Lentivirus Tau (P301S) expression in adult amyloid precursor protein (APP)-transgenic mice leads to tangle formation


  • M. Osinde,

    1. Department of Neuroscience, Novartis Institutes for BioMedical Research, Novartis Pharma,
    Search for more papers by this author
    • 1

      These authors contributed equally.

  • F. Clavaguera,

    1. Pathology Institute, Neuropathology, Basel, Switzerland,
    Search for more papers by this author
    • 1

      These authors contributed equally.

  • R. May-Nass,

    1. Department of Neuroscience, Novartis Institutes for BioMedical Research, Novartis Pharma,
    Search for more papers by this author
  • M. Tolnay,

    1. Pathology Institute, Neuropathology, Basel, Switzerland,
    Search for more papers by this author
  • K. K. Dev

    Corresponding author
    1. Department of Neuroscience, Novartis Institutes for BioMedical Research, Novartis Pharma,
    2. Department of Anatomy, Molecular Neuropharmacology, and
    3. Biosciences Institute, University College Cork, Cork, Ireland
      Kumlesh K. Dev, Department of Anatomy, Molecular Neuropharmacology, University College Cork, Windle Building, Cork, Ireland. Tel. + 353 21 4902028; Fax: + 353 21 4273518; E-mail: k.dev@ucc.ie
    Search for more papers by this author

Kumlesh K. Dev, Department of Anatomy, Molecular Neuropharmacology, University College Cork, Windle Building, Cork, Ireland. Tel. + 353 21 4902028; Fax: + 353 21 4273518; E-mail: k.dev@ucc.ie


Aims: In this study, we aimed to investigate the interaction between amyloid- and Tau-associated pathologies to gain further insights into the development of Alzheimer's disease. We examined the formation of neurofibrillary tangles (NFT) in adult mouse brain without the prior overexpression of Tau at embryonic or early post-natal stages to dissociate any developmental mechanisms. Methods: Lentivirus technology was used to examine the effects of overexpressing mutant Tau-P301S in the adult mouse brains of both wild-type and amyloid precursor protein (APP)-transgenic mice. Results: We find that injection of lentivirus Tau-P301S into the hippocampus of adult wild-type mice increases levels of hyperphosphorylated Tau, as early as 3 months post injection. However, no NFT are found even after 13 months of Tau expression. In contrast, the overexpression of Tau-P301S in adult APP-transgenic animals results in the formation of Gallyas-stained NFT. Conclusions: Our current findings are the first to show that overexpression of Tau-P301S in adult mice overexpressing APP, but not wild-type mice, leads to enhanced Tau-related pathology.


Tau is a microtubule-associated protein that polymerizes tubulin and is involved in intracellular trafficking [1,2]. This protein is located in axons under normal conditions, but is hyperphosphorylated and found in cell bodies and dendrites in various Tauopathies including Alzheimer's disease (AD). Neurofibillary tangles (NFT) are composed of filaments of hyperphosphorylated Tau and the temporospatial spreading of NFT in AD correlates well with the cognitive decline [3]. In addition, mutations in the Tau gene are associated with a subset of frontotemporal dementias with parkinsonism linked to chromosome 17 (FTDP-17T) [4–6].

A number of Tau-transgenic mouse models show that overexpression of wild-type Tau results in higher levels of its hyperphosphorylated form (so called the ‘pretangle stage’) as seen in human Tauopathies [7], but no or limited formation of NFT [8–12]. In contrast, robust NFT pathology develops in transgenic mice expressing a FTDP-17T-linked P301L Tau mutant [13,14]. Additionally, abundant filaments made of hyperphosphorylated Tau can be observed in transgenic mice expressing the FTDP-17T-linked P301S Tau mutant [15]. NFT are also found in transgenic mice overexpressing V337M Tau [16] and R406W Tau [17]. Not surprisingly, transgenic mice with double mutations in Tau (G272V, P301S) [18] or triple mutations (G272V, P301L, R406W) [19] also display NFT formation.

In AD pathogenesis, the amyloid cascade hypothesis holds that accumulation of amyloid-β-peptide in the brain is an early and critical event that triggers a cascade of events leading to hyperphosphorylation and somatodendritic segregation of Tau, and formation of NFT [20]. To examine the interaction between amyloid- and Tau-associated pathologies, two approaches have been used: (i) amyloid precursor protein (APP)-transgenic animals have been mated with Tau-transgenic mice [21–23] and (ii) seeding studies with injected Aβ-containing preparations into Tau-transgenic animals [21,24]. Collectively, these studies indicate that Tau-related pathology is augmented in the presence of abnormal levels of Aβ. Although conditional Tau-transgenic animals exist, they have not been crossed with APP transgenics or used in seeding studies. These mice express Tau P301L at an early age starting at P7 and develop NFT [25,26]. Discontinuing Tau expression in these mice in adult stages results in stabilization of neuronal numbers and recovery of memory function but, surprisingly, NFT continue to accumulate [25].

To our knowledge, there has been no study that examines the effects of Aβ and/or APP on the formation of NFT without the overexpression of Tau at the embryonic or early post-natal stage. To dissociate any effects of Tau overexpression during development, we examined the effects of overexpressing Tau-P301S (a FTDP-17T-linked mutant) starting at a much later age in adult brains of wild-type and APP-transgenic animals by starting expression at 6 months of age using lentivirus technology.

Materials and Methods

Lentiviral vector and virus preparation

The vector was constructed based on the backbone of pLL3.7 [27] using standard molecular-cloning techniques. The cytomegalovirus (CMV) promoter was replaced by phospho-glycerate-kinase (PGK) promoter (obtained by polymerase chain reaction amplification of the PGK promoter from pSIREN-Retro Q vector) (BD biosciences, NJ, USA) generating the enhanced green fluorescent protein (EGFP) pLL4.0-PKG lentiviral (LV) transfer vector. The cDNA encoding the isoform (4R0N) [28] of the human Tau mutant P301S was then cloned into the pLL4.0-PGK vector (LV-hTauP301S) by replacing the EGFP with hTauP301S. The viral particles were produced as described previously for both LV-hTauP301S and LV-EGFP [29]. The viral suspension was concentrated by ultracentrifugation and re-suspended in phosphate buffer saline (PBS) with 2% bovine serum albumin.

Analysis of Tau expression in cell cultures

Chinese hamster ovary (CHO) cells were cultured in HAM F12 media supplemented with 10% foetal calf serum and 1% penicillin and streptomycin (Invitrogen, Carlsbad, CA, USA). Cells were plated on glass coverslips in 12-well plates at 1.5 × 105 cells per ml. Two days after, transduction was performed with LV-hTauP301S using one transfecting particle per cell. Three days post transduction, the cells were fixed in 4% paraformaldehyde in PBS (PFA), permeabilized in 0.2% Triton X-100 (TX-100) followed by incubation in blocking buffer (2% BSA and 5% goat serum in PBS) for 2 h at room temperature before addition of primary antibody. Anti-Tau (Tau-5) antibody (Abcam, Cambridge, UK) was used (1:2000) for human Tau detection. Alexa fluor 488 anti-mouse (1:1000) (Molecular Probes, Invitrogen, CA, USA) was used as secondary antibody. Coverslips were mounted with Vectashield mounting media and pictures taken with an ORCA-AG CCD camera (Hamamatsu, Massy, France). Neuronal cortical cultures were prepared from embryonic day 18 Sprague Dawley rats as previously described [30]. The cortices were dissected in cold Hank's buffered salt solution (Invitrogen), and dissociated using trypsin for 10 min at 37°C. Cultures were plated onto poly D-lysine culture dishes (BD biosciences) or chamber slides (BD biosciences) and maintained at 37°C, 5% CO2 in Neurobasal media (Invitrogen) with B27 supplement (Invitrogen). The neuronal cultures were transfected after 7 days in vitro (DIV) and immunocytochemistry was performed 2–3 days after addition of lentivirus as described for CHO cells.

Stereotaxic injection

Lentiviruses (either LV-EGFP or LV-hTauP301S) were stereotaxically injected into the right hippocampus of adult male C57Bl/6 mice (Charles River Laboratories, I'Arbresle, France), APP23-transgenic animals and wild-type littermates of 6 months of age [31]. Viral suspensions (1-μl volume) were injected with a 10-μl Hamilton syringe at a speed of 0.4 μl/min with an automatic injector (Stoelting, Wood Dale, IL, USA), and the needle was left in place for an additional 5 min before being withdrawn. Stereotaxic injections were delivered within the hippocampus with the following coordinates in millimetres: posterior, 1.75; lateral, 1.5; and ventral, 2.3. The posterior and lateral coordinates were calculated from the Bregma and the ventral coordinates were calculated from the skull surface. Animal studies were conducted in accordance with Swiss animal welfare authorities.


At the time points indicated post injection of lentivirus, animals were anaesthetized with sodium pentobarbital and transcardially perfused with saline and 4% PFA. Brains were removed, kept overnight at 4°C in 4% PFA and paraffin-embedded (Medite TPC 15 Duo, Nunningen, Switzerland). Sections of five microns were made with a microtome (Microm HM 340, Volketswil, Switzerland) and placed onto Superfrost glass slides (Menzel Gläser, Braunschweig, Germany). Sections were rehydrated and additionally heated at 90°C for 10–30 min in citrate buffer (Zymed, Invitrogen, Carlsbad, CA, USA) for epitope retrieval for T14 and AT8 antibodies. Permeabilization was performed in 0.2% TX-100 followed by incubation in blocking buffer (2% BSA and 5% goat serum in PBS) for 2 h at room temperature before addition of primary antibodies. Primary antibodies to detect Tau expression were anti-human Tau T14 clone (1:1000) (Zymed) and anti-phosphorylated Tau AT8 (1:1000) (Pierce, Rockford, IL, USA) and to detect EGFP expression was anti-EGFP (1:1000) (Molecular Probes). Secondary antibody anti-mouse IgG conjugated to biotin (1:200) (Vector) was applied 1 h at room temperature. Sections were stained with ABC kit (Vector) and AEC (Sigma, St Louis, MO, USA) following manufacturer instructions. Additional haematoxilin staining was performed to visualize nuclei and general morphology. After mounting (Vectamount, Vector), sections were visualized with Zeiss microscope. Gallyas silver technique, which detects most NFT in human Tauopathies [32–34], was used to detect NFT in injected mouse brains.

Analysis of Tau expression in hippocampal samples

Thirteen months post injection of lentivirus, mice were anaesthetized with sodium pentobarbital and decapitated. Brains were removed, left and right hippocampi were separately isolated and homogenized on ice. Lysis buffer containing 50 mM Tris, 150 mM NaCl, 1% TX-100, 10% Glycerin, protease (Roche, Mannheim, Germany) and phosphatase inhibitor cocktails (Calbiochem, Merck, Nottingham, UK) was added at a ratio of 100 mg tissue/ml buffer and the homogenates were sonicated repeatedly at maximum 50 Watts. After 1-h rotation on a wheel at 4°C, the total homogenates were aliquoted and stored at −80°C until used for Western blotting.

Western blotting

Chinese hamster ovary, neuronal cultures and hippocampal samples were run on NuPAGE 7% or 4–20% Tris-Acetate pre-cast gels (Invitrogen), transferred onto PVDF membrane (Invitrogen) and blocked with 5% non-fat milk powder in Tris-Buffered Saline supplemented with 0.05% Tween-20 (TBST). Membranes were then incubated overnight at 4°C with primary antibodies in 2.5% non-fat milk powder in TBST. Following washing in TBST, membranes were incubated for 1 h in 2.5% non-fat milk powder in TBST with horseradish peroxidase (HRP)-conjugated secondary antibodies. When required, blots were stripped using Restore Western Blot Buffer (Pierce). The primary antibodies were: rabbit polyclonal anti-total Tau, N-Tau5 (1:2000; in-house antibody raised against a Tau peptide, amino acids 1–16) [35], mouse monoclonal anti-human Tau clone HT7 (1:1000; Innogenetics, Gent, Belgium), mouse monoclonal anti-phospho Tau clone AT8 (1:1000; Pierce); anti-human Tau T14 clone (1:1000; Zymed) and mouse monoclonal anti-βTubulin isotype III clone SDL.3D10 (1:2000; Sigma). The secondary antibodies were: goat anti-rabbit IgG (Sigma) and goat anti-mouse IgG (Sigma). Supersignal West Femto (Pierce) was used for HRP detection on Hyperfilm ECL (Amersham, Buckinghamshire, UK).


Distribution of lentivirus-mediated transgene expression

To determine the distribution of protein expression mediated by lentivirus in mouse brain, the expression of an EGFP transgene was evaluated under the control of the PGK promoter (LV-EGFP). When the LV-EGFP was delivered into the hippocampus, EGFP expression spread over a region of approximately 5 mm cubic area (Figure 1A). At the cellular level, EGFP expression occurred in cell bodies of pyramidal layers and the dentate gyrus as well as axons and dendrites (Figure 1A). On comparison, the injection of LV-EGFP into the corpus callosum gave a more extensive distribution which is likely explained by a spread of EGFP throughout the fibre tracts of the corpus callosum (Figure 1B). When injected into the corpus callosum, expression also reached adjacent cortical cells and the pyramidal cells of hippocampal subfield CA1 (Figure 1B). The EGFP expression was additionally found in the ventricular linings, most probably the ependymal cells, when injected into the corpus callosum (Figure 1B).

Figure 1.

PGK-EGFP lentivirus expression in mouse brain. Wild-type B6 mice were injected with 2 μl of lentivirus expressing EGFP under PGK promoter into the hippocampus. After 2 weeks, the animals were perfused with 4% PFA. Fifty micron sections were analysed for EGFP expression by immunostaining using the EGFP antibody. (A) Strong EGFP signal was observed in the injected side of the hippocampus (upper panel). Magnification of the hippocampus is shown (lower panel). (B) Lentivirus was injected in the Corpus callosum showing a large spreading along the fibres tracts and some pyramidal neurones of CA1 as well as ependymal cells of the ventricles. Cortex (c), corpus callosum (cc), CA1/3 field of hippocampus (ca1/3), hippocampus (h), dentate gyrus (dg), ventricle (v), × 2.5 and × 5 magnification. PGK, phospho-glycerate-kinase; EGFP, enhanced green fluorescent protein; PFA, paraformaldehyde.

The LV-hTauP301S drives stable expression of Tau in neuronal cultures and in the hippocampus

Lentivirus-mediated expression of the human TauP301S (LV-hTauP301S) was driven by the PGK promoter, similar to LV-EGFP. The LV-hTauP301S was first tested for its ability to drive expression in vitro. CHO cells transduced with LV-hTauP301S stained positively with an anti-Tau T5 antibody 72 h post transduction (Figure 2A). In addition, rat cortical cultures (E18) transduced at DIV 7 with LV-hTauP301S showed immunoreactivity with a human-specific anti-Tau T14 antibody at 48 h (data not shown) and 96 h post transduction (Figure 2B). Furthermore, Western blot analysis of rat cortical cultures transduced with LV-hTauP301S showed expression of hTauP301S, where the expression level correlated with the amount of lentivirus used for transduction (Figure 2C).

Figure 2.

Lentivirus (LV)-hTauP301S mediates stable expression of Tau in cell cultures. (A) Chinese hamster ovary (CHO) cells were infected with lentivirus TauP301S for 48 h, fixed and incubated with anti-Tau (T5) antibody. No endogenous Tau was detected in non-infected CHO cells while detectable expression was found in the infected cells. (B) Same transduction procedure was applied on primary cortical cultures incubated with anti-human Tau (T14) antibody. (C) Primary cortical culture lysates were analysed on Western blot with anti-human Tau (T14) antibody. Expression levels were proportional to the amount of transducing particles used. The amount of LV-hTauP301S used for transfection is indicated, none (non-transfected), 2 μl and 4 μl.

The LV-hTauP301S was injected unilaterally into the hippocampus of wild-type mice and the expression level of total hTauP301S was detected in lysates prepared from each isolated hippocampus by Western blotting using N-Tau5 (total Tau) and HT7 (human-specific) antibodies (Figure 3). Hippocampi, injected and non-injected side, from the same animals were used as positive and negative controls, respectively. There was an increase in total Tau expression as detected by the N-Tau5 antibody in the injected hippocampi as compared with the non-injected hippocampi. Accordingly, there was a dramatic increase in HT7 immunoreactivity in lysates prepared from injected hippocampi compared with non-injected, indicating the overexpression of hTauP301S. The expression of hyperphosphorylated Tau was also determined using the AT8 antibody (Figure 3). Levels of hyperphosphorylated Tau were markedly augmented in LV-hTauP301S-injected side compared with non-injected side controls. The expression of hTauP301S remained strong as detected by all three antibodies until 13 months post injection compared with non-injected side controls (Figure 3).

Figure 3.

Lentivirus-hTauP301S mediates stable expression of Tau in the hippocampus. Hippocampi of wild-type mice were dissected 13 months post injection and lysates of injected and non-injected side analysed by Western blot. Tau protein expression was measured with N-Tau5 that detects total Tau, HT7 for human Tau detection and AT8 for human hyperphosphorylated Tau. H1 and H2 are hippocampus from two different mice. Beta-tubulin showed equal protein amounts loaded.

Notably, the T14 and HT7 (human-specific) antibodies showed weak immunoreactivity in non-infected neuronal cultures (Figure 2) and non-injected hippocampi (Figure 3), respectively. In the neuronal cultures, the non-transfected and transfected cells were prepared in different wells; the weak staining found in the non-transfected samples is likely due to cross-reactivity of the T14 antibody with rodent Tau (Figure 2). There is a possibility that the weak immunoreactivity found in non-injected hippocampal samples using the HT7 antibody was due to the partial spread of the virus in the mouse brain to the non-injected hippocampus (Figure 3). In light of the finding that all three N-Tau5, HT7 and AT8 antibodies showed increased immunoreactive bands in the LV-hTauP301S-injected side, the increased immunoreactivity is likely a consequence of Tau overexpression and a subsequent increase in phosphorylated Tau levels, compared with the non-injected side.

Overexpression of hTauP301S in adult APP-transgenic mice promotes tangle formation

Histological analysis using the T14 antibody was performed to examine distribution of total hTauP301S expression after injection of LV-hTauP301S in wild-type animals (Figure 4A). Total hTauP301S expression started to appear at 2 months post injection and increased, remaining strong at 7 months (not shown) and 13 months post injection (Figure 4A). Injection of the LV-hTauP301S towards the CA3 region of the hippocampus resulted in the expression of hTauP301S predominantly in the CA3 subfield of the hippocampus and on occasion in the dentate gyrus (Figure 4A). The hTauP301S expression was found in cell bodies and axonal tracts of the mossy fibres in the dentate gyrus. The AT8 antibody was used to determine the distribution of hyperphosphorylated Tau (Figure 4B). Few neurones expressing hyperphosphorylated Tau were found at 2 months post injection, which was abundant at 7 months (not shown) and also 13 months (Figure 4B). Hyperphosphorylated Tau was found predominantly in the areas of high total hTauP301S expression, mainly in the CA3 region. These data were in agreement with Western blotting results (Figure 3) showing that the number of neurones expressing hyperphosphorylated Tau was proportional to those expressing total hTauP301S levels.

Figure 4.

Expression of Tau P301S in APP23 mice drives tangles formation. Images are from wild-type or APP23 mice sacrificed 13 months after injection with lentivirus-hTauP301S. T14 immunostaining (1:1000; × 100 magnification) showed the expression of human Tau P310S in the hippocampus of wild-type (WT) (A) and APP23 (D) mice. Hyperphosphorylated Tau was detected using the AT8 antibody (1:1000; × 100 magnification) in both wild-type (B) and APP23 (E) animals. Inserts in B and E (× 630) show high magnification of AT8-positive neurones in the CA3 region of the hippocampus. Note that more intense cytoplasmic AT8 staining in the APP23 animal. Numerous Gallyas staining (× 400 magnification, CA3 region of the hippocampus) neurofibrillary tangles were found in the APP23 (C) but not in the wild-type (F) mice. The figures shown are representatives of five to six animals for each condition.

To determine synergy between Aβ deposition and hyperphosphorylated or aggregation of Tau, the LV-TauP301S was injected into APP23-transgenic mice, which overexpresses human APP751 with the Swedish double mutation (K670N/M671L) under the control of a neurone-specific Thy-1 promoter [31]. Animals were injected with LV-hTauP301S (and with LV-EGFP for control groups) before onset of plaques at 6 months of age and examined 13 months post injection, where plaque pathology was evident. The extent of Aβ deposition was examined using an antibody specific for Aβ40 (NT12 antibody). Plaques were detected mainly in the cortical area of APP23 mice whereas no staining was observed in the brain of wild-type animals (not shown). The expression and distribution of total human TauP301S (T14 antibody) was similar in wild-type (Figure 4A) and APP23 (Figure 4D) mice injected with LV-TauP301S, but more intense cytoplasmic AT8 staining was observed in APP23 mice (Figure 4E) when compared with wild-type animals (Figure 4B). No or very few NFT were detected using the Gallyas staining in wild-type animals injected with LV-TauP301S, suggesting that Tau pathology in these mice remained in a ‘pretangle stage’ (Figure 4C). In contrast, APP23-transgenic mice injected with LV-TauP301S revealed a high number of Gallyas-positive cells, indicating that formation of NFT is augmented in animals overexpressing APP (Figure 4F). No difference was found in the number of plaques in the hippocampus of APP23-transgenic mice injected with LV-TauP301S compared with those injected with LV-EGFP, indicating that expression of mutated Tau does not influence Aβ deposition (not shown).


To our knowledge, this is the first report showing that the overexpression of Tau starting in adult mouse brain leads to the formation of NFT in APP-transgenic mice but not wild-type mice. The development of NFT was seen at 13 months after expression of LV-TauP301S, occurring at a later age than other mouse models expressing mutated forms of Tau [13–19], possibly due to differences in expression levels and/or the fact that we avoid expression of Tau at embryonic or early post-natal stages, which may increase the rate development of NFT. The findings also show that directed overexpression of TauP301S can result in the development of tangles in the CA3 region of the hippocampus. These results are in agreement with previous reports, confirming that while APP and/or Aβ augments Tau pathology, the overexpression of Tau has little impact on plaque pathology.

The first Tau-transgenic models that overexpressed wild-type Tau showed limited or no NFT pathology despite an elevated level of hyperphosphorylation of Tau, reminiscent of a ‘pretangle’ phenotype [8–12]. In contrast, a number of transgenic animals that overexpress pathogenic Tau mutants have been shown to develop NFT pathology. Two sets of studies, namely, the direct injection of Aβ fibrils and the overexpression of APP, support the notion that Aβ and/or APP can increase NFT formation. In particular, the injection of fibrillar Aβ42 into the cortex and the CA1 region of the hippocampus increases phosphorylation of Tau and the formation of NFT in P301L mice but not in mice expressing wild-type Tau; the NFT are found in cell bodies of the amygdala, where the injected neurones are known to project [24]. To examine the interaction of amyloid- and Tau-associated pathologies, APP-transgenic animals (Tg2576 mice) have been mated with Tau P301L-transgenic mice (JNPL3 mice) [22]. These double-transgenic animals (TAPP mice) develop both plaques and tangles and show that Tau overexpression has no effect on amyloid pathology while APP overexpression enhances NFT formation [22], in agreement with our findings. In TAPP mice, NFT develop by around 6 months of age at an increased number compared with P301L single-transgenic mice [22]. In a triple-transgenic model, the 3xTg-AD mice harbour mutations of APP (Swedish), presenilin PS1 (M146V) and Tau (P301L), and develop plaques in the neocortex at 3 months of age and in the hippocampus by 6 months. NFT appear in the hippocampus at 12 months, then spread to the cortex [23]. In this mouse model, Aβ immunotherapy reduces extracellular Aβ plaques and intracellular Tau pathology. Tau clearance is thought to be mediated by the proteasome and to be dependent on the phosphorylation state of Tau; hyperphosphorylated Tau aggregates are unaffected by Aβ antibody treatment [35]. The regulation of Tau pathology by amyloid is in agreement with our current findings.

In summary therefore, a number of studies have shown that Tau expression (mutant forms) leads to NFT formation and that amyloid pathology increases the levels of NFT. These studies have used mainly transgenic animals where APP and/or Tau are expressed at otongeny. In the current study, we have shown the overexpression of mutant Tau (P301S) in the brains of adult transgenic mice overexpressing APP (but not wild-type mice) leads to NFT formation. Our studies support the notion that APP and/or Aβ can promote Tau-related pathology even in the adult brain, indicating that amyloid-induced NFT formation can be independent of embryonic and/or post-natal developmental mechanisms.


We wish to thank André Schade, Simone Danner and Dorothee Abramowski for technical assistance and Karl-Heinz Wiederhold for advice on histological analysis. We are also grateful to Dr Matthias Staufenbiel for the kind gift of the APP23-transgenic animals and appreciate Dr Peter Frey for insightful discussions. FC and MT are supported by the Swiss National Science Foundation (3100-068328).