We first analysed the subcellular localization of Foxo3a in motoneurons using an antibody that recognizes Foxo3a independently of its phosphorylation state. Motoneurons were purified from mouse embryonic spinal cord at E12.5, at the beginning of the period of naturally-occurring cell death, and cultured in the presence of a cocktail of neurotrophic factors (NTFs; see Methods) to strongly activate the PI3K/Akt pathway (Perez-Garcia et al., 2004; Dolcet et al., 2001; Dolcet et al., 1999). In these conditions, Foxo3a is predominantly detected in the cytoplasm (Fig. 1). To confirm that exogenous Foxo3a adopts the same subcellular localization, we electroporated purified motoneurons with a vector encoding HA-tagged wildtype Foxo3a (HA-wt-Foxo3a) together with a vector encoding GFP to identify transduced neurons. When motoneurons were grown in the presence of NTFs, only cytoplasmic staining for HA-wt-Foxo3a was detected (Fig. 2A).

We then tested the effects of inhibiting survival signalling through the PI3K/Akt pathway by treatment with the PI3K inhibitor LY294002 either in the presence (not shown) or the absence (Fig. 2A) of neurotrophic factors. In both cases, the majority of HA-wt-Foxo3a becomes localized in the nucleus of motoneurons (Fig. 2A and not shown). This pharmacological evidence suggests that Foxo3a localization reflects its phosphorylation by Akt. To confirm this, we studied the cellular distribution of the non-phosphorylatable form of Foxo3a, triple mutant (TM-) Foxo3a, in which all three Akt phosphorylation sites are mutated (Brunet et al ref). In the absence of survival signalling, HA-TM-Foxo3a showed a very similar nuclear distribution to HA-wt-Foxo3a (Fig. 2B). However, in contrast with the wildtype form, no redistribution of HA-TM-Foxo3a to the cytoplasm was observed in the presence of the cocktail of neurotrophic factors (Fig. 2B).

Results in all conditions were quantified by counting 2 independent experiments (Fig. 2C). They clearly show that, in motoneurons as in other cell types, Akt-induced phosphorylation of Foxo3a is required to prevent its accumulation in the nucleus.

To investigate the ability of Foxo3a to trigger death of cultured motoneurons, we took advantage of the electroporation technique we recently developed for high-efficacy transduction of primary neurons 16. We first confirmed that electroporation did not intrinsically inhibit motoneuron survival, by overexpressing a constitutively active form of Akt (Akt ca) in which the PH domain is replaced by a myristyl moiety which constitutively targets Akt to the membrane. Survival of electroporated motoneurons was reproducibly increased by overexpression of Akt ca as compared to the empty vector (Fig. 3A). Indeed, survival with Akt ca was greater than with NTFs alone. This was a result of increased Akt activity, since motoneurons electroporated with wild-type Akt showed survival values no greater than those with empty vector (not shown).

We therefore used electroporation to analyse the effects of Foxo3a on motoneuron survival. Overexpression of HA-wt-Foxo3a with GFP had no significant effect as compared to GFP alone or GFP with the empty vector (GFP: 100 ± 10; GFP + HA-wt-Foxo3a: 116 ± 12). To mimic Foxo3a activation and nuclear translocation, we then overexpressed HA-TM-Foxo3a together with GFP. The triple mutant triggered significant death of motoneurons cultured with or without NTFs (Fig. 3B). In the presence of NTFs, survival was reduced by 65%, which is similar to the proportion of motoneurons that die in the absence of trophic support (Fig. 3B).

One candidate gene downstream of Foxo3a was FasL, known to be regulated by Foxo3a and shown by us to be able to trigger motoneuron death. Motoneurons were therefore isolated from mice bearing mutations that lead to reduced signalling through the Fas pathway: gld mice (point mutation in FasL) and lpr mice (regulatory defect in Fas). As before, electroporation of HA-TM-Foxo3a in control motoneurons led to loss of 60 to 70 % of them compared to HA-wt-Foxo3a (Fig. 4). This figure was reduced from 70 to 50 % in gld motoneurons, and from 60 to 35% in lpr mutants. Thus, induction of Fas signalling is responsible for approximately half of the cell killing induced by Foxo3a.

In other cell types, Foxo3a can upregulate FasL either by direct interaction with the Foxo3A promoter 817 or indirectly through JNK activation 18, which itself leads to upregulation of FasL 7. Motoneurons were therefore electroporated with either wild-type or TM-Foxo3a, and treated immediately after plating with an inhibitor of JNK, L-JNKI1, which has been shown to inhibit the interaction between JNK and its substrates. In the presence of NTFs, JNKI1 had no effect on the survival of motoneurons electroporated with wt Foxo3a or the empty vector. In contrast, inhibition of JNK led to a reduction of 20% in the number of motoneurons triggered to die by TM-Foxo3a (Fig. 5). Since this percentage is lower than the fraction saved by reduced Fas signalling, it is likely that both direct and indirect (through JNK) mechanisms are used by Foxo3a to trigger Fas-dependent death of motoneurons.

Normal CD1 and C57BL/6 mice were obtained from Iffacredo (L'Arbresle, France). lpr/lpr and gld/gld mice were purchased from the Jackson Laboratory (Bar Harbor, ME). lpr/lpr mutants present an insertion of an early transposon into the fas gene, resulting in fas transcriptional repression 20. gld/gld mutants show a loss-of-function mutation in the fasL gene 26. All mutants were maintained on a C57BL/6 genetic background. Controls were done in either C57BL/6 or in CD1 mice (we previously showed that their response to Fas activation was identical).

LY 294002 was purchased from Calbiochem and used at a final concentration of 100 μM. L-JNKl1 was purchased from Alexis Biochemicals and used at a final concentration of 1 μM. Rabbit polyclonal antibody to HA-tag was from Clontech. Antibodies against Foxo3A were as described previously 8 and were a generous gift from A. Brunet (Stanford).

We are grateful to Anne Brunet for donating vectors used as the basis for our expression constructs. The vectors encoding HA-tagged Akt and Akt ca were as described previously 27. The vectors encoding HA-tagged wt and triple-mutant forms of Foxo3A were developed by Brunet et al8 The cDNAs were excised from the original clones and subcloned in the pCAGGS expression vector at ClaI site, since this vector gives higher expression levels in motoneurons.

Motoneuron cultures were prepared from E12.5 mouse spinal cords essentially as described 28, except that the magnetic column step was omitted and motoneurons in the enriched metrizamide fraction were identified by morphological criteria. Motoneurons were plated in the presence or not of a cocktail of neurotrophic factors (referred to as "NTFs": 1 ng/mL BDNF, 100 pg/mL GDNF, 10 ng/mL CNTF), added at the time of cell seeding. L-JNKl1 was added at the time of seeding, and LY294002 was added for 90 mins after 22 hours of culture.

Cells dissociated from E12.5 mouse ventral spinal cords were centrifuged over a 6.5% (v/v) Metrizamide cushion at 2000 rpm for 15 min. Cells at the interface were collected and washed on a BSA cushion at 1500 rpm for 5 min. Cells were resuspended in electroporation buffer at a density of 50,000 cells per 100 μL. 100 μL aliquots of the suspension were transferred to 4 mm gap cuvettes (Eppendorf) and five μg of the pCAGGS-GFP vector as well as the same molar amount of the vector of interest were added. After 15 min of incubation at room temperature, cells were electroporated using three pulses of 5 ms at 200 V with intervals of 1 s. Immediately after electroporation, 1.5 ml of culture medium was added to dilute the cells which were plated in four four-wells plates 16.

For anti-HA staining, motoneurons electroporated with HA-tagged wt or TM Foxo3A were seeded on polyornithine-laminin-coated 12-mm diameter glass coverslips and cultured for 24 hr at 37°C in complete Neurobasal medium supplemented or not with NTFs. Motoneurons cultured without NTFs were treated with LY294002 for the last 90 min. They were then fixed in 3.6% (v/v) formaldehyde for 30 min, washed in PBS-50 mM L-Lysine, and blocked for 1 hr in 5% donkey serum, 4% BSA, 0.1% Triton X-100 in PBS-50 mM L-Lysine. Cells were incubated with the anti-HA antibody (dilution 1:500) in blocking buffer, followed by fluorochrome-conjugated anti-rabbit secondary antibody. The cells were then observed under a fluorescence microscope. Cells with typical motoneuron morphology that were both GFP- and HA-positive were analyzed. The fraction of motoneurons expressing HA-wt-Foxo3A or HA-TM-Foxo3A exclusively in the nucleus, exclusively in the cytoplasm or in both compartments was evaluated in a total of 20 cells per well.

Staining for endogenous Foxo3A (dilution 1:100) was performed using the same method on non-electroporated motoneurons cultured for 24 hr with NTFs.

Motoneurons were counted 48 hr after electroporation under a fluorescence microscope. Only green cells with healthy motoneuron morphology, i.e. with large cell bodies and long non-fragmented neurites, were taken into consideration. Two different wells for each condition were counted (between 50 and 350 green motoneurons) in all the experiments. Cross percentages of test versus control (being set at 100%) were calculated yielding 4 values, used to calculate the medium and the SD for each experiment.