Signal transduction pathways implicated in neural recognition molecule L1 triggered neuroprotection and neuritogenesis

Authors

  • Gabriele Loers,

    1. Zentrum für Molekulare Neurobiologie Hamburg, Universität Hamburg, Hamburg, Germany
    Search for more papers by this author
  • Suzhen Chen,

    1. Zentrum für Molekulare Neurobiologie Hamburg, Universität Hamburg, Hamburg, Germany
    Search for more papers by this author
    • 1

      The present address of Suzhen Chen is the Division of Neuroscience, Children's Hospital, Harvard Medical School, Enders 270, 300 Longwood Ave., Boston, MA 02115, USA.

  • Martin Grumet,

    1. Department of Cell Biology and Neuroscience and W. M. Keck Center for Collaborative Neuroscience, Nelson Laboratory, Rutgers State University of New Jersey, Piscataway, New Jersey, USA
    Search for more papers by this author
  • Melitta Schachner

    1. Zentrum für Molekulare Neurobiologie Hamburg, Universität Hamburg, Hamburg, Germany
    Search for more papers by this author

Address correspondence and reprint requests to Melitta Schachner, Zentrum für Molekulare Neurobiologie Hamburg, Universität Hamburg, Falkenried 94, D-20251 Hamburg, Germany.
E-mail: Melitta.Schachner@zmnh.uni-hamburg.de

Abstract

The signal transduction pathways involved in adhesion molecule L1-triggered neuritogenesis and neuroprotection were investigated using the extracellular domain of mouse or human L1 in fusion with the Fc portion of human immunoglobulin G or L1 purified from mouse brain by affinity chromatography. Substrate L1-triggered neuritogenesis and neuroprotection depended on distinct but also overlapping signal transduction pathways and on the expression of L1 at the neuronal cell surface. PI3 kinase inhibitors, Src family kinase inhibitors as well as mitogen-activated protein kinase kinase inhibitors reduced both L1-triggered neuritogenesis and neuroprotection. In contrast, fibroblast growth factor receptor inhibitors, a protein kinase A inhibitor, and an inhibitor of cAMP-mediated signal transduction pathways, blocked neuritogenesis, but did not affect L1-triggered neuroprotection. Proteolytic cleavage of L1 or its interaction partners is necessary for both L1-mediated neuritogensis and neuroprotection. Furthermore, L1-triggered neuroprotection was found to be associated with increased phosphorylation of extracellular signal-regulated kinases 1/2, Akt and Bad, and inhibition of caspases. These observations suggest possibilities of differentially targeting signal transduction pathways for L1-dependent neuritogenesis and neuroprotection.

Abbreviations used
AMP

adenosine monophosphate

Apaf-1

apoptotic protease activating factor-1

BDNF

brain-derived neurotrophic factor

cytoD

cytochalasin D

DAG

diacylglycerol

dATP

2′-deoxyadenosine 5′-triphosphate

ERK

extracellular signal-regulated kinase

FCS

foetal calf serum

FGF

fibroblast growth factor

GABA

gamma-aminobutyric acid

GAPDH

glucose-6-phosphate dehydrogenase

L1

neuronal cell adhesion molecule L1

HBSS

Hanks' balanced salt solution

hL1Fc

extracellular domain of human L1 fused to Fc from human IgG

H2O2

hydrogen peroxide

mL1Fc

extracellular domain of mouse L1 fused to Fc from human IgG

MAP

kinase, mitogen-activated protein kinase

MEK

mitogen-activated protein kinase kinase

MTT

3-(4,5-dimethyl-thiazol-2-yl)-2,5-diphenyl-tetrazolium bromide

NCAM

neural cell adhesion molecule

PLC

phospholipase C

NGF

nerve growth factor

PDGF

platelet-derived growth factor

PI3-K

phosphatidylinositol-3′ kinase

PIP-3

phosphatidylinositol (3,4,5)-triphosphate

PKA

protein kinase A

PLL

poly-l-lysine

PMSF

phenylmethylsulfonyl fluoride

Rp-cAMP

Rp-adenosine-3′,5′-cyclic monophosphorothioate

SDS

sodium dodecyl sulfate

TBS-T

Tris-buffered saline/0.05% Tween 20

The neural cell adhesion molecule L1 (Lindner et al. 1983; Rathjen and Schachner 1984), a member of the immunoglobulin superfamily (Brümmendorf and Rathjen 1994), stimulates neuronal migration and survival, neuritogenesis, fasciculation, myelination in the peripheral nervous system during development (Hulley et al. 1998; Chen et al. 1999; Ourednik et al. 2001; for review, see Hortsch 2003), and synaptic plasticity and regeneration in the adult (Dihnéet al. 2003; Roonprapunt et al. 2003; for reviews, see Fields and Itoh 1996; Schachner 1997). The phenotype of L1-deficient mice and humans carrying mutations in the L1 gene underscore the critical role of L1 for normal development of the nervous system (Dahme et al. 1997; Cohen et al. 1998; Demyanenko et al. 1999, 2001; Kenwrick et al. 2000).

L1 consists of six immunoglobulin (Ig)-like domains at the amino terminal end of the molecule followed by five fibronectin type III homologous repeats, a single transmembrane region and a short intracellular domain (Moos et al. 1988). This intracellular domain contains an ankyrin-binding region (Davis and Bennett 1994), and a neuron-specific sequence, RSLE, which is critical for sorting L1 to axon growth cones (Schaefer et al. 2002; Kamiguchi 2003). This sequence is likely to play an important role in L1-triggered signal transduction pathways during neuritogenesis and neuritogenesis-associated membrane recycling (Kamiguchi et al. 1998; Schmid et al. 2000; Schaefer et al. 2002) along with signalling through basic fibroblast growth factor receptor (Green et al. 1997; Williams et al. 1994; Kiryushko et al. 2004).

Because of the ability of L1 to promote neuritogenesis and neuroprotection, it represents a potential therapeutic agent for treating neurodegenerative diseases in humans. We were therefore interested in whether L1-triggered neuritogenesis and L1-triggered neuroprotection depend on similar or different signal transduction pathways. Although Src family kinase and MAP kinase pathways are required for L1-triggered neuritogenesis (Schaefer et al. 1999; Schmid et al. 2000), nothing is known about the intracellular signal transduction events related to L1-triggered neuroprotection. We observed that overlapping, but also distinct signal transduction pathways mediate neuritogenesis and neuroprotection.

Materials and methods

Cell culture medium, L1 proteins and inhibitors

Cells were cultured in chemically defined serum-free medium containing basal medium Eagle's supplemented with 2 mm l-glutamine (PAA Laboratories, Cölbe, Germany), 1 mg/mL bovine serum albumin (Sigma, Deisenhofen, Germany), 12.5 μg/mL insulin (Sigma), 4 nm thyroxine (Sigma), 100 μg/mL transferrin (Merck Biosciences, Schwalbach, Germanay), 30 nm sodium selenite (Sigma), 0.1 mg/mL streptomycin, 10 U/mL penicillin (PAA Laboratories).

The L1Fc fusion protein was produced and purified as described by Chen et al. (1999) and Haspel et al. (2000). It contains the extracellular domain of human or mouse L1 in fusion with human Fc. Gel filtration of the L1Fc fusion protein preparation showed that L1Fc migrates as a monomer. L1 derived from mouse brain was purified from 7-day-old wild-type animals by affinity chromatography using a monoclonal L1 antibody (Rathjen and Schachner 1984). The Fc-part of human immunoglobulin G (Fc) served as a negative control (Dianova, Hamburg, Germany).

The following inhibitors/inhibitor concentrations were used in this study: FGF receptor inhibitors PD173074 (500 nm; Parke-Davis, Ann Arbor, MI, USA) and SU5402 (50 μm; Merck Biosciences), PI3K inhibitors LY294002 (50 μm; Cell Signaling Technology, Beverly, MA, USA) and wortmannin (25 nm; Sigma), PKA inibitor KT5720 (100 nm; Sigma), Rp-cAMP, an inhibitor of cAMP-mediated signal transduction pathways (100 μm; Sigma), non-selective protein kinase inhibitor K252a (100 nm; Merck Biosciences), PKC inhibitor peptide 20–28 (50 μm; Merck Biosciences), Src kinase family inhibitors PP2 (50 nm; Merck Biosciences) and Herbimycin (5 μm; Sigma), MEK inhibitors U0126 (10 μm; Cell Signaling Technology) and PD98059 (10 μm; Merck Biosciences), PLC inhibitor U73122 (10 μm; Sigma), DAG kinase inhibitor R59022 (10 μm; Sigma), caspase inhibitors z-VAD-fmk (50 μm; Merck Biosciences) and z-DEVD-fmk (50 μm; Merck Biosciences), protease inhibitors GM6001 (10 μm; Merck Biosciences), E64 (10 μm; Merck Biosciences), leupeptin (10 μm; Sigma), pepstatin (1 μm; Sigma) and AEBSF (100 μm; Merck Biosciences), F-actin depolymerizing cytochalasin D (100 nm; Merck Biosciences) and intracellular calcium chelator BAPTA-AM (10 nm−10 μm; Merck Biosciences).

Neuritogenesis

Dissociated cerebellar granule cell cultures were prepared from the cerebellum of 6- to 8-day-old C57BL/6J mice or L1-deficient mice. In this mutant, the expression of L1 was completely abolished by insertion of a tetracycline-controlled transactivator (Gossen and Bujard 1992) into the second exon of the L1 gene (M. Kutsche and M. Schachner, unpublished data). One millilitre of a suspension containing 105 cells/mL was transferred to a 12-well plate (Nunc, Roskilde, Denmark) containing one glass coverslip per well (15 mm in diameter; Hecht, Sondheim, Germany) substrate-coated with poly-l-lysine (PLL; Sigma) followed by either L1 derived from mouse brain (L1; 2 μg/mL), mouse or human L1Fc fusion proteins (10 μg/mL), human Fc (10 μg/mL) or laminin (Sigma; 1 μg/mL) and cultured in chemically defined serum-free medium (see above). For analysis of signal transduction mechanisms, cells were treated 30–60 min after seeding with the inhibitors mentioned above. By then, most of the plated cells had attached to substrate-coated glass coverslips and the inhibitors were not found to interfere with the stabilization of the attachment of the cells. After 22 h of maintenance in vitro, cells were fixed with 2.5% glutaraldehyde in phosphate buffered saline, pH 7.5, and stained with toluidine blue/trypan blue. The length of total neurites per cell and the number of cells with neurites were measured by the KS image analysis system (Kontron, Zeiss, München, Germany). In each experiment, at least 50 cells on each of two coverslips were analysed.

Neuroprotection

Dissociated cerebellar granule cell cultures were prepared from the cerebellum of 6- to 8-day-old C57BL/6J mice and 250 μL of a cell suspension containing 1 × 106 cells/mL were plated into each well of 48-well tissue culture plates (Nunc) substrate coated with PLL followed by L1 (2 μg/mL), mouse or human L1Fc fusion proteins (10 μg/mL) or human Fc (10 μg/mL). The cells were cultured in chemically defined serum-free medium as described above. Cell survival was determined 4 h after plating and set to 100% without additives. Cells were also maintained in the same medium supplemented with 10% foetal calf serum (FCS) and 5% horse serum (Invitrogen, Karlsruhe, Germany) as control. For cell death induced by serum deprivation, inhibitors were added 16 h, and 3, 5 and 7 days after plating. Cell survival was determined at day 8 or every day during the culture period. For induction of cell death, hydrogen peroxide (cell death through oxidative stress; 10 μm; Merck Biosciences) or staurosporine (broad spectrum kinase inhibitor; 500 nm; Sigma) were used. Again, cells were treated with inhibitors 16–24 h after seeding. Four hours after inhibitor treatment hydrogen peroxide or staurosporine were added and the cells were cultured for an additional 24 h before cell death was determined. Viability of cells was assessed in all cases by counting the numbers of calcein AM (Molecular Probes, Leiden, the Netherlands) versus propidium iodide (Sigma) -positive cells. For calcein/propidium iodide staining, cells were treated with 1 μg/mL calcein and 1 μg/mL propidium iodide for 1 h at 37°C. The cells from two randomly chosen areas of a microscopic field (magnification 10 × 20) in each well were counted, and for each experimental value four wells were measured.

For the determination of total caspase activity, cells were treated 2–4 h after induction of cell death or after 8 days in culture under serum-free conditions with the cell-permeable caspase substrate (DMe)2R (50 μm; Merck Biosciences) for 30–60 min at 37°C. Cells were then washed with fresh medium to remove excess dye, and rhodamine fluorescence was examined by laser light excitation at 488 nm (Stroh and Schulze-Osthoff 1998; Hug et al. 1999).

Preparation of cell lysates and immunoprecipitation

Two millilitres of a suspension containing 106 cerebellar neurons/mL were transferred to each well of a 6-well tissue culture plate (Nunc) substrate-coated with with PLL followed by either L1 (2 μg/mL), L1Fc (10 μg/mL) or human Fc (10 μg/mL). Cells were treated with inhibitors 20 h after seeding. Two to four hours after inhibitor treatment, either 10 μm hydrogen peroxide or 500 nm staurosporine were added and the cells were cultured for an additional 20 min, and 2 h, 4 h or 24 h. Cells were then cooled on ice, washed twice with ice-cold HBSS and harvested in lysis buffer (20 mm Tris/HCl, pH 7.5, 150 mm NaCl, 1 mm EDTA, 1 mm EGTA, 1% Triton X-100, 2.5 mm sodium pyrophosphate, 1 mmβ-glycerol phosphate, 1 mm sodium vanadate, 1 μg/mL leupeptin, and 1 mm PMSF). Cell extracts were briefly sonicated on ice (15 s) and cleared by centrifugation at 20 000 g for 15 min at 4°C. The protein concentration in the supernatant was determined by the bicinchinonic acid method (BCA; Perbio, Bonn, Germany). Cell lysates were kept frozen at −20°C until use.

For immunoprecipitation of Bad, 200 μL of the cell lysate was taken and incubated with 1 μL of a polyclonal Bad-antibody (Merck Biosciences) overnight at 4°C. Then, 10 μL protein A-Sepharose beads were added and the mixture was incubated for 2 h at 4°C. The mixture was then centrifuged for 30 s at 100 g and 4°C. The pelleted beads were washed four times with 500 μL lysis buffer and re-suspended in SDS-sample buffer, boiled for 5 min and stored at −20°C until use.

Western blot analysis

Equal amounts of protein (10 or 25 μg) from each cell lysate sample were loaded and run on 8, 10 or 12% Tris/glycine polyacrylamide gels and transferred to nitrocellulose membranes (Protran, Schleicher & Schuell, Dassel, Germany). Membranes were blocked in 4% milk powder (Frema Reform, DE-VAU-GE, Lüneburg, Germany) in Tris-buffered saline/0.05% Tween 20 (TBS-T) for 2 h at room temperature (25°C) and incubated with the primary antibodies in blocking buffer overnight at 4°C. Primary antibodies used in this study were directed against phospho-Bad (Ser136), Bad, Bcl-xL, phospho-Akt (Thr308), Akt, phospho-ERK and ERK, phospho-PDK1 docking motif, caspase 9 (all from Cell Signaling Technology), GAPDH and Bcl-2 (Sigma). Concentrations of antibodies were used as suggested by the supplier. The appropriate HRP-conjugated secondary antibodies (Dianova) were applied after six washing steps with TBS-T. After 1 h incubation time at room temperature with secondary antibodies, membranes were washed again with TBS-T and blots were developed using enhanced chemiluminescence (ECL; Perbio, Bonn, Germany). In some cases, membranes were stripped and re-probed with different primary antibodies. Western blots were scanned and densitometric analysis was performed using TINA Image software (Version 2.0).

Statistical analysis

All values are expressed as means ± standard error of the mean. The standard error values indicate the variation between mean values obtained from at least three independent experiments. Statistical comparisons between two groups were made using the Kolmogorov-Smirnov two sample test. p-values < 0.05 were classified as statistically significant.

Results

Neuritogenesis

Neuritogenesis was determined 22 h after plating the cells. The average length of neurites was between 20 and 30 μm for both poly-l-lysine (PLL) alone and PLL with 10 μg/mL Fc. Neuritogenesis was significantly enhanced in a dose-dependent and saturable manner by both L1, mL1Fc and hL1Fc, but not by Fc alone (Fig. 1a). The optimal concentrations for stimulation were 10 μg/mL in the case of mL1Fc and hL1Fc and 2 μg/mL in the case of L1, with an average neurite length around 80 μm on all three substrates. On the laminin substrate, neurites showed an average length between 90 and 100 μm (Fig. 1b). The percentage of cells with neurites followed a similar pattern, showing 30% of all cells with neurites on PLL, 60–70% of all cells with neurites on either L1 or L1Fc and 80% of cells with neurites on laminin (Fig. 1c). No difference was seen between hL1Fc and mL1Fc.

Figure 1.

Effects of L1 on neuritogenesis of cerebellar neurons. Cerebellar neurons were cultured on PLL alone or PLL followed by Fc, L1, L1Fc or laminin and evaluated 22 h after plating. (a) Dose–response curve of neuritogenesis stimulated by substrate-coated L1. Cells were plated on PLL followed by Fc, L1 or mouse (m) or human (h) L1Fc. (b) Neuritogenesis on PLL or PLL followed by Fc (10 μg/mL), L1 (2 μg/mL), mL1Fc or hL1Fc (10 μg/mL), or laminin (1 μg/mL). Total length of neurites per cell is shown. (c) Number of cells with neurites measured as percentage of all cells counted. Substrates are indicated as in (b). Data represent mean ± SEM of three independent experiments. *p = 0.05, **p = 0.01, ***p = 0.001 significant difference from PLL.

Neuroprotection

Programmed cell death can be induced by a variety of factors, including ligand activation of death receptors, growth factor deprivation, oxidative stress, oncogenes, cancer drugs, and staurosporine. In our experiments, we investigated the ability of L1 to protect neurons from growth factor deprivation, oxidative stress and staurosporine treatment. For evaluation of cell death, cerebellar neurons were treated with propidium iodide and calcein. Calcein stains only live cells and propidium iodide necrotic and late apoptotic cells that have lost membrane integrity. Thus, apoptotic and necrotic mechanisms were both monitored.

First, a time course of neuroprotection on PLL versus L1 was determined (Fig. 2). Cells survived significantly better on the L1 substrates, even from the second culture day onward. After 7 days in culture, only about 25% of the attached cells survived on PLL, and about 65% on L1. The number of cells that attached to and survived on the Fc substrate was the same as on PLL alone. Representative pictures are shown for serum-starved cerebellar neurons grown for 7 days in culture on PLL substrate (Fig. 2b) and hL1Fc substrate (Fig. 2c). Next, we investigated, whether L1 was also able to protect neurons from oxidative stress or staurosporine-induced cell death (Fig. 3). For further cell death experiments, cell survival was determined 4 h after plating and the number of surviving cells was set to 100%. Then, for induction of cell death, hydrogen peroxide or staurosporine were added to the cells 16–20 h after plating and cells were maintained for additional 24 h before cell death was determined. As seen for serum deprivation, L1 enhanced cell survival after oxidative stress or staurosporine treatment by a factor of approximately two. When cells were cultured on control substrate in the presence of foetal calf serum for 8 days, 80–90% of all cerebellar neurons that had attached survived and thus were protected against oxidative stress and staurosporine induced apoptosis. Thus, L1 is nearly as potent as foetal calf serum in maintaining cerebellar neurons under the conditions chosen in this study.

Figure 2.

Time-course of cerebellar neuron survival on PLL and L1 under serum free conditions. Cerebellar neurons were plated on PLL followed by 10 μg/mL Fc or 10 μg/mL mL1Fc or hL1Fc and cultured under serum free conditions for 8 days. (a) Cell survival was determined every day by counting calcein and propidium iodide positive cells. Data represent mean ± SEM of three independent experiments. *p = 0.05, **p = 0.01 and ***p = 0.001 significant difference from PLL. (b, c) Calcein (green cells, arrows) and propidium iodide (red cells, arrowheads) staining of cerebellar neurons grown on PLL substrate (b) and hL1Fc substrate (c) for 7 days in culture.

Figure 3.

Survival of cerebellar neurons induced by L1 after serum deprivation or treatment with hydrogen peroxide or staurosporine. Cerebellar neurons were plated on PLL or PLL followed by Fc (10 μg/mL), mL1Fc (10 μg/mL), hL1Fc (10 μg/mL) or L1 (2 μg/mL). Control cells plated on PLL were cultured in the presence of 10% FCS. (a) Cell survival was determined after 7 days by counting calcein versus propidium iodide positive cells. (b, c) Cells were cultured under serum-free conditions for 16–20 h before induction of cell death by addition of 10 μm hydrogen peroxide (b) or 500 nm staurosporine (c) to the culture medium. Cell survival was determined 24 h after cell death induction by counting calcein versus propidium iodide positive cells. Data represent mean ± SEM of three independent experiments. ***p = 0.001 significant difference from PLL.

To investigate whether L1 has to be present on the surface of cerebellar neurons for L1-triggered neuroprotection, cells from wild-type and L1-deficient mice were grown on PLL or L1 and cultured under serum-free conditions for 7 days (Fig. 4). Only cells from wild-type mice showed enhanced neuroprotection when cultured on L1, but not cells from L1-deficient mice. Survival of L1-deficient neurons was normal, when cells were maintained on PLL in the presence of 10% serum, indicating that L1-deficient neurons are not compromised in their vitality.

Figure 4.

Cell survival of wild-type and L1-deficient cerebellar neurons on PLL or L1. Cerebellar neurons were plated on PLL or PLL followed by Fc (10 μg/mL), mL1Fc (10 μg/mL) or L1 (2 μg/mL). Control cells were cultured in the presence of 10% FCS. Survival was determined after 7 days by counting calcein versus propidium iodide positive cells. Data represent mean ± SEM of three independent experiments. ***p = 0.001 significant difference from PLL.

Thus, L1 has to be present at the cell surface for both L1-triggered neuritogenesis and neuroprotection.

Signal transduction pathways involved in neuritogenesis and neuroprotection triggered by L1

To characterize signal transduction pathways triggered by L1, we tested a set of specific inhibitors of signal transduction molecules for their effects on neuritogenesis and neuroprotection. Possible toxic effects of the used inhibitors on cerebellar neurons were excluded by testing them on cells grown on PLL alone and PLL plus Fc. None of the inhibitors significantly decreased neuritogenesis or neuroprotection under these conditions.

The fibroblast growth factor (FGF) receptor-specific inhibitors PD173074 and SU5402 abolished neuritogenesis mediated by L1 (Fig. 5a), but had no effect on L1-triggered neuroprotection (Figs 5b and c). The PI3-kinase inhibitors LY294002 and wortmannin inhibited both L1-triggered neuritogenesis and neuroprotection. Rp-cAMP, which inhibits cAMP-dependent activation of PKA, and the PKA inhibitor KT5720 reduced neuritogenesis induced by L1, but did not affect L1-triggered neuroprotection. Inhibitors of the Src family kinases, PP2 and herbimycin, strongly reduced L1-triggered neuritogenesis and neuronal survival to control levels (Fig. 5). The PLC inhibitor U73122 and the DAG kinase inhibitor R59022 repressed both L1-triggered neuritogenesis and neuroprotection, whereas the inactive analogue of U73122 (U73343) had no effect (data not shown). The non-selective protein kinase inhibitor K252a inhibited L1-mediated neurite outgrowth more strongly than neuronal survival, whereas the PKC inhibitor peptide 20–28 equally reduced both outgrowth and survival. Inhibition of MAP kinases with U0126 and PD98059 inhibited L1-triggered neuritogenesis and neuroprotection. The F-actin depolymerizing cytochalasin D also reduced both L1-triggered neurite outgrowth and neuronal survival. Treatment with the caspase inhibitors z-VAD-fmk and z-DEVD-fmk rescued cerebellar neurons from cell death when grown on PLL substrate, but had no further protecting effect on neurons grown on L1 substrate. As expected, the caspase inhibitors had also no effect on neuritogenesis (Fig. 5a). These results suggest an involvement of PI3-kinase, PLC, PKC, Akt, Bcl-2 family members and the MAP-kinase pathways in L1-triggered survival, partly overlapping with the pathways induced in L1-triggered neuritogenesis.

Figure 5.

Effect of inhibitors of signal transduction pathways on L1-induced neuritogenesis and neuroprotection. Cerebellar neurons were cultured under serum-free conditions on PLL alone or PLL followed by 10 μg/mL mL1Fc. (a) For neuritogenesis experiments, inhibitors were added 30–60 min after plating of the cells. Cultures were fixed 22 h after plating. Total length of neurites per cell is shown. (b) For cell death induced by serum deprivation, cells were treated with inhibitors 16–20 h, and 3, 5 and 7 days after plating. Survival of neurons was determined after 7 days by counting calcein versus propidium iodide positive cells. (c) For induction of cell death by hydrogen peroxide treatment 16–20 h after seeding, 10 μm hydrogen peroxide was added to the culture medium. Cell survival was determined 24 h after cell death induction by counting calcein versus propidium iodide positive cells. Data represent mean ± SEM of three independent experiments. *p = 0.05 and **p = 0.01 significant difference from untreated control.

To investigate if soluble L1 signals through the same pathways as substrate bound L1, soluble L1 was added to the cell culture medium of cerebellar neurons grown on PLL. A four times higher L1 concentration had to be added to obtain the stimulatory effects seen with substrate-bound L1 (data not shown).

Proteolytic cleavage of L1 by proteinases is important for L1-mediated cell adhesion, migration, neuritogenesis and NMDA receptor-dependent synaptic plasticity (Nayeem et al. 1999; Gutwein et al. 2000; Gutwein et al. 2003; Kalus et al. 2003; Matsumoto-Miyai et al. 2003). Whether proteolytic processing of L1 is also important for L1-mediated cell survival is, however, unknown. We first confirmed that L1-mediated neuritogenesis is inhibited by the metalloprotease inhibitor GM6001 (Kalus et al. 2003), but not by serine and cysteine protease inhibitors (Fig. 6a). We then investigated the effect of different protease inhibitors on L1-mediated neuroprotection (Fig. 6b and c). GM6001 significantly reduced neuronal survival on the L1 substrate, but not on the PLL substrate. The cysteine protease inhibitor E64, the serine protease inhibitor AEBSF, the serine and cysteine protease inhibitor leupeptin and the inhibitor of acidic proteases pepstatin A affected neither L1-mediated survival nor survival on the PLL control substrate. These results show that proteolytic processing by metalloproteases are involved in L1-mediated survival of cerebellar neurons.

Figure 6.

Effect of protease inhibitors on L1-induced neuritogenesis and neuroprotection. Cerebellar neurons were cultured on PLL alone or PLL followed by 10 μg/mL mL1Fc. (a) For neuritogenesis experiments, inhibitors were added 30–60 min after plating of the cells. Cultures were fixed 22 h after plating. Total length of neurites per cell is shown. (b) For cell death induced by serum deprivation, cells were treated with inhibitors 16–20 h, and 3, 5 and 7 days after plating. Survival of neurons was determined after 7 days by counting calcein versus propidium iodide positive cells. (c) For induction of cell death by hydrogen peroxide treatment 16–20 h after seeding, 10 μm hydrogen peroxide was added to the culture medium. Cell survival was determined 24 h after cell death induction by counting calcein versus propidium iodide positive cells. Data represent mean ± SEM of three independent experiments. ***p = 0.001 significant difference from untreated control.

To investigate whether L1 enhances cell survival by down-regulation of caspase protein levels or proteolytic cleavage/activation of caspases, cells were treated with a cell-permeable caspase substrate, the methyl ester derivative of D2R (DMe)2R (Fig. 7). When cells grown on L1 and PLL were compared, caspase activity was significantly reduced in neurons maintained on L1 (Fig. 7), indicating an involvement of signalling pathways that lead to inhibition of caspases in L1-triggered neuroprotection.

Figure 7.

Detection of caspase activity in cerebellar neurons. Cerebellar neurons were plated on PLL or PLL followed by Fc (10 μg/mL), mL1Fc (10 μg/mL) or L1 (2 μg/mL). Control cells plated on PLL were cultured in the presence of 10% FCS. Cells were cultured under serum-free conditions for 16–20 h before addition of 10 μm hydrogen peroxide to the culture medium. Caspase activity was determined 2 h after induction of cell death by adding 50 μm (DMe)2R to the cell culture medium. Cells were incubated for 30–60 min at 37°C and washed with fresh culture medium to determine rhodamine fluorescence. Data represent mean ± SEM of three experiments. **p = 0.01 and ***p = 0.001 significant difference from PLL.

Our inhibitor studies showed that the PI3-kinase activity is essential for L1-triggered neuroprotection. Possible downstream targets, which are strongly involved in survival mechanisms, are the protein kinases PDK1 and Akt. Akt promotes survival by phosphorylation of Bad, which then binds to 14-3-3 instead of Bcl-2 and Bcl-xL and inhibits the activation of the initiator caspase, caspase 9. To verify the involvement of this pathway and also the activation of the MAP kinase pathway, western blot analysis was performed to determine the phosphorylation status, protein amount and/or proteolysis (Fig. 8). Cells grown on L1 showed 1.5- to 3-fold higher levels of phosphorylated ERK, phosphorylated PDK1, phosphorylated Akt and phosphorylated Bad, when compared with cells grown on PLL or PLL plus Fc. After induction of cell death by hydrogen peroxide treatment, levels of phosphorylated PDK1, phosphorylated Akt and phosphorylated Bad in cells grown on L1 were reduced when compared with signals before induction of cell death. The phosphorylation levels of these proteins were significantly higher on L1 substrates than on PLL (Fig. 8). Addition of PI3 kinase inhibitors to cultured cells before induction of cell death abolished the L1-triggered activation of Akt and Bad, suggesting that activation of PI3 kinase is necessary for their activation. Addition of MEK inhibitors reduced the level of phosphorylated ERK to control levels. Addition of soluble L1 to cerebellar neurons grown on PLL also stimulated activation of the MAP kinase pathway and phosphorylation of Akt and Bad (data not shown). Also in this case, four times more soluble L1 was needed than substrate-bound L1 to achieve comparable effects.

Figure 8.

Western blot analysis of the phosphorylation of ERK, Akt, PDK1 and Bad proteins and protein levels of Bcl-2, Bcl-xL and caspase 9 after induction of cell death with hydrogen peroxide. Cerebellar neurons were grown on PLL alone or PLL plus 10 μg/mL mL1Fc and treated with inhibitors 20 h after seeding. Four hours after inhibitor treatment, 10 μm hydrogen peroxide was added, and the cells were cultured for additional 20 min (ERK, PDK1, Akt, Bcl-2 family) or 3 h (caspase 9), and then treated with lysis buffer. In each lane, 10 μg protein (for ERK and Akt analysis) or 25 μg protein (for Bad analysis) was loaded. Examples of typical western blots are shown in (a). (ai) ERK phosphorylation was analysed after culture in the presence or absence of 10 μm MEK inhibitor PD98059 using phosphorylated ERK and ERK protein antibodies. Lane 1, PLL; lane 2, PLL + H2O2; lane 3, PLL + PD98059 + H2O2; lane 4, mL1Fc; lane 5, mL1Fc + H2O2; lane 6, mL1Fc + PD98059 + H2O2. (aii) PDK1 phosphorylation was analysed after culture in the absence and presence of 50 μm PI3-K inhibitor LY294002 using antibodies recognizing phosphorylated PDK1 and GAPDH. Lane 1, PLL; lane 2, PLL + H2O2; lane 3, PLL + LY294002 + H2O2; lane 4, mL1Fc; lane 5, mL1Fc + H2O2; lane 6, mL1Fc + LY294002 + H2O2. GAPDH was used as loading control for pPDK1. (aiii) Akt phosphorylation was analysed after culture in the absence and presence of 50 μm PI3-K inhibitor LY294002 using antibodies recognizing phosphorylated Akt and Akt protein. Lane 1, PLL; lane 2, PLL + H2O2; lane 3, PLL + LY294002 + H2O2; lane 4, mL1Fc; lane 5, mL1Fc + H2O2; lane 6, mL1Fc + LY294002 + H2O2. (aiv) Bad phosphorylation was analysed using phosphorylated Bad and Bad protein antibodies. Lane 1, PLL; lane 2, PLL + H2O2; lane 3, mL1Fc; lane 4, mL1Fc + H2O2, lane 5, PLL + LY294002 + H2O2; lane 6, mL1Fc + LY294002 + H2O2. (av) Caspase 9 protein was analysed using a polyclonal antibody recognizing pro-caspase 9 and activated (cleaved) caspase 9. GAPDH was used as loading control. Lane 1, PLL; lane 2, PLL + H2O2; lane 3, mL1Fc; lane 4, mL1Fc + H2O2. (avi) Bcl-2 protein was analysed using Bcl-2 protein and GAPDH antibodies. GAPDH was used as loading control. Lane 1, PLL; lane 2, PLL + H2O2; lane 3, mL1Fc; lane 4, mL1Fc + H2O2. (avii) Bcl-xL protein was analysed using Bcl-xL protein and GAPDH antibodies. GAPDH was used as loading control. Lane 1, PLL; lane 2, PLL + H2O2; lane 3, mL1Fc; lane 4, mL1Fc + H2O2. (aviii) Immunoprecipitation from total cell lysates of cells grown on PLL and L1 and after induction of cell death with hydrogen peroxide with a Bad protein antibody. Bad, Bcl-2 and Bcl-xL protein antibodies were used for detection of proteins. Lanes 1, 3 and 5, cells grown on mL1Fc; lanes 2, 4 and 6, cells grown on PLL. Antibodies used for detection: lanes 1 and 2, Bcl-2; lanes 3 and 4, Bcl-xL; lanes 5 and 6, total Bad. (b–d) Quantification of phosphorylation (ERK, PDK1, Akt and Bad) and protein levels (caspase 9, Bcl-2 and Bcl-xL) were performed by densitometric analysis. The signals of non-phosphorylated total protein or GAPDH were used as control to normalize the signal intensity to amounts of protein loaded. Data represent mean ± SEM of three independent experiments. *p = 0.05 and **p = 0.01 significant difference from PLL.

Bcl-2 and Bcl-xL protein levels in cells grown on L1 were slightly increased when compared with levels of cells grown on PLL alone. Co-immunoprecipitation of Bad with Bcl-2 and Bcl-xL showed that lower amounts of Bcl-2 and Bcl-xL were pulled down together with Bad from cells cultured on L1 than from cells cultured on PLL (Fig. 8).

Discussion

L1 stimulates at similar concentrations both neuritogenesis and survival of cultured cerebellar neurons by a factor of two to three and in a saturatable manner. For both effects, substrate-coated L1 only affects neurons, when L1 is expressed at the cell surface of the stimulated cells: L1 does not enhance neuritogenesis or survival of neurons from L1-deficient mice. Soluble L1 also triggered neurite outgrowth and enhanced neuroprotection, but considerably higher concentrations needed to be applied to yield the same effects. It remains to be determined whether L1 triggers signal transduction by a homophilic trans interaction with L1 or whether L1 interacts with an unknown heterophilic partner at the neuronal cell surface that depends on L1 in cis interaction to induce downstream signalling.

It could be argued that enhanced neuritogenesis is a consequence of increased neuroprotection. However, several observations suggest that these two functions are not necessarily linked. Studies on the effects of neurotrophins on neuroprotection and promotion of neuritogenesis have shown that the two parameters can be dissociated. Using the neuroblastoma cell line SH-SY5Y as a model, Encinas and co-workers showed that BDNF promoted cell survival and neurite outgrowth, but only neurite outgrowth but not cell survival was blocked by the MAP kinase inhibitor PD98059 (Encinas et al. 1999). BDNF increased target innervation, but not survival of cutaneous sensory neurons (LeMaster et al. 1999). Sympathetic neurons depended on P13-kinase for TrkB-mediated survival and, to a lesser extent, on MAP kinase activity, while neurite outgrowth depended on both MAP kinase and PI3-kinase activities (Atwal et al. 2000). Furthermore, we have previously observed that L1 enhances neuroprotection, but does not influence differentiation of the dopamine synthesis pathway of mesencephalic neurons, whereas fibroblast growth factor does not enhance neuroprotection but promotes neuritogenesis (Hulley et al. 1998). Finally, the observations on signal transduction pathways studied in the present investigation point to disparities in L1-triggered signal transduction pathways involved in neuritogenesis and neuroprotection, although there is also overlap.

The integrity of the neuronal cytoskeleton is important for both neurite outgrowth and neuronal survival (Kamiguchi and Lemmon 1997; Panicker et al. 2003; Cabado et al. 2004). Also, L1-mediated neurite outgrowth requires a functional interplay between L1 and the cytoskeleton: L1 binds to the membrane-cytoskeletal linker ankyrin (Burden-Gulley et al. 1997; Gil et al. 2003; Nishimura et al. 2003) and to members of the ERM family of proteins (ezrin, radixin and moesin), which serve as linkers between L1 and actin (Dickson et al. 2002; Mintz et al. 2003). In this study, the F-actin depolymerizing cytochalasin D reduced L1-dependent neurite outgrowth and neuronal survival, indicating that cytoskeletal integrity and reorganization is important, not only for L1-mediated neuritogenesis but also for neuroprotection.

Different signal transduction molecules are involved in neuritogenesis and neuroprotection of L1

We have tested several inhibitors of signal transduction pathways for their ability to affect neuroprotection and neuritogenesis mediated by L1. The PI3 kinase inhibitors LY294002 (Vlahos et al. 1994) and wortmannin, the Src family kinase inhibitors PP2 (Gutwein et al. 2000) and herbimycin, the PLC inhibitor U73122 (Bleasdale et al. 1990; Zapf-Colby et al. 1999), the DAG kinase inhibitor R59022 (Camiña et al. 1999; Lopez-Andreo et al. 2003) and the MAP kinase inhibitors PD98059 (Pang et al. 1995) and U0126 inhibit both neuroprotection and neuritogenesis mediated by L1. For both L1-mediated neuroprotection and neuritogenesis, the Ras/MAP kinase and PI3 kinase pathways, are important mediators. For another member of the immunoglobulin superfamily, the neural cell adhesion molecule NCAM, which stimulates neurite outgrowth through the Ras/MAP and PI3 kinase pathways (reviewed by Crossin and Krushel 2000), survival of primary neurons and PC12 cells also depended on these pathways (Ditlevsen et al. 2003). Additionally, N-cadherin, laminin and basic fibroblast growth factor were shown to activate ERK in embryonic chick retinal neurons (Perron and Bixby 1999) and laminin-dependent stimulation of spiral ganglion neuron outgrowth was activated by the MAP/ERK pathway (Aletsee et al. 2002). These combined results show that ERK activation is one point of convergence for signalling pathways generated by a variety of axon growth inducers. It is interesting in this context that neurotrophins also use these pathways for neurite outgrowth and neuronal survival. The Ras/MAP kinase pathway is required for NGF-induced differentiation and neuritogenesis of PC12 cells (Cowley et al. 1994; Pang et al. 1995) and rat sympathetic neurons (Creedon et al. 1996), while the Ras/MAP and PI3 kinase pathways are necessary for BDNF-induced neurite outgrowth and survival of SH-SY5Y cells (Encinas et al. 1999). Src family kinases, especially Src, mediate L1-triggered signal transduction and neuritogenesis of cerebellar neurons (Ignelzi et al. 1994; Schmid et al. 2000). Our study confirms these data and shows that Src family kinases are also essential for L1-mediated neuroprotection.

A clear dissociation in signal transduction pathways is indicated by the observation that the FGF receptor inhibitors, PD173074 and SU5402, and the PKA inibitors, Rp-cAMP and KT5720, block neuritogenesis triggered by L1, but do not affect neuroprotection. PD173074 exhibits a high degree of specificity towards the FGF receptor in the nanomolar range (Mohammadi et al. 1998; Niethammer et al. 2002). SU5402 also inhibits the FGF receptor tyrosine kinase, but is less specific, because it also weakly affects the PDGF receptor (Mohammadi et al. 1997). The involvement of the FGF receptor in stimulation of neuritogenesis by cell adhesion molecules and the involvement of growth factors and their receptors in cell survival are well established: NCAM homophilic binding leads to activation of a signal transduction pathway involving Ca2+ through activation of the fibroblast growth factor receptor and to activation of the mitogen-activated protein kinase pathway (for review see Povlsen et al. 2003). Activation of the FGF tyrosine kinase receptor and the downstream effector PLCγ are also required for the responses stimulated by L1 (for review see Green et al. 1997). Indeed, direct stimulation of L1 can regulate PLC activity as shown by Schuch et al. (1989): treatment of PC12 rat pheochromocytoma cells with polyclonal antibodies directed against L1 resulted in reduced intracellular inositol-1,4-bisphosphate and inositol-1,4,5-trisphosphate, increased intracellular Ca2+ and regulation of PLC activity. Our results confirm these observations and extend them by showing that L1-triggered neuroprotection is independent of the FGF receptor activation.

In addition, cyclic adenosine monophosphate and protein kinase A trigger NCAM-mediated neuritogenesis (Povlsen et al. 2003). Here, we show that L1-triggered neuritogenesis also depends on PKA. In contrast, L1-triggered neuroprotection is independent of PKA. That neuroprotection of cerebellar neurons can be achieved without activation of PKA was recently demonstrated by Vaudry et al. (2003): Pituitary adenylate cyclase-activating polypeptide prevents C2 ceramide-induced apoptosis of cerebellar granule cells through activation of ERK in a PKA- and PKC-independent mechanism. Neuritogenesis and neuroprotection mediated by L1 were also abolished by treatment of the cells with the intracellular calcium chelating agent BAPTA-AM (1 and 10 μm) (data not shown). However, as survival on the PLL control substrate and in the presence of serum was equally inhibited, no conclusion about an L1-specific effect could be deduced.

Signal transduction of L1-triggered neuroprotection

To investigate whether L1 is not only protective for cerebellar neurons when growth factors are withdrawn (Chen et al. 1999), but also when cell death is actively induced, cells were treated with hydrogen peroxide and staurosporine. Hydrogen peroxide exposes cells to oxidative stress and staurosporine induces cell death as a broad spectrum kinase inhibitor (Kahns et al. 2002; Popescu et al. 2002). Our results show that L1 is able to protect neurons from cell death induced by oxidative stress or staurosporine. Analysis of the signal transduction pathways involved in this neuroprotection revealed that phosphorylation of PDK1, Akt, ERK1/ERK2 and Bad is increased in the presence of L1. These results suggest that the MAP kinase pathway contributes to both neuroprotection and neuritogenesis, correlating with findings on neurotrophin signalling (Encinas et al. 1999; Atwal et al. 2000; Kaplan and Miller 2000).

The PI3 kinase/Akt pathway is a potent mediator of cell survival (Dudek et al. 1997; Kennedy et al. 1997). Activated PI3 kinase is responsible for the phosphorylation of inositol phospholipids in generating phosphatidyl-inositol (3,4,5)-triphosphate (PIP-3). The increase in PIP-3 at the plasma membrane recruits and activates Akt, which in turn, phosphorylates and inactivates a variety of substrates, including the pro-apoptotic Bcl-2 family member Bad and caspase 9 (Del Peso et al. 1997; Cardone et al. 1998). That the PI3 kinase/Akt pathway is an effector of L1-triggered neuroprotection was shown by the activation of Akt upon L1 stimulation and by the inhibition of neuroprotection in the presence of the two PI3K inhibitors LY294002 and wortmannin. The L1-triggered increase in phosphorylation of Akt and Bad was inhibited by the PI3 kinase inhibitor LY294002. Similar results were obtained for BDNF-induced and NCAM-mediated cell survival (Encinas et al. 1999; Ditlevsen et al. 2003). The combined observations indicate that the PI3-kinase/Akt pathway is implicated in L1-triggered neuroprotection.

Members of the Bcl-2 family are crucial integrators of survival and death signals in higher eukaryotes. The Bcl-2 protein family members Bcl-2 and Bcl-xL suppress apoptosis, whereas Bad, Bax, Bid and BAK promote apoptosis. These molecules control the permeability of the outer mitochondrial membrane and, by doing so, the release of apoptogenic factors, such as cytochrome c and Smac/Diablo (for reviews see Scorrano and Korsmeyer 2003; Tsujimoto 2003). Cytochrome c interacts directly with the apoptotic protease activating factor-1 (Apaf-1) in the cytoplasm leading to dATP-dependent formation of the apoptosome (Li et al. 1997). This complex recruits and activates caspase 9. Activated caspase 9 can activate additional caspase 9 molecules, as well as the downstream caspases, such as caspases 3 and 7. Caspases are key promotors of apoptosis in most cells (reviewed in Slee et al. 1999), but cells can also die in an apoptotic manner independently of this enzyme activity (Borner and Monney 1999). We therefore investigated protein and phosphorylation levels of prominent members of the Bcl-2 family and examined caspase protein levels and caspase activity. Bcl-2 and Bcl-xL protein levels were slightly increased when cells were grown on L1 instead of PLL. Also, levels of phosphorylated Bad were enhanced. Bad phosphorylation by Akt has been shown to be necessary for the release of Bad from its association with Bcl-xL (Tan et al. 2000). That this is also the case in our study could be shown by immunoprecipitation experiments. Precipitation of Bad, together with Bcl-2 and Bcl-xL, showed that, in cells cultured on L1, the anti-apoptotic Bcl-2 family members were released from Bad after its phosphorylation. Determination of caspase activity in neurons grown on either L1 or control substrate revealed that total caspase activity in cells grown on L1 substrate was markedly reduced in comparison with cells grown on the PLL substrate. Western blot analysis of caspase 9 protein confirmed this by showing that more proteolytically cleaved, that is active, caspase 9 was detectable in cells grown on PLL than on L1. The combined observations suggest that enhanced Bcl-2 and Bcl-xL protein levels, their dissociation from Bad and the inhibition of caspases are important steps in L1-triggered neuroprotection.

Furthermore, proteolytic cleavage of L1 also appears to be necessary not only for L1-mediated neuritogenesis (Kalus et al. 2003) and NMDA receptor-dependent synaptic plasticity (Matsumoto-Miyai et al. 2003), but also for L1-mediated neuroprotection. Matrix metalloproteinases have been shown to regulate the turnover of extracellular matrix components and to play an important role in morphogenesis and tissue remodelling, as well as in tumor invasion and metastasis (Mitsiades et al. 2001). For instance, Herren et al. (1998) provided evidence that cleavage of beta-catenin and plakoglobin and shedding of vascular endothelial cadherin act in concert to disrupt structural and signalling properties of adherend junctions and to interrupt extracellular signals required for endothelial cell survival. Here, we show that treatment with a metalloproteinase inhibitor reduced survival of wild-type cerebellar granule neurons on L1 but not on PLL substrate, whereas inhibition of cysteine proteases or acidic proteases had no influence on L1-mediated survival. Furthermore, the metalloproteinase inhibitor had no effect on the survival of cerebellar neurons derived from L1-deficient mice grown on either L1 or PLL substrate. Because survival of cerebellar neurons derived from L1-deficient mice cannot be stimulated by L1, it is likely that L1-induced neurite outgrowth and neuroprotection are both mediated by homophilic interaction and signalling via neuronal cell surface expressed L1 and that proteolytic cleavage of L1 is necessary for the effects. The mechanisms by which proteolysis affects neuronal survival remain to be established.

The combined observations suggest that L1 uses similar signal transduction pathways, such as PI3 kinase, PLC, Src family kinase and MAP kinase pathways for both neuritogenesis and neuroprotection, but evokes distinct signal transduction pathways for neuronal survival being promoted by Akt, Bcl-2 and Bcl-xL. Proteolytic cleavage of L1 or its interaction partners by metalloproteinases is not only necessary for L1-mediated neurite outgrowth but also for neuron survival (for a hypothetical scheme, see Fig. 9). For L1-triggered neuritogenesis and neuroprotection, L1 needs to be expressed at the neuronal cell surface, thus activating multiple signalling pathways that may interact with other cues during neural development and are also likely to function under pathological conditions. It will be interesting to investigate whether similar signal transduction pathways are effective in L1-triggered neuronal survival and re-growth of axons after a lesion and synaptic plasticity in the adult.

Figure 9.

Proposed signal transduction pathways implicated in L1-triggered neuroprotection and neuritogenesis. Proteins involved in L1-mediated neuritogenesis and neuroprotection are shown in black, proteins only involved in neuritogenesis in red (encircled) and proteins only involved in neuroprotection in blue (boxed). Dashed lines represent putative cross-talk between signalling cascades. L1 interacts homophilically at the neuronal surface with L1 present on another surface. This interaction stimulates the FGF receptor signalling cascade and the MAP kinase pathway (for reviews see, Kenwrick et al. 2000; Panicker et al. 2003). L1 can interact directly with actin stress fibres, AP-2 adaptor complex, ankyrin and members of the ERM family, and these interactions are important for L1-mediated neurite outgrowth (Kamiguchi et al. 1998; Dickson et al. 2002; Gil et al. 2003). The cytoplasmic domain of L1 is subject to phosphorylation by p90rsk, ERK2 and casein kinase II (Sadoul et al. 1989; Wong et al. 1996a,b; Schaefer et al. 2002) and its phosphorylation influences cytoplasmic interactions, L1 mobility in the surface membrane and internalization. L1-mediated neuroprotection involves activation of PI3K and Akt, up-regulation of Bcl-2, inhibition of Bad and down-regulation of caspase 9 activity. Proteolytic cleavage of L1 is involved in L1-mediated neurite outgrowth (Kalus et al. 2003), neuronal survival and NMDA receptor-dependent synaptic plasticity (Matsumoto-Miyai et al. 2003).

Acknowledgements

The authors would like to thank Alexander Nikonenko for help in performing statistical analysis. This work was supported by a research grant from the Deutsche Forschungsgemeinschaft (SCHA 185/17–1).

Ancillary