation protocol for mouse brain tissue.
3. Immunoblots showing enrichment of hallmark proteins in individual fractions.
4. Immunoblots showing the levels of protein expression in the Cyto and PSD fractions prepared from wild‐type and APP23 mouse brains.
5. Quantification of Western blots in (D) (mean ± s.e.m., two‐tailed t‐test, wild‐type, n = 3 mice; APP23, n = 4 mice; Tau5 in Cyto, t ₍₅₎ = 5.331, **P = 0.0031; Tau in PSD, t ₍₅₎ = 2.918, *P = 0.0331; p‐SFK in Cyto, t ₍₅₎ = 7.389, ***P = 0.0007; p‐SFK in PSD, t ₍₅₎ = 6.123, **P = 0.0017; Fyn in Cyto, t ₍₅₎ = 4.294, **P = 0.0078; Fyn in PSD, t ₍₅₎ = 1.368, P = 0.2296; p‐ERK/ERK in Cyto, t ₍₅₎ = 7.536, ***P = 0.0007; p‐ERK/ERK in PSD, t ₍₅₎ = 4.59, **P = 0.0059; p‐S6/S6 in Cyto, t ₍₅₎ = 2.292, ^# P = 0.0705; p‐S6/S6 in PSD, t ₍₅₎ = 0.9439, P = 0.3886).

Data information: ^# P = 0.0705, *P < 0.05, **P < 0.01, ***P < 0.001. Scale bar: (A): 50 μm. Source data are available online for this figure.

We next obtained a cytosolic protein‐enriched fraction (Cyto), a synaptophysin‐enriched non‐PSD fraction, and a PSD‐95‐enriched PSD fraction from brains (Fig 5B and C). We found significantly elevated Tau levels in both the Cyto and PSD fractions in old (24–28 months old) APP23 brains compared to age‐matched wild‐type controls, and Fyn/ERK/S6 signaling was massively activated in both fractions from APP23 animals, with relatively higher activity found in the Cyto compared to the PSD fractions (Fig 5D and E). In young (6‐month‐old) APP23 animals, Tau increases were only observed in the Cyto and S1 (total protein; TP) fractions, and not the PSD fraction, suggesting that Tau may first accumulate in the soma, and with an increasing Aβ burden subsequently translocate into the PSD‐related compartment (Fig EV5A and B). Young mice also showed higher SFK activity in the TP fraction, whereas ERK/S6 activation was only observed in the Cyto fraction.

Figure EV5. Early Tau accumulation in APP overexpressing mouse brain.

1. Representative Western blots showing levels of protein expression in Cyto, PSD, and TP fractions (described in Fig 5B) prepared from 6‐month‐old wild‐type and age‐matched APP23 mouse brains.
2. Quantification of results in (A) (mean ± s.e.m., n = 4 per group; two‐tailed t‐test, Tau in Cyto, t ₍₆₎ = 0.8534, ***P = 0.0001; Tau in PSD, t ₍₆₎ = 0.7583, P = 0.477; Tau in TP, t ₍₆₎ = 2.593, *P = 0.0411; p‐SFK in Cyto, t ₍₆₎ = 0.4896, P = 0.6418; p‐SFK in PSD, t ₍₆₎ = 2.741, *P = 0.0337; p‐SFK in TP, t ₍₆₎ = 2.147, ^# P = 0.0755; Fyn in Cyto, t ₍₆₎ = 0.1885, P = 0.8567; Fyn in PSD, t ₍₆₎ = 1.914, P = 0.1041; Fyn in TP, t ₍₆₎ = 0.2015, P = 0.847; p‐ERK/ERK in Cyto, t ₍₆₎ = 3.674, *P = 0.0104; p‐ERK/ERK in PSD, t ₍₆₎ = 3.544, *P = 0.0122; p‐ERK/ERK in TP, t ₍₆₎ = 0.5342, P = 0.6124; p‐S6/S6 in Cyto, t ₍₆₎ = 1.649, P = 0.1502; p‐S6/S6 in PSD, t ₍₆₎ = 0.03538, P = 0.9729; p‐S6/S6 in TP, t ₍₆₎ = 0.443, P = 0.6733).

Data information: ^# P = 0.0755, *P < 0.05, ***P < 0.001.

Stereotaxic infusion of Aβo into the hippocampus causes Fyn/ERK/S6 activation and Tau overexpression

APP23 mice not only form Aβ, but also express elevated levels of APP from which Aβ is derived. To determine the role of Aβ without the confounding effects of APP, we stereotaxically injected Aβo and artificial cerebrospinal fluid (aCSF), respectively, into the hippocampus of wild‐type mice. Successful infusion of Aβo confined to the hippocampus and neighboring areas was confirmed by immunofluorescence staining using the anti‐Aβ antibody 6E10, which revealed diffuse Aβ immunoreactivity, with many neurons being strongly stained in the somatodendritic domain (Fig 6A). Counterstaining for the active form of S6 (p‐S6) and Aβ indicated that these Aβ‐positive neurons also displayed stronger p‐S6 immunoreactivity than Aβ‐negative neurons in their vicinity (Fig 6B). We then investigated additional components of the Fyn/ERK/S6/Tau pathway and consistently found more and strongly p‐SFK‐, p‐ERK‐, and Tau5‐positive neurons in Aβo‐injected hippocampi compared to aCSF‐injected brains. The immunopositive neurons displayed colocalization with a strong p‐S6 signal in the soma (Fig 6C–E). In summary, this model therefore provides additional support for a role of the Fyn/ERK/S6 signaling pathway in Aβo‐induced Tau accumulation in vivo.

Figure 6. Stereotaxic infusion of Aβ causes colocalization of activated Fyn/ERK and Tau with p‐S6 in the hippocampus.

• A
  Representative immunofluorescence images showing infusion of Aβo in the mouse hippocampus. aCSF served as the injection control. Magnified close‐ups showing individual Aβ‐positive neurons on the right (Aβ, red; DAPI, blue).
• B
  Representative immunofluorescence staining showing that injected Aβo binds to neurons that display higher p‐S6 activity.
• C–E
  Representative immunofluorescence images showing slices from injected brains stained with the indicated antibodies. Staining for p‐SFK (C), p‐ERK (D), and Tau5 (E) identifies neurons that on average display higher p‐S6 activity, with prevalent staining of the cell body. DAPI‐labeled nuclei (blue). All experiments were repeated at least three times.

Data information: Scale bar: 50 μm (A–E).

Overexpression of activated Fyn leads to ERK/S6 activation and Tau accumulation in vivo

The consistent observations made in APP transgenic animals and Aβo‐injected wild‐type mice led us to explore whether Fyn activation itself could activate ERK/S6 and cause subsequent Tau accumulation. We have previously generated a transgenic mouse strain that overexpresses a Myc‐tagged constitutively active form of Fyn (FynCA Tg) driven by the neuronal mThy 1.2 promoter (Xia & Götz, 2014). Owing to the severe premature lethality (lifespan < 7 weeks), we collected brains from FynCA Tg and littermate animals already at 3 weeks of age, followed by subcellular fractionation, yielding TP, Cyto and PSD fractions as described above (Fig 5B and C). The FynCA Tg brains displayed increased levels of endogenous Tau in both the TP fraction (P = 0.06) and the Cyto fraction (P = 0.0432), with the relatively subtle extent of the increase probably reflecting the very young age at which the brains had to be collected. Significant activation of ERK and S6 was also detected, again demonstrating a pronounced role for Fyn in activating the ERK/S6 signaling cascade (Fig 7A and B, Appendix Fig S5). In agreement, the total lysate from Fyn‐knockout (FynKO) primary neuronal cultures exhibited a 64% reduction in Tau levels compared to that of wild‐type neurons (Fig 7C and D). Also, Tau levels were consistently reduced in PSD fractions of cortices dissected from 2‐month‐old FynKO compared to wild‐type mice (Appendix Fig S6A and B). In addition, FynCA‐GFP‐transfected neurons showed a moderate albeit significant increase in Tau expression compared to GFP‐only transfection (Fig 7C and D). These lines of evidence strongly support the notion that Fyn regulates Tau expression via ERK/S6 activation in vivo.

Figure 7. Overexpression of activated Fyn leads to Tau accumulation and ERK/S6 activation in vivo .

1. Protein levels in the TP, Cyto, and PSD fractions (see Fig 5B) prepared from FynCA transgenic (FynCA Tg) and littermate mouse brains determined by Western blotting. Sh, short exposure; lo, long exposure.
2. Quantification shown in (A) (mean ± s.e.m., two‐tailed t‐test, n = 4 per group; Tau/actin in TP, t ₍₆₎ = 2.314, ^# P = 0.06; in Cyto, t ₍₆₎ = 2.555, *P = 0.0432; in PSD, t ₍₆₎ = 1.45, P = 0.1972; p‐SFK/actin in TP, t ₍₆₎ = 7.924, ***P = 0.0002; in Cyto, t ₍₆₎ = 7.601, ***P = 0.0003; in PSD, t ₍₆₎ = 7.15, ***P = 0.0004; Fyn/actin in TP, t ₍₆₎ = 11.68, ****P < 0.0001; in Cyto, t ₍₆₎ = 10.93, ****P < 0.0001; in PSD, t ₍₆₎ = 10.05, ****P < 0.0001; p‐ERK/ERK in TP, t ₍₆₎ = 3.935, **P = 0.0077; in Cyto, t ₍₆₎ = 3.13, *P = 0.0203; in PSD, t ₍₆₎ = 2.345, ^$ P = 0.0574; p‐S6/S6 in TP, t ₍₆₎ = 2.116, P = 0.0787; in Cyto, t ₍₆₎ = 2.699, *P = 0.0356; in PSD, t ₍₆₎ = 1.62, P = 0.1564.
3. Immunoblots for endogenous Tau expression in GFP‐ or FynCA‐GFP‐transfected FynKO or wild‐type primary mouse cortical neurons. Cells were collected 48 h after transfection.
4. Quantification of Tau expression in (C) (mean ± s.e.m., n = 3 per group. One‐way ANOVA, F _(2,6) = 337.2, Dunnett's multiple comparisons, GFP‐transfected FynKO versus wild‐type, ****P = 0.0001; wild‐type transfection with GFP versus FynCA‐GFP, **P = 0.0034).

Data information: ^# P = 0.06, ^$ P = 0.0574, *P < 0.05, **P < 0.01, ***P < 0.001, ****P ≤ 0.0001.

Activity inhibition or genetic deletion of Fyn abolishes Aβo‐induced Tau overexpression via ERK/S6 suppression

To further validate the upstream role of Fyn in the ERK/S6 signaling pathway in neurons, we employed the widely used SFK inhibitor PP2, with the inert analog PP3 serving as a negative control. When neurons were exposed to PP3 or PP2 both for 1 h before and throughout Aβo treatment, Aβo‐induced Tau elevation was prevented by PP2, as shown by immunoblotting. More remarkably, once p‐SFK was blocked by PP2, the downstream ERK/S6 activation completely disappeared, suggesting an upstream position of Fyn in regulating ERK/S6 signaling (Fig 8A and B). However, in PP3‐treated neurons, Aβo caused a significant elevation in Tau protein levels, together with a more highly activated Fyn/ERK/S6 signaling pathway (Fig 8A and B). We also cultured hippocampal neurons from FynKO mice to further validate the role of this kinase in regulating Tau after exposure to Aβo. Whereas, in FynKO neurons, Aβo still caused a reduction in MAP2 at the higher dose of 3 μM (as observed above in wild‐type neurons), it failed to induce Tau overexpression at either the 1 or 3 μM dose (Fig 8C and D). As expected, FynKO neurons failed to express Fyn and also failed to yield a p‐SFK signal, indicating that the signal observed earlier for Src activation (Fig 4A) is an indicator of the specific activation of Fyn by Aβo (Um et al, 2012). In addition, and somewhat surprisingly, 1 μM Aβo caused the mild but significantly higher phosphorylation of both ERK1/2 and S6 in FynKO neurons, whereas this effect was absent at the higher Aβo dose (3 μM) (Fig 8C and D). These data indicate that, although Aβo can trigger ERK‐S6 activation independent of Fyn expression, this Fyn‐independent activation is not sufficient to cause the overexpression of Tau. In summary, this shows that Fyn activity is required to mediate the effect of Aβ on the somatodendritic accumulation of Tau.

Figure 8. Inhibition of activity or genetic deletion of Fyn abolishes the Aβo‐induced increase in Tau levels via ERK/S6 suppression.

1. Immunoblots for the indicated antibodies in Aβo‐treated neurons (20 min) in the presence of the Src family kinase inhibitor PP2 (10 μM). PP3 (10 μM) served as a negative control for PP2.
2. Quantification of immunoblots in (A). (mean ± s.e.m.; Tau/actin, n = 6,6,5,6, two‐way ANOVA, PP3 or PP2 treatment, F _(1,19) = 8.465, P = 0.0090, control or Aβo treatment, F _(1,19) = 2.71, P = 0.1161, interaction, F _(1,19) = 15.6, P = 0.0009, Sidak's multiple comparisons test, Control+PP3 versus Aβo+PP3, **P = 0.0040, versus Control+PP2, ***P = 0.0009, Aβo+PP2 versus Control+PP2, P = 0.5603, versus Aβo+PP3, P = 0.9752; p‐SFK/Fyn in PP3, n = 6 per group, two‐tailed t‐test, t ₍₁₀₎ = 1.983, ^# P = 0.00814; p‐ERK/ERK in PP3, n = 6 per group, two‐tailed t‐test, t ₍₁₀₎ = 4.465, **P = 0.0012; p‐S6/S6 in PP3, n = 6 per group, two‐tailed t‐test, t ₍₁₀₎ = 5.089, ***P = 0.0005).
3. FynKO neuronal cultures were treated with Aβo for 20 min and analyzed for levels of the indicated protein by Western blotting.
4. Quantification of Western blots in (C) (mean ± s.e.m., n = 4 per group, one‐way ANOVA, Tukey's post hoc test; Tau, F _(2,9) = 0.4245, P = 0.6666; MAP2, F _(2,9) = 5.846, control versus 3 μM Aβo, *P = 0.031; p‐ERK/ERK, F _(2,9) = 13.35, control versus 1 μM Aβo, **P = 0.0075, 1 μM versus 3 μM Aβo; **P = 0.0032; p‐S6/S6, F _(2,9) = 8.515, control versus 1 μM Aβo, *P = 0.0238, 1 μM versus 3 μM Aβo; *P = 0.014).

Data information: n.s., not significant, ^# P = 0.00814, *P < 0.05, **P < 0.01, ***P < 0.001. Source data are available online for this figure.

Discussion

Neurodegenerative disorders are all characterized by the accumulation of signature proteins that have been linked to the pathogenesis of these diseases; however, what leads to their accumulation is only incompletely understood. To date, in the case of AD, the prevailing view has been that Aβ contributes to the hyperphosphorylation and detachment of Tau from axonal microtubules, allowing it to pass through the axon initial segment, which under physiological conditions acts as a diffusion barrier, and to enter the somatodendritic compartment where it slowly accumulates and becomes insoluble (Li et al, 2011; Zempel et al, 2013; Zempel & Mandelkow, 2014). Our data do not exclude such a relocalization mechanism; rather, they suggest a novel pathomechanism of local Tau synthesis in the somatodendritic compartment (Fig 9). The notion that inhibition of protein synthesis was found to block Aβo‐mediated “mis‐sorting of overexpressed Tau” (Zempel et al, 2013) lends support to our study that in fact presents an alternative mechanism, with Aβo causing local Tau translation in the somatodendritic domain.

Figure 9. Alternative scheme of the somatodendritic accumulation of Tau induced by Aβo via Fyn/ERK/S6 activation.

Neuronal neurofibrillary tangle (NFT) formation is initiated by the transformation of a healthy neuron that progressively accumulates hyperphosphorylated Tau. A conventional view is displayed in Model I: hyperphosphorylated Tau in the axon detaches from the microtubules (MT) and relocalizes to the somatodendritic domain and accumulates, a process that is partly mediated by Aβ. Model II (this work) involves Aβ‐mediated local protein translation of Tau in the somatodendritic domain, mediated by the Fyn/ERK/S6 signaling cascade, a process concomitant with phosphorylation of Tau.

Given that in mature neurons, Tau and MAP2 have long been known to segregate into axonal and dendritic domains, respectively (Kosik & Finch, 1987), it was important to address the localization of their corresponding mRNAs. It has previously been shown that, unlike Map2 mRNA, which is confined to the soma and dendrites in young primary neuronal cultures, Mapt mRNA localizes to the soma and the proximal axon (Litman et al, 1993). The presence of Mapt in the soma was validated in our study; we additionally detected Mapt mRNAs in dendrites (identified based on MAP2‐positivity) using FISH, as well as in MAP2‐negative neurites that we considered to be axons. This localization is supported by previous evidence showing abundant MAPT mRNAs in the somatodendritic compartment of human neurons (Kosik et al, 1989). The FISH analysis also supports our observation of newly synthesized Tau protein in the somatodendritic compartment, demonstrated by us in two complementary systems, dispersed hippocampal cultures and neurons cultured in microfluidic systems, in combination with the Puro‐PLA assay. It remains a possibility that in AD, the axonal transport machinery becomes overloaded by Tau accumulating in the somatodendritic compartment, thereby leading to further sequestration and accumulation of Tau in this compartment.

Fyn is a master regulator that interacts with many proteins in the central nervous system and peripheral tissue via its SH2 and SH3 domains, providing a point of convergence for a host of signaling pathways (Ohnishi et al, 2011). Recent studies have shown that Aβo‐induced Fyn activation causes increased levels of total Tau, although no mechanism for this accumulation has been suggested (Larson et al, 2012). Whereas until now, little was known about how Fyn regulates protein translation in neurons, in oligodendrocytes, RNA‐binding proteins have been shown to release mRNAs after their phosphorylation by activated Fyn, thereby allowing local protein translation (White et al, 2008). Our finding reveals that oligomeric forms of Aβ activate a signaling pathway that involves Fyn, a MAPK (ERK), and S6, causing local translation of the axonally enriched protein Tau in the somatodendritic domain, thereby linking the two molecules that accumulate in AD brains, Aβ and Tau, via the kinase Fyn. In support of this, Fyn was shown to mediate the Aβo‐induced phosphorylation of the translational regulator eEF2 in neurons (Um et al, 2013).

In two cellular systems (primary neurons and HEK293T cells), by both activating and inhibiting MAPK and Fyn, we identified the Fyn/ERK/S6 pathway as being crucial in mediating elevated Tau levels. This analysis was complemented by in vivo models.

1. By fractionating brains of Aβ‐forming APP23 mice, we revealed an activation of the Fyn/ERK/S6 pathway in the cytoplasmic and, importantly, also the PSD fraction. In support of our observations, total endogenous Tau levels in CSF are threefold increased in APP23 mice at a stage when Aβ pathology becomes prominent (Maia et al, 2013). Also, in a second AD mouse model (APP/PS1), increased levels of Tau and pY18‐Tau have been reported in the PSD‐containing fraction (Larson et al, 2012). This even extends to human patients, where Tau levels in the CSF of AD patients, but not patients with frontotemporal dementia (another tauopathy), rise steeply pointing at Aβ as a driver of Tau increases in AD (Hampel et al, 2004; Olsson et al, 2005).

2. Stereotaxic injection of Aβo into the hippocampus of wild‐type mice revealed a prominent increase in Fyn, ERK, and S6 activity, as well as an increase in endogenous Tau in neurons to which Aβ had bound or which had taken up Aβ.

3. By analyzing fractions from very young FynCA transgenic mice that express a constitutively active form of Fyn, elevated levels of active Fyn, ERK, and S6, as well as increased Tau, were observed in three fractions. Interestingly, it has been reported in Tau transgenic mice that an ERK inhibitor not only reduced the levels of hyperphosphorylated, but also total human Tau, suggesting a role for ERK in regulating Tau expression (Le Corre et al, 2006). Together, these findings support the notion that Aβ activates the Fyn/ERK/S6 pathway to cause Tau translation.

MAP kinases and the ribosomal protein S6 have been implicated in the general regulation of new protein synthesis (Kelleher et al, 2004; Pende et al, 2004), with strong evidence that S6 activation occurs downstream of MAPK and requires Fyn activity (Salmond et al, 2009). In agreement, we show that Tau translation via MAPK and S6 activation requires Fyn activity in both HEK293T cells and primary neurons, albeit via an mTOR‐independent mechanism. Interestingly, activation of the ERK/S6 cascade required co‐transfection of Tau, that is, the cascade was not activated when Fyn was transfected alone, suggesting some form of scaffolding function for Tau. This is not surprising, as Tau is known to exert such a function and interact with a range of different proteins (Klein et al, 2002; Sharma et al, 2007; Morris et al, 2011). More specifically, an increased association between Tau and ribosomes has been reported in AD compared to control brains (Meier et al, 2016), as well as the presence of two new ERK2 docking motifs in Tau that regulate ERK binding and Tau phosphorylation (Qi et al, 2016). Together with our finding that Tau, Fyn, and ERK1/2 can be co‐immunoprecipitated, this would suggest that Tau serves a scaffolding function, allowing it to participate in the feedback regulation of its own synthesis. S6 kinase (and the mTOR pathway) has been shown to be also activated in progressive supranuclear palsy and corticobasal degeneration, two tauopathies characterized by pronounced glial pathology (Gentry et al, 2016). Furthermore, genetic reduction of S6 kinase in mice was found to improve synaptic plasticity and spatial memory deficits and reduce Aβ and Tau accumulation (Caccamo et al, 2015).

In AD, Tau is massively phosphorylated, particularly at serine and threonine residues. By phosphorylating MAPKs including ERK, not only is the protein translational machinery activated, but the newly synthesized Tau also becomes phosphorylated at multiple residues, initiating a vicious cycle (Fig 9). As a consequence, Fyn increases not only tyrosine phosphorylation of Tau, but also serine/threonine phosphorylation at many epitopes, indicating that Fyn activation, by regulating both Tau synthesis and phosphorylation, can potentially enhance Tau pathology in an AD context.

Considering that Tau accumulates in dendritic spines under conditions of neuronal activation (Frandemiche et al, 2014), together with the role that local protein translation has in the dendrite (Tom Dieck et al, 2014), and the fact that Fyn regulates activity‐dependent trafficking (Brigidi et al, 2015), it is plausible that Fyn and local protein synthesis are tightly coupled via Tau. Our data suggest a role for Tau synthesis in the somatodendritic domain in addition to the relocalization mechanism of hyperphosphorylated Tau from the axon to this compartment. Notably, we reveal for the first time that Aβ directly triggers endogenous Tau overexpression via protein translation in vitro and in vivo, via activation of the Fyn/ERK/S6 signaling pathway in the somatodendritic domain. These findings significantly advance our knowledge of the fundamental role of protein translation in transducing an extracellular toxic signal (Aβ) into an intracellular pathology (Tau), both of which are defining features of AD. With disease progression, Tau levels increase, mediating the toxic effects of Aβ in the dendritic compartment, and thereby establishing a feedback loop (Ittner et al, 2010; Roberson et al, 2011). Whether de novo protein synthesis mediated by Fyn also occurs in neurodegenerative diseases other than AD remains to be determined.

Materials and Methods

Animals

Wild‐type (WT) C57BL/6 mice, Fyn‐knockout (FynKO) mice (Stein et al, 1992), Fyn transgenic mice (FynCA Tg) expressing a constitutively active Y531F mutant form of human Fyn (Xia & Götz, 2014), and APP23 mice expressing the human APP transgene containing the Swedish mutation (KM670/671NL) (Sturchler‐Pierrat et al, 1997) were used in this study. Mice were maintained on a 12‐h light/dark cycle with ad libitum access to food and water and were housed 3–5 mice per cage. All experiments were carried out with ethical approval from the University of Queensland Animal Ethics Committee (approval numbers QBI/027/12/NHMRC and QBI/412/14/NHMRC).

Primary cell culture and cell lines

Hippocampal neurons from WT and FynKO mouse embryos were obtained at embryonic day 17 (E17) and plated in 12‐well plates on coverslips coated with poly‐D‐lysine (PDL) at a density of 80,000 cells/well (Fath et al, 2009). Neurobasal medium (Thermo Fisher Invitrogen) was used as the plating medium, supplemented with 5% fetal bovine serum (FBS; Hyclone), 2% B27, 2 mM GlutaMAX (Thermo Fisher Invitrogen) and 50 U/l penicillin/streptomycin (Thermo Fisher Invitrogen). The medium was changed to serum‐free Neurobasal medium 24 h post‐seeding, and half the medium was changed twice a week. Microfluidic chambers (SND450, Xonamicrofluidics) were dropped onto the PDL‐coated coverslips to form non‐plasma bonding with light pressure. Dissociated hippocampal neurons were suspended at a 4–5 × 10⁶ cells/ml density and seeded into one side of the device chamber. The medium was supplemented 10 min after the cells had settled, and half of it was changed twice a week. Mature neurons [days in vitro (DIV) 14–28] were subjected to treatments before collection or fixation, except that neurons grown in microfluidic chambers were treated on DIV9‐11 due to the relatively high cell density. For transfection of pEGFP‐N1 and FynCA‐GFP into primary murine cortical neurons, electroporation was performed using the Nucleofector™ 2b device and the Mouse Neuron Nucleofector Kit (VPG‐1001). Neurons were collected two days after transfection.

The human embryonic kidney (HEK293T) cell line has been authenticated and tested for mycoplasma contamination. Cells were cultured in Dulbecco's modified Eagle's medium (DMEM) (Thermo Fisher Invitrogen) supplemented with 10% FBS, 1% penicillin and l‐glutamine at 37°C, 5.0% CO₂ with saturated humidity. Prior to transfections, cells were cultured to 70–80% confluency. Transfection was done using lipofectamine LTX (Thermo Fisher Invitrogen) following the manufacturer's instructions. Cells were collected in 1× radioimmunoprecipitation assay (RIPA) buffer (Cell Signaling Technology) supplemented with a protease/phosphatase inhibitor cocktail (Cell Signaling Technology) or 2× Laemmli sample buffer (Bio‐Rad). They were then thoroughly lysed by sonication and heat denatured at 95°C for 5 min before loading onto gels for analysis.

Vectors and reagents

Tau‐V5 and Fyn‐Myc constructs were generated from pENTR donor vectors that contained full‐length human 2N4R Tau (hTau40, 441 aa) and full‐length human Fyn as described previously (Xia et al, 2015). Mutagenesis of Tau to generate P301L‐Tau and Fyn to generate FynKD (K299A, kinase dead) and FynCA (Y531F, constitutively active) constructs were performed following the manufacturer's instructions (Q5 site‐directed mutagenesis kit, New England Biolabs, E0554S). The pEGFP‐N1 plasmid was obtained from Clontech Laboratories (Catalog #6085‐1). GFP‐tagged FynCA (FynCA‐GFP), and N‐terminally V5‐tagged and GFP‐tagged Tau were generated using the Gateway cloning method (Thermo Fisher Invitrogen) by cloning Tau cDNA into the pcDNA6.2/V5‐DEST and pcDNA6.2/N‐EmGFP‐DEST vectors. The pCMV‐myc‐ERK2‐L4A‐MEK1_fusion plasmid (Addgene #39197) and the pCMV‐myc‐ERK2‐MEK1_fusion plasmid (Addgene #39194) were kind gifts from Dr. Melanie Cobb. All plasmid vectors were prepared using a plasmid DNA purification kit (Qiagen, Cat # 27106) prior to transfection. U0126 (#9903) was obtained from Cell Signaling Technology. Cycloheximide (CHX) (C7698), puromycin (Puro) (P8833), actinomycin D (ActD) (A9415), phorbol 12‐myristate 13‐acetate (PMA) (P8139), and anisomycin (A9789) were purchased from Merck Sigma Aldrich. 1‐tert‐Butyl‐3‐(4‐chlorophenyl)‐1H‐pyrazolo[3,4‐d]pyrimidin‐4‐amine (PP2) (ab120308) and 4‐amino‐1‐phenyl‐1H‐pyrazolo[3,4‐d]pyrimidine (PP3) (ab120617) were obtained from Abcam.

Puromycin‐proximity ligation assay (Puro‐PLA)

A technique that combines PLA with puromycin (Puro) labeling was recently developed to trace newly synthesized target proteins in situ (Tom Dieck et al, 2015). To detect newly synthesized Tau protein, we carried out the Puro‐PLA reaction using an anti‐puromycin antibody and anti‐Tau antibody as the primary antibody pair, followed by detection using Duolink reagents (Sigma Duolink In Situ Orange Starter Kit Mouse/Rabbit DUO92102) according to the manufacturer's instructions (see diagram of Puro‐PLA labeling in Fig 3A). For puromycin labeling, neurons were incubated with or without 1 μM puromycin for 5 min before or after treatment with Aβ oligomers as shown in the scheme. Puromycin was then immediately washed off using prewarmed medium before the next step. Neurons were pre‐incubated with 40 μM anisomycin for 30 min before the addition of puromycin in the protein synthesis inhibitor group.

Cells were then fixed with 4% paraformaldehyde (PFA) and washed, followed by permeabilization with 0.3% Triton X‐100 in PBS. They were then blocked in blocking buffer and incubated overnight with the anti‐puromycin antibody (Merck Millipore, MABE343, 1:1,000 dilution) and an anti‐Tau rabbit polyclonal antibody (Dako, A0024, 1:1,000 dilution). The same anti‐Tau rabbit antibody was paired with the Tau5 mouse monoclonal antibody (Merck Millipore, MAB361, 1:1,000 dilution) as a positive control, and an anti‐histone 3 antibody (Cell Signaling Technology, #3638, 1:500 dilution) as a negative control to reveal the lack of colocalization. An anti‐actin rabbit (Merck Millipore, #04‐1116, 1:1,000 dilution) and an anti‐C‐terminal Tau rabbit (Abcam, ab150736, 1:1,000 dilution) antibody were also used to pair with anti‐puromycin antibody in PLA reactions. Probes were applied, followed by ligation and signal amplification using the standard protocol. Total Tau expression was visualized by directly applying an Alexa Fluor 488‐conjugated goat anti‐rabbit secondary antibody at room temperature for 1 h (Thermo Fisher, A‐11008, 1:1,000 dilution) after the PLA reaction. Neuronal dendrites were counterstained with chicken anti‐MAP2 antibody (Abcam, ab92434, 1:1,000 dilution) and an Alexa Fluor 647‐conjugated goat anti‐chicken secondary antibody (Thermo Fisher, A‐21449, 1:1,000 dilution). Alexa Fluor 488‐conjugated anti‐phospho‐S6 antibody (Cell Signaling Technology, #4803, 1:100 dilution) was also used for counterstaining after Puro‐PLA application. Coverslips were then stained with DAPI and mounted for imaging.

Fluorescence in situ hybridization (FISH) analysis

A mixture of RNA probes (46 oligonucleotides) against the target region (3,233–4,908 bp) of the mouse Mapt transcript variant 1 (NM_001038609.2) was designed using Stellaris Probe Designer to hybridize with all Mapt transcript variants (probe sequences are listed in Table EV2). Custom oligonucleotides were synthesized and conjugated with the CAL Fluor Red 590 (CFR 590) dye that excites at 566 nm and emits at 588 nm (Biosearch, Stellaris RNA FISH probes). FISH analysis in neurons was performed following a standard protocol provided by Stellaris. In brief, cells were fixed in 4% PFA, permeabilized with 70% ethanol at 4°C and incubated with the FISH probe and Tau5 and MAP2 primary antibodies in hybridization buffer at 37°C in the dark. The cells were then washed, Alexa Fluor 488‐labeled goat anti‐rabbit and Alexa Fluor 647‐labeled goat anti‐mouse IgG secondary antibodies (Thermo Fisher, 1:500 dilution) were applied, and after another round of washing, the cells were mounted using mounting solution (Vectorshield H‐1000), and DAPI was used to visualize the nuclei. Fluorescence images were captured with a 20× or 63× objective on a Zeiss LSM710 laser scanning confocal microscope.

Aβ oligomer preparation

1,1,1,3,3,3‐hexafluoro‐2‐propanol (HFIP)‐treated amyloid‐β_1‐42 peptide (JPT, SP‐Ab‐07_0.5) was resuspended at 5 mM in freshly opened DMSO as monomers, with sonication to facilitate suspension. Aβ peptides were mixed thoroughly with cold F12 medium at a final concentration of 100 μM and then incubated at 4°C for 24 h to allow oligomerization or incubated at room temperature for 7 days to allow fibrillization (Stine et al, 2003). For subsequent experimentation, high‐speed centrifugation was performed to obtain Aβ oligomers (24‐h incubation). Incubations with a scrambled amyloid‐β_1‐42 peptide (AIAEGDSHVLKEGAYMEIFDVQGHVFGGKIFRVVDLGSHNVA) (GenicBio, A‐42‐S‐1) were processed in parallel. The concentration of Aβ oligomers was expressed in monomer equivalents. To validate the Aβ composition from the preparations, samples (monomers, oligomers and fibrils) were loaded onto Mini‐Protein TGX precast gels (Bio‐Rad), followed by Western blotting using the anti‐Aβ antibody 6E10 (Covance, SIG‐39300, 1:2,000 dilution). For the Aβ preparation used in the stereotaxic injection, the HFIP‐treated Aβ peptide was resuspended at 5 mM in artificial cerebrospinal fluid (aCSF) (8.66 g NaCl, 0.224 g KCl, 0.206 g CaCl₂·2H₂O, 0.163 g, MgCl₂·6H₂O, 0.214 g Na₂HPO₄·7H₂O, 0.027 g NaH₂PO₄·H₂O in 1L pyrogen‐free sterile water) followed by sonication. Diluted Aβ (100 μM in aCSF) was obtained at 4°C for 24 h, with oligomeric species validated by immunoblotting (data not shown). aCSF served as the control for Aβ injections. For the analysis by transmission electron microscopy, a drop of sample was placed onto a glow‐discharged carbon‐coated grid and allowed to adsorb for 2 min. The grid was then dipped into UHQ water to remove the majority of PBS before being negatively stained with 1% ammonium molybdate at pH 7. The grids were viewed in a JEOL 1010 transmission electron microscope operated at 80 kV, and images were captured using an Olympus Soft Imaging Veleta digital camera.

Subcellular fractionation

Subcellular fractionation was performed with slight modifications of a previously described method (Milnerwood et al, 2010). Male APP transgenic animals were used in this study. FynCA mice were 3 weeks old, and both genders were used. Brain tissue was first homogenized on ice in sucrose buffer (0.32 M sucrose, 10 mM HEPES, pH 7.4), and the homogenate was centrifuged at 1,000 g for 10 min at 4°C, yielding the supernatant fraction and the nuclear enriched pellet. An aliquot of supernatant was taken for further analysis as the total protein (TP) fraction, with the remainder being centrifuged at 14,000 g for 20 min at 4°C to obtain the crude synaptosomal fraction P2 and the cytosolic protein (Cyto) enriched supernatant. The P2 pellet was washed twice with wash buffer (4 mM HEPES, 1 mM EDTA, pH 7.4) by resuspension and centrifugation at 12,000 g for 20 min at 4°C and then resuspended in buffer A (20 mM HEPES, 100 mM NaCl, 0.5% Triton X‐100, pH 7.2). After rotation at 4°C for 1 h, the suspension was centrifuged at 12,000 g for 20 min at 4°C to yield the non‐PSD fraction containing extrasynaptic proteins. The resultant pellet was washed twice in the wash buffer and resuspended in buffer B (20 mM HEPES, 0.15 mM NaCl, 1% Triton X‐100, 1% SDS, 1 mM dithiothreitol, 1% deoxycholate, pH 7.5) for 1 h at 4°C, followed by centrifugation at 10,000 g for 20 min at 4°C to obtain the PSD fraction containing synaptic proteins. All buffers were freshly supplemented with protease and phosphatase inhibitor cocktail (Cell Signaling Technology) prior to use, and fractions were stored as aliquots at −80°C.

Western blotting

Equal amounts of protein were loaded (10–50 μg depending on the assay) and resolved by SDS–PAGE in Tris‐glycine‐SDS buffer (Bio‐Rad), followed by transfer onto low fluorescence polyvinylidene fluoride (LF‐PVDF) membranes (Bio‐Rad). The membranes were blocked with 5% non‐fat milk in Tris‐buffered saline with 0.05% Tween‐20 (TBS‐T) or Odyssey blocking buffer TBS (LICOR) for 1 h, and then reacted with primary antibodies in TBS‐T with 2% skimmed milk or Odyssey blocking buffer (LICOR) overnight at 4°C under gentle rocking. Primary antibodies are listed in Table EV1. After washing, the corresponding horseradish peroxidase (HRP)‐conjugated secondary antibodies (Thermo Fisher, 1:10,000 dilution) or IR Dye 680RD/800CW secondary antibodies (LICOR, 1:10,000 dilution) were added for 1 h at room temperature under constant agitation. After washing, the membrane was either probed with enhanced chemiluminescence (ECL) Western blotting substrate (Bio‐Rad) and with X‐ray film audioradiography for detection of luminescence (FujiFilm), or scanned in the Odyssey Fc Imaging system (LICOR) for detection of an infrared signal. Where many antibodies were employed, the blots were cut into stripes before probing; however, to confirm the specificity and lack of background staining, representative full immunoblots are provided for brain lysates for antibodies frequently used throughout this study (Appendix Fig S5).

Stereotaxic surgery and microinjections

Mice were anesthetized in an induction chamber with an oxygen/isoflurane mixture (Laser Animal Health, Pharmachem). The pedal reflex was used to monitor anesthesia before placing the mouse in the stereotaxic frame (Kopf Instruments) that was fitted with a mask supplying a continuous flow of oxygen/isoflurane mixture (1 l/min oxygen, 1.5 ml/min isoflurane). Once the animal was deeply anesthetized, a 1.5 cm incision was made to expose the skull. For each microinjection, a small hole was pierced (~0.2 mm width) at the correct coordinates (A‐P: −2.50 mm [bregma]; M‐L: ± 2.0 mm [bregma]; D‐V: −1.80 mm [skull surface]). A pulled glass pipette (GC100TF‐15, Harvard Apparatus) was prefilled first with mineral oil and then with injection solution and fitted into a Nanoject II injector (Drummond 7 Scientific) before being slowly inserted vertically until it reaches its target. Once in the target, after a 2 min waiting interval, one injection of 1 μl Aβ solution (100 μM in aCSF) was delivered into the right brain hemisphere and 1 μl of aCSF into the left hemisphere. Each injection consisted of 15 pulses of 69 nl spaced out by 30 s. Two minutes after the last pulse, the pipette was gently pulled out, and the incision was closed with surgery suture silk. Animals were kept anesthetized in the induction chamber for one hour until they were first perfused transcardially with PBS. Brains were dissected and post‐fixed overnight in 4% PFA at 4°C. Consecutive 40 μm coronal sections spanning the hippocampus including the injection sites were obtained for each animal using a vibratome (VT1000s, Leica Microsystems). Free‐floating slices were stored at −20°C in a cryoprotectant solution (30% ethylene glycol, 30% glycerol, 0.1 M sodium phosphate buffer) until processing for immunofluorescence.

Immunofluorescence

Cells grown on coverslips or microfluidic chamber devices were washed gently with PBS and fixed in 4% PFA, followed by permeabilization with 0.3% Triton X‐100. The cells were then blocked for 1 h in 5% goat serum, followed by incubation overnight with primary antibodies at 4°C. For primary antibodies used in this study, see Table EV1. Secondary antibodies were Alexa Fluor (488, 555, 647)‐labeled goat anti‐mouse, goat anti‐rabbit, and goat anti‐chicken IgG antibodies (Thermo Fisher, 1:500 dilution). Free‐floating slices were rinsed three times in TBS (containing 0.1 mM NaF when phospho‐specific antibodies were employed) for 10 min each, incubated in blocking solution (10% normal goat serum in TBS supplemented with 0.2% Triton X‐100), and placed on an orbital shaker for 2 h. Slices were incubated with primary antibodies at 4°C for 48 h. Alexa Fluor 488‐labeled goat anti‐rabbit and Alexa Fluor 555‐labeled goat anti‐mouse secondary antibodies were used (Thermo Fisher, 1:1,000 dilution). DAPI was used to visualize the nuclei. Fluorescence images were taken with a 20× or 63× objective on a Zeiss LSM710 laser scanning confocal microscope.

Image analysis

Degeneration of neuronal neurites was semi‐quantitatively determined by assessing the average number of independent neurites that contained obvious punctate staining, with neurites being visualized by immunofluorescence staining of MAP2 protein, which is a putative dendritic marker. The number of degenerated neurites in each 20× image was quantified and then normalized to counts of cell nuclei from the same image as visualized with DAPI. The data were finally presented as the number of degenerated neurites per mm² area in each group.

For a general analysis of fluorescence intensity by Image J (FIJI), before measurement, all image intensities were thresholded using the same value for each fluorescence channel within one experiment. The integrated intensity was obtained from multiple images and then averaged to the control group and presented as the relative integrated intensity. For immunoblots, the intensity of blots without obvious bands was considered to be zero (e.g., Fig 8A). The sample size for each group is indicated in the figure legends.

For Western blotting, band intensities were quantified using ImageJ software. They were first normalized to the intensity of an internal control protein and then normalized to the control group as indicated, except that phosphorylated proteins were generally normalized to the amount of total protein instead of the internal control.

Statistical analysis

The molecular and biochemical analysis was performed using a minimum of three biological replicates per condition. The statistical analyses were done using the Prism 6 software (GraphPad). No data points were excluded from the analysis, and randomization of experimental groups was not required. The data distribution was assumed to be normal with similar variance between groups; however, datasets were first assessed with normality tests. Two groups were analyzed with the unpaired Student's t‐test when the datasets were normally distributed; otherwise, nonparametric tests were used with the Mann–Whitney test. Three or more groups were assessed with one‐way ANOVA followed by a post hoc test (Tukey's test, Dunnett's test, or Sidak test depending on the groups) for multiple comparisons, or a nonparametric Dunn's test when the datasets were not normally distributed. Two‐way ANOVA was performed when two independent variables were applied (e.g., Fig 8B). The interaction between two variables was evaluated followed by multiple comparisons. Values are presented as the mean ± standard error of mean (s.e.m.).

Data availability

Additional data that support the findings of this study are available from the corresponding author upon request.

Author contributions

CL and JG conceived and designed the experiments; CL performed all the experiments, imaging, and data analysis; JG provided mice, reagents, and funding; CL and JG wrote the manuscript.

Conflict of interest

The authors declare that they have no conflict of interest.

Supporting information

Appendix

Expanded View Figures PDF

Table EV1

Table EV2

Review Process File

Source Data for Figure 1

Source Data for Figure 2

Source Data for Figure 4

Source Data for Figure 5

Source Data for Figure 8

The EMBO Journal (2017) 36: 3120–3138