Region-specific dendritic simplification induced by Aβ, mediated by tau via dysregulation of microtubule dynamics: a mechanistic distinct event from other neurodegenerative processes

We employed an ex vivo model of AD using organotypic hippocampal slice cultures from APP_SDL transgenic mice in combination with Sindbis virus-mediated expression of EGFP-tagged constructs. We demonstrated previously that this approach permits analysis of cell death and synaptic changes with respect to the functional interaction between Aβ and tau pathology [5]; here we have extended the approach to determine region-specific changes in dendritic morphology. Virus infections were performed at 12 days in vitro (DIV) and slices were analyzed 3 days later (Fig. 1a). At these conditions efficient neuronal expression of EGFP with low toxicity was obtained [21]. We have also shown previously that the expression levels of different EGFP-tau constructs are similar [5, 21]. Virus titer was adjusted to achieve low infection rate in CA1, CA3 and DG, which permitted 3D imaging of single neurons by high-resolution cLSM with minimal overlap of dendritic arbors (Fig. 1b). Image tile stacks were stitched into single neurons and the morphology of whole neurons was reconstructed in 3D by automated digitalization (Fig. 1c) [22]. The detection gain was adjusted to capture the full arbor, including smaller secondary and tertiary dendrites, irrespective of the EGFP-construct used. Comparing neuronal morphology after virus infection in slices prepared from APP_SDL transgenic mice versus B6 controls permitted determination of the effect of the APP transgene on dendritic morphology in the apical (red) and basal (blue) arbor in principal neurons from the different hippocampal regions (Fig. 1d).

For a quantitative assessment of potential changes in dendritic arborization, we determined the total path length and the number of branching points in principal neurons from CA1, CA3 and DG. Expression of EGFP alone on APP_SDL transgenic background induced only slight dendritic simplification, which did not reach significance. In contrast, we observed that EGFP-htau expression on APP_SDL transgenic background induced pronounced and significant dendritic simplification in CA1 neurons as indicated by a reduction of total path length and number of branching points by 38 % and 30 %, respectively (Fig. 2a, left). No change was seen in CA3 pyramidal neurons and granule cells of the DG (Fig. 2a, middle and right). Human tau expression on non-transgenic background did not induce morphological changes, indicating that dendritic simplification in the CA1 region requires the combination of human tau and APP_SDL transgenic background.

CA1 pyramidal neurons receive different inputs in specific subregions of the dendritic tree (Fig. 2b). To determine subregional differences in the induction of dendritic simplification, 3D Sholl analysis [23] was performed on basal and apical branches to quantitate changes in the intersection frequency of dendrites as a function of the distance from the cell body (Fig. 2c, left). On the APP transgenic background we observed a reduction of dendritic complexity in the entire basal part of the tree (Fig. 2c, right). In contrast, in the apical part, dendritic simplification was induced specifically in two segments of the tree. Responsive segments were localized at ~15-25 % and ~60-70 % of total apical length and correlated with regions, where most input from Schaffer collaterals occurs (see Fig. 2b).

Dendritic simplification could be induced by the presence of mutated APP or might require the generation of Aβ through γ- and β-secretase mediated cleavage of APP_SDL. To distinguish between these possibilities, we pretreated the culture with the non-transition state γ-secretase inhibitor DAPT in order to reduce Aβ generation [24]. Treatment with DAPT was started one day before infection with EGFP-htau expressing virus (Fig. 3a, top). DAPT reduced the negative effect of tau and APP on dendritic complexity in CA1 neurons as judged by a significant increase in total path length compared to untreated cultures of the same genotype (Fig. 3a, bottom). Sholl analysis of CA1 neurons revealed that dendritic simplification was completely abolished in basal dendrites and reduced in apical segments (Fig. 3b). The data indicate that dendritic simplification is induced by Aβ rather than mutated APP.

It has previously been shown that Aβ acts via NMDA receptor (NMDAR) activation to induce spine changes and neuron loss [5, 25, 26]. To test whether NMDAR activation is also involved in mediating dendritic simplification, we added the competitive NMDAR antagonist CPP [27] to the cultures one day before infection with EGFP-htau expressing virus (Fig. 4a, top). CPP also abolished the negative effect of tau and Aβ on dendritic complexity in CA1 neurons similar to DAPT (Fig. 4a, bottom, and b), demonstrating that dendritic simplification is mediated by NMDAR activation. It should also be noted that both drugs induced some dendritic simplification in CA3 neurons, which might indicate sensitivity of the CA3 neurons to the drugs themselves.

Data from this study could suggest that Aβ induces microtubule destabilization in a tau-dependent manner, which results in a decrease of dendritic complexity. If true, one would expect that microtubule-stabilizing drugs would prevent dendritic simplification. To test this hypothesis we treated the cultures prior to infection with the microtubule-stabilizing drug epothilone D (EpoD), a brain-penetrant small molecule potential therapeutic candidate for AD [28, 29]. Epothilones may have neuroprotective activity; however dose-dependent neurotoxic effects have also been reported [30]. In pilot experiments we observed loss of neurons in control cultures at concentrations of 1 and 5 nM of EpoD indicating some toxicity for the cells. The number of neurons was not affected at 0.2 nM EpoD, a concentration where still induction of microtubule polymerization was observed in model neurons (data not shown). Thus this concentration was subsequently used for the experiments. Surprisingly, EpoD induced dendritic simplification in control cultures as indicated by significantly decreased total path length and branching compared to the respective untreated cultures (Fig. 5a, Table 2). No obvious differences in the dendritic morphology between htau-expressing APP-transgenic and control cultures were observed, which was – except some simplification in basal dendrites – confirmed by Sholl analysis (Fig. 5b, c).

Since microtubule stabilization did not have any effect in Aβ-producing cultures, we explored whether tau phosphorylation, which is known to affect microtubule dynamics, is at all involved in mediating dendritic simplification. We employed a phosphoblocking (Ala htau) and a phosphomimicking (PHP htau) construct in which 10 AD-relevant sites were modified to alanine to prevent, or to glutamate to simulate, phosphorylation at these residues (Fig. 6a, top) [31]. Surprisingly, expression of Ala htau induced dendritic simplification per se whereas the phosphomimicking PHP htau did not induce morphological changes (Fig. 6a, bottom, Table 1). This is in sharp contrast to the previous observation that PHP tau is the active species to induce cell death, and importantly indicates that the development of dendritic simplification and tau-dependent cell death are mechanistically distinct. The finding that the effect of Ala htau expression closely resembles the impact of EpoD-treatment could suggest that Ala htau induces dendritic simplification by hyperstabilizing dendritic microtubules. To test this hypothesis, we determined the influence of wt htau, Ala htau and PHP htau expression on microtubule stability by determining the ratio of acetylated to total tubulin, since tubulin acetylation is considered to be a marker for microtubule stability [32]. Expression of Ala htau induced a significantly increased ratio of acetylated to total tubulin compared to wt htau or PHP htau expressing cultures (Fig. 6b). This indicates that non-phosphorylatable htau induces dendritic simplification due to the inherent activity to promote microtubule stabilization. It should however be noted that Ala htau expression prevented dendritic simplification in CA1 neurons from APP_SDL transgenic animals (Fig. 6a, bottom), which might indicate that it counteracts Aβ-induced microtubule destabilization in these neurons.

Generally, higher Aβ concentrations are considered to increase rather than decrease tau phosphorylation, but the effect of low concentrations of secreted Aβ on the phosphorylation of human tau is less clear. A previous study using a coculture system has shown that secreted Aβ induces an increased phosphorylation at the AT8 (S202/T205) and the AT180 epitope (T231/S235) [33]. The concentration of Aβ was comparable with that found in cerebrospinal fluid of AD patients, probably in the higher nanomolar range. However, a rough estimate suggests that the amount of Aβ in the brain of APP_SDL transgenic mice that have been used for preparation of the organotypic slices is in the low nanomolar range (L. Bakota, unpublished observation), which reflects more a presymptomatic scenario. To test the hypothesis that low nanomolar concentrations of naturally secreted Aβ cause a change in the phosphorylation pattern of tau that might lead to increased microtubule stabilization, we turned to a cellular model, where PAGFP-tagged human tau is stably expressed in rat PC12 cells (Fig. 6c). The cells were seeded in a defined density, neuronally differentiated with NGF and incubated with a supernatant of APP_Swe-transfected HEK cells. Western blot analysis showed that 3.5 nM Aβ caused a decreased phosphorylation at S262, while some other sites such as T181 and S214 were not affected. Interestingly, S262 is located in the microtubule-binding region of tau and phosphorylation of S262 largely abolishes its interaction with microtubules [34]. Thus, even a small decrease in S262 phosphorylation could induce tau-dependent microtubule stabilization thereby promoting dendritic simplification.

Heterozygous APP_SDL transgenic mice [48] expressing human APP₆₉₅ with three familial AD mutations were used. APP_SDL mice express equimolar concentrations of Aβ40 and Aβ42 already at embryonal age [49]. Genotyping was performed as described previously [5]. APP_SDL transgenic and C57BL/6 J (B6) control animals were maintained and sacrificed according to National Institutes of Health Guidelines and German Animal Care Regulations.

Chemicals were purchased from Sigma (Steinheim, Germany), culture medium and supplements from Sigma and Invitrogen (Darmstadt, Germany), culture dishes and plates from Nunc (Langenselbold, Germany), and membrane culture inserts from Millipore (Billerica, MA, USA). γ-secretase inhibitor N-[N-(3,5-difluorophenacetyl)-1-alanyl]-S-phenylglycine t-butyl ester (DAPT) was purchased from Merck (Darmstadt, Germany). Epothilone D (EpoD) was prepared as previously described [50, 51]. The spectroscopic properties of the compound were identical to those reported in the literature. Compound purity was >95 % as determined by LC-MS and NMR analyses.

Construction of virus was performed as described previously [21]. The following constructs were used for the experiments: pSinRep5-EGFP, pSinRep5-EGFP-352wt htau, pSinRep5-EGFP-352 Ala htau, pSinRep5-EGFP-352 PHP htau. For Ala and PHP htau, respectively, 10 sites (Ser198, Ser199, Ser202, Thr231, Ser235, Ser396, Ser404, Ser409, Ser413, and Ser422) were mutated to alanine to block or to glutamate to mimic phosphorylation at the respective residues [31].

Hippocampal slice cultures were prepared from mice of either sex and processed as described previously [5, 52]. On day 12 in vitro slice cultures were infected with Sindbis virus. For some experiments, slices were treated with 0.5 μM DAPT, 20 μM 3-(2-carboxypiperazin-4-yl)propyl-1-phosphonic acid (CPP), or 0.2 nM EpoD starting on day 11 in vitro. Cultures were fixed at day 3 post infection. For preparation of lysates, cultured hippocampal slices were visually inspected on day 15 in vitro. Slices with similar infection rates were lysed in RIPA buffer as described previously [5].

Confocal high-resolution microscopy of complete principal hippocampal neurons was performed as described previously [22]. Each single neuron was imaged in 8 to 12 overlapping image stacks with voxel size of 0.14 × 0.14 × 0.44 μm in x-y-z directions. 3D reconstruction of whole neurons was performed using Neuromantic software (University of Reading, Reading, UK) in semi-automated mode.

For the determination of dendritic tau levels, non overlapping basal dendritic segments of 2nd or 3rd order of CA1 neurons expressing EGFP-tagged tau constructs were selected. Image stacks between first and last appearance of the respective dendritic segment were projected onto a single layer and mean intensity values in the middle of the process were determined and corrected for the detector gain.

PC12 cells stably transfected to express PAGFP-tagged wt 352 htau were cultured in serum-DMEM in the presence of 250 μg/ml Geneticin as described previously [53]. Cells were plated onto 10-cm collagen-coated TC-plates at 2 × 10⁴ cells/cm². On the next day, medium was replaced with low serum DMEM (DMEM with 1 % (vol/vol) serum) containing 100 ng/ml 7S mouse NGF (Alomone Laboratories). After 4 days, medium was replaced with serum-free medium (NB/B-27, Thermo Fisher Scientific) containing 100 ng/ml 7S mouse NGF and supernatant of Aβ-producing HEK cells (HEK-SW) to yield 0.7 or 3.5 nM of Aβ. Same amounts of supernatant of untransfected HEK cells (HEK-con) were added as control. Cell lysates were prepared 24 h. later as described by [53].

HEK-SW cells (HEK 293-APP_swe; [54]), which were stably transfected with APP carrying the Swedish familial AD mutation resulting in increased Aβ secretion, were obtained from C. Haass (University of Munich) and cultured in DMEM containing 10 % fetal calf serum. As a control, HEK 293FT cells (Life Technologies) were used. Cells were cultured until 60–70 % confluency, medium was exchanged and the supernatant collected 24 h. later. The supernatant was filtered (0.22 μm pore size) and shock frozen in aliquots in liquid nitrogen. The concentration of Aβ_1–40 and Aβ_1–42 was determined by ELISAs according to the manufacturer’s description (EZHS40, EZHS42; Millipore). The ratio of Aβ_1–42 to Aβ_1–40 in the supernatant of HEK-SW cells was 1:4.

Same volumes of lysates were subjected to SDS-PAGE, transferred to Immobilon-P (Millipore) and stained. Immunodetection used the following antibodies: Phosphorylation-independent tau antibodies: HT7 (human tau, mouse; Pierce); phosphorylation-dependent tau antibodies: AT270 (pThr181, mouse; Pierce), p214 (Ser-214, rabbit; Biosource), pS262 (pSer262, rabbit; Biosource). Also used were anti-GFP (rabbit; Invitrogen), anti-tubulin (mouse, DM1A), anti-acetylated tubulin (mouse; 6-11B-1), and anti-actin (mouse, JLA20; Amersham) antibodies. As secondary antibodies, peroxidase–conjugated goat anti–mouse and anti–rabbit antibodies (Jackson ImmunoResearch Laboratories, Inc.) were used. Detection using enhanced chemiluminiscence and quantitation were performed as described previously [17].

Statistical evaluation was performed using one way analysis of variance (ANOVA) with post hoc Fisher's least significant difference (LSD) test for multiple comparisons and two-tailed, unpaired Student’s t test for comparison of the two genotypes at control conditions. For the tau phosphorylation experiments, one-sample Student’s t test was performed. Data are shown as mean and s.e.m. Sholl analysis [23] was performed separately for basal and apical parts of dendritic trees.