Downregulation of thymosin β4 in neural progenitor grafts promotes spinal cord regeneration

Thymosin β4 (Tβ4), the most abundant member of the thymosin-β (Tβ) protein family, is regarded as the principal G-actin-sequestering peptide in the cytoplasm, modulating the availability of actin monomers in a large variety of cells (Erickson-Viitanen et al., 1983; Huff et al., 2001; Safer and Nachmias, 1994; Sun and Yin, 2007). In the brain, Tβs are highly expressed (Devineni et al., 1999), mediate neuroprotection (Choi et al., 2007; Popoli et al., 2007) and are believed to regulate normal patterns of neurite outgrowth (Boquet et al., 2000; Carpintero et al., 1999; Choe et al., 2005; Lin and Morrison-Bogorad, 1990; Roth et al., 1999a; Roth et al., 1999b; van Kesteren et al., 2006), a process that is highly dependent on the reorganization and dynamics of the actin cytoskeleton (Bradke and Dotti, 1999; Bradke and Dotti, 2000; Chen et al., 2000; Gallo and Letourneau, 2004; Luo, 2000).

Neural stem cells (NSCs) have the capacity for self-renewal and multipotency (Gritti et al., 1996; Reynolds and Weiss, 1992; Suda et al., 1987; Svendsen et al., 1999; Weiss and Orkin, 1996). In vitro, in the presence of growth factors, NSCs proliferate and form free-floating colonies called neurospheres, which include a minority of neural stem cells together with various progenitor and more differentiated cells. Here we refer to neural stem cells and their progeny, progenitor cells, as NPCs. NPCs might be directed to a neuronal lineage and used to restore the function of damaged neurons in cases of neurodegenerative disorders and injuries of the brain or spinal cord (Cao et al., 2002; Pallini et al., 2005; Pluchino et al., 2005).

Because it can modulate the actin cytoskeleton, Tβ4 influences a variety of cellular processes. However, little is known about the involvement of Tβ4 in the proliferation, survival and differentiation of NPCs. In this study, we asked whether a downregulation of Tβ4 expression could result in NPCs with altered growth and differentiation abilities. To this end, NPCs were stably transduced with lentiviral vectors in order to overexpress either the Tβ4 antisense vector (Tβ4as) or the empty vector (EV) carrying only the enhanced green fluorescent protein (EGFP) as control. We then evaluated their in vitro behavior and in vivo survival and differentiation in a model of spinal cord injury.

Spinal cord injury results in gliosis and in limited cellular regeneration (Fitch et al., 1999; Silver and Miller, 2004) and has been an especially challenging target for stem cell therapies (Eftekharpour et al., 2008; Schultz, 2005). We found that grafts of Tβ4as-transduced NPCs (Tβ4as-NPCs) result in improved spinal cord repair, both functionally and morphologically, as compared with EV-transduced NPCs (EV-NPCs) or no cells. In addition, we found that grafted Tβ4as-NPCs contribute, in part, to functional restoration via mechanisms that involve the overexpression of cell adhesion molecule L1.

In conclusion, our data show that Tβ4 is functionally involved in neuronal differentiation in mouse embryonic NPCs and might have a therapeutic potential for the treatment of spinal cord injury.

In vitro NPCs grown in suspension in the presence of epidermal growth factor (EGF) and basic fibroblast growth factor (bFGF) form floating clusters of nestin-positive cells, termed neurospheres (Fig. 1A). We first confirmed the expression of Tβ4 in murine embryonic neurospheres by PCR and western blot analysis (Fig. 1B).

After removal of growth factor and plating on adhesive substrates, neurospheres adhere, their spherical structure is broken up, and cells differentiate and acquire the morphological properties of neurons and glia (Reynolds and Weiss, 1992). Using the differentiation culture conditions, for more than a week we followed the differentiation of NPCs into neurons by the expression of specific markers, such as β-tubulin III. The β-tubulin-III-positive cells had a characteristic neuronal phenotype, with a small soma and two or a few neurites, and lay above a layer of larger glial fibrillary acidic protein (GFAP)-positive cells (Fig. 1C, left panel).

In order to analyze the subcellular localization of Tβ4 in differentiating neurons derived from mouse NPCs, we performed immunofluorescence staining (illustrated in supplementary material Fig. S1). Confocal images of β-tubulin-III-positive developing neurons showed Tβ4 staining distributed in the cell body, growth cone and distal tips of neurites. In agreement with previous observations (Choe et al., 2005; Moccia et al., 2003; van Kesteren et al., 2006), the specific localization in both growth cones and processes indicates that Tβ4 is involved in neurite outgrowth of differentiating NPCs.

We next measured the mRNA level of Tβ4 in NPCs cultured in differentiating conditions, using real-time PCR (RT-PCR) and found that they have a significantly lower level of mRNA than neurospheres, indicating that a downregulation of Tβ4 occurs during differentiation (Fig. 1C, right panel).

After establishing that a reduction in Tβ4 levels occurs during differentiation of NPCs, we investigated whether a downregulation of Tβ4 expression could result in NPCs with altered growth and differentiation properties. To study the effects of the downregulation of Tβ4 expression, we therefore exploited an antisense strategy mediated by lentiviral infection. Because the lentiviral vector carries the EGFP reporter gene, we were able to sort the cell populations and obtain clones transduced with the EV and clones with the Tβ4as vector.

We used RT-PCR to verify the presence of the exogenous Tβ4as transcript, the expression of the EGFP transcript and the reduction in Tβ4 mRNA levels in transduced neurospheres (Fig. 2A). We then selected an antisense clone that showed the highest reduction in Tβ4 mRNA when compared with both the EV clones and the untreated neurospheres, although similar results were obtained with two other analyzed clones. Western blot analysis on the same clone confirmed a significant reduction in Tβ4 protein due to overexpression of the antisense construct (Fig. 2B).

Because the process of mitosis is highly dependent on actin dynamics (Glotzer, 2005), we first explored whether downregulation of Tβ4 could affect cellular division and, more specifically, the phase of cytokinesis during which a ring of actin has to be formed to allow physical cell division. We performed cell-cycle analysis using fluorescence-activated cell sorting (FACS) of the randomly cycling transduced neurospheres. We did not detect any accumulation of aneuploid cells in Tβ4as-transduced neurospheres, which showed a cell-cycle profile similar to the control cells, even after several passages in culture (supplementary material Fig. S2).

Next, we focused on the morphology of transduced neurospheres and analyzed the expression of progenitor markers. We found that Tβ4as-transduced neurospheres maintain a normal morphology and cellular composition (supplementary material Fig. S3).

The overexpression of Tβ4 in colon carcinoma cells can cause downregulation of E-cadherin (Wang et al., 2004), as a result of the disruption of the adherence junction due to the depolymerization of actin microfilaments triggered by this peptide (Huang et al., 2006). We therefore investigated whether the downregulation of Tβ4 expression in neurospheres could induce changes in N-cadherin expression, which mediates calcium-dependent adhesion in the central nervous system. Western blot analysis revealed a fivefold increase in the expression of N-cadherin in Tβ4as-transduced neurospheres (Fig. 3). In parallel, we also found an increased expression of β-catenin, which is normally complexed with N-cadherin. Densitometric analysis showed a 4.6-fold increase in β-catenin in Tβ4as extracts compared with controls (Fig. 3).

In conclusion, the antisense strategy was efficient at reducing Tβ4 expression levels in mouse embryonic NPCs without affecting their growth or morphology, and without affecting expression of markers generally expressed in undifferentiated neurospheres. However, Tβ4 downregulation in NPCs significantly elevated the expression of N-cadherin and β-catenin.

Using the expression of either β-tubulin III or microtubule-associated protein 2 (MAP2), the differentiation of neurons derived from transduced NPCs was examined for over one week. Triple-labeled confocal images showed fewer neurons from the EV-transduced cultures than from similar fields of Tβ4as-NPCs (Fig. 4A, arrows). This difference in neuronal numbers was confirmed by labeling with both neuronal markers. In particular, neurons derived from the EV-transduced clones accounted on average for 5% of the cell population in culture, whereas the cell population derived from the Tβ4as-transduced neurospheres showed a twofold increase in the number of neurons. Furthermore, western blot demonstrated a twofold increase in MAP2 in differentiating cultures transduced with Tβ4as instead of the EV (Fig. 4B).

In our relatively experimental conditions, no differences in astrocytic and oligodendrocytic differentiation were evident in cell-counting assessment using specific markers (GFAP and GalC) (data not shown).

Tβ4as-transduced neurons were morphologically different from neurons derived from the EV-transduced neurospheres. Starting from the early phases of differentiation, Tβ4as differentiating neurons generally extended processes more rapidly and had more prominent neurites (Fig. 4A; supplementary material Fig. S4A). Moreover, β-tubulin III immunoreactivity was generally greater within growing neurites of Tβ4as-transduced neurons than with controls. Interestingly, as shown in Fig. 4A, Tβ4as-transduced neurons established a connecting network and had more numerous neuronal processes emanating from the cell body (Fig. 4C; supplementary material Fig. S4b), thus acquiring a multipolar aspect. In addition, neurites of Tβ4as-transduced neurons were about twice as long as those of EV-transduced neurons (Fig. 4C) and this difference was observed up to day 7 (Fig. 4D). Together, these results indicate that Tβ4 plays a role in inhibiting neuronal differentiation and process outgrowth.

Previous studies have reported that N-cadherin is essential for differentiation and nerve cell morphology, and that its overexpression is sufficient to initiate neuronal differentiation in P19 and PC12 cells (Chen, Q. et al., 2005; Doherty et al., 2000; Gao et al., 2001; Utton et al., 2001). We therefore investigated whether Tβ4 downregulation could increase N-cadherin expression in differentiating cultures. Western blot analysis on protein extracts of cultures at day 3 of differentiation showed a 1.6-fold increase in N-cadherin in Tβ4as cultures compared with the EV cells (Fig. 5; right, upper panel). In addition, β-catenin levels were significantly increased (1.8-fold) (Fig. 5; right, middle panel).

Activation of extracellular-signal-regulated kinase (ERK) is required for full neurite outgrowth induced by N-cadherin (Bixby, 1990; Perron and Bixby, 1999; Saffell et al., 1997). Therefore, in the next step, we analyzed ERK phosphorylation levels in differentiating culture extracts. Quantification of the immunoreactive levels of the activated kinases, normalized to the total amount of the respective kinase, revealed a twofold increase in the phosporylation state of ERK1 in Tβ4as extracts (Fig. 5; right, lower panel). Thus, Tβ4 downregulation might influence neuronal differentiation of NPCs by increasing expression of the adhesion complex N-cadherin–β-catenin and by increasing ERK activation.

We then analyzed some electrophysiological parameters of EV- and Tβ4as-derived neurons to confirm neuronal identity. Furthermore, single-cell RT-PCR was performed on recorded cells for analysis of MAP2 expression (data not shown). By performing current-clamp experiments, a comparative analysis showed that the resting membrane potential for the two neuronal populations was not significantly different (P>0.05). In fact, at day 1 the mean resting potential of control neurons (n=6) was –27.66±1.15 mV, and for Tβ4as-transduced neurons was –29.12±4.48 mV (n=8). Similarly, the resting potential at day 7 was –39.47±1.71 mV for control neurons (n=19) and –36.11±3.05 for Tβ4as-transduced neurons (n=22). Tβ4as-transduced neurons and control neurons at day 4 (n=19 and 12, respectively) and at day 7 (n=19 and 12, respectively) showed one or two action potentials evoked by 50 pA of current injection (200 milliseconds). For all injected currents equal to or above +50 pA, all neurons presented adaptation (Contreras, 2004) (Fig. 6A).

In voltage-clamp mode, depolarizing steps from a holding potential of –60 mV (from –50 mV to +40 mV, in steps of 10 mV) evoked currents with an early inward peak (the voltage-dependent sodium currents) followed by outward components (the voltage-dependent potassium currents) (Fig. 6B). The early component (inset of Fig. 6B) was blocked by 1 μM of tetrodotoxin (not shown), indicating that these early currents were due to the voltage-activated sodium channels. The differences observed in the amplitude of both inward and outward currents were not statistically significant at either day 4 and 7. In fact, at day 7, the mean amplitude of the inward currents recorded at the command potential of –10 mV, from the holding potential of –60 mV, was 171±127 pA and 170±131 pA for Tβ4as-transduced and control neurons, respectively; at day 4 it was 127.2±65.6 pA and 101.2±52.3 pA, respectively. Similarly, the mean outward peak currents recorded at the command potential of +40 mV, from the holding potential of –60 mV, at day 7 was 1203.4±661 pA for Tβ4as-transduced neurons and 1358±565 pA for control neurons; at day 4 the mean was 1289±355 pA and 1308±784 pA, respectively.

Subsequently, we tested the cellular response to kainate injection. All patched neurons responded to perfusion of kainate (200 μM) with a current that was blocked by GYKI 52466 (100 μM), a selective AMPA receptor antagonist (Fig. 6C), confirming that the activation of these receptors was responsible for the current. An example of current evoked by 200 μM kainate at a holding membrane potential of –60 mV, in a control cell with the EV and in a Tβ4as-NPC at day 2 in vitro, is shown in Fig. 6D. In control neurons, the amplitude of the kainate-induced currents increased with the time in culture, and at day 7 the mean current was 110±15 pA (n=19). Tβ4as-transduced neurons responded to kainate with a current that was significantly higher than that of control neurons, starting from day 1 (Fig. 6E). However, on day 7 the amplitude of the currents for control and Tβ4as-transduced neurons was not significantly different (Fig. 6E). Such an increase in current density could be the result of the insertion of more channel receptors at the plasma membrane, or of changes in the intrinsic properties of the channels. Western blot analysis of differentiated cells at day 3 showed a twofold increase in subunits of glutamate receptors 2 and 3 (GluR2/3) in Tβ4as cultures (Fig. 6F), suggesting an increased expression of AMPA receptor. Based on the timing and higher number of glutamate-AMPA receptors, we concluded that Tβ4 downregulation induces an early expression of these receptors in NPCs.

Based on the in vitro results showing increased differentiation of Tβ4as-NPCs, we investigated the effects of their transplantation on locomotory recovery in a mouse model of spinal cord injury. In this model of spinal cord contusion injury, a complete palsy of the hindlimbs lasts for 2 weeks, followed by a stepping gait with partial recovery of locomotory activity over the subsequent 6-8 weeks (Pallini et al., 1988; Pallini et al., 1989).

NPCs, genetically modified to express the EV or Tβ4as, were transplanted as small neurosphere suspensions just rostrally to the injury site. Dissociation to generate a single-cell suspension was avoided to exclude detrimental effects on cell survival. At the time of transplantation, cells expressed nestin, and fewer than 1% expressed the astroglial marker GFAP (supplementary material Fig. S3).

To determine whether transplantation of NPCs improved recovery of function in mice bearing spinal cord contusion injury, we performed footprint analysis (Fig. 7A; upper panel) using the automated CatWalk system that monitors different gait parameters. Compared with observational open field methods, which monitor the animals' reluctance to move about in an open field arena and have been used in similar studies, the footprint analysis provides a more reliable assessment of hindlimb movements (Goldberger et al., 1990; Hamers et al., 2001; Hamers et al., 2006; Kunkel-Bagden et al., 1993; Pallini et al., 2005). Mice transplanted with either EV-NPCs (n=8) or Tβ4as-NPCs (n=8), recovered a stepping gait one week after spinal cord injury, whereas animals that received vehicle injection (n=8) experienced this recovery after 2 weeks (Fig. 7A; lower panel). Analysis of gait parameters showed that the stride length (right and left) and intensity both recovered significantly better in mice grafted with Tβ4as-NPCs than in mice grafted with EV-NPCs. Stride length (right and left) of mice grafted with Tβ4as-NPCs at 4 and 8 weeks after injury was 53.3±1.8 and 57.3±1.5 mm, respectively, compared with the stride length (right and left) of mice grafted with EV-NPCs, which was 48.9±3.4 and 53.2±1.7 mm (^*P<0.0001 one-way ANOVA test) (Fig. 7B; lower panel, right). The gait intensity of mice grafted with Tβ4as-NPCs at 4 and 8 weeks after injury was 31.9±5.8 and 41.9±6.3 respectively, whereas that of mice grafted with EV-NPCs was 20.3±6.9 and 29±1.8 (^*P<0.0005, ^**P<0.0001 one-way ANOVA test) (Fig. 7B; upper panel, center). There were no significant differences in the two groups of animals in the other parameters measured, including the base of support, maximum area, print width, print length, print area and angle of rotation (Fig. 7B).

Overall, this analysis revealed a better outcome in injured mice grafted with Tβ4as-NPCs.

To investigate the morphological basis and cellular mechanisms that might contribute to the motor recovery, we analyzed sagittal serial sections from mice grafted with Tβ4as-NPCs and EV-NPCs at 8 weeks after spinal cord contusion. The gross morphology of the injured spinal cord was visualized by hematoxylin-eosin staining, which identified the damaged area marked by the accumulation of a cellular and connective tissue scar (Fig. 8A). The quantification of the lesion area in consecutive sections showed that the lesion remains unchanged in the two groups of grafted and injured mice (data not shown). We then performed double and triple immunofluorescence labeling studies to identify the survival, localization and differentiation of transplanted NPCs in the two groups of injured mice.

Grafted NPCs were detected by direct examination of EGFP fluorescence in the injured spinal cords. Examination of sagittal sections 8 weeks after injury in both groups of mice identified nuclei positive for EGFP surrounding and within the lesion site, suggesting that, following transplantation, many of the cells were attracted towards the lesion site. Fig. 8B,C show triple-stained examples of the distribution of EGFP-positive cells in sagittal sections from animals receiving transplants of the EV- or Tβ4as-NPCs. In the images, survival and localization of the EGFP-positive cells were visualized along with the nuclei and GFAP-positive cells which, as previously reported (Camand et al., 2004), form a glial scar consisting mainly of reactive astrocytes surrounding the center of the lesion. The survival in the injured area of Tβ4as-NPCs was notably higher than that of EV-NPCs (Fig. 8B,C), which in two mice were barely detectable and appeared unhealthy as judged by the appearance of their fluorescence. To confirm the presence of transplanted cells and to rule out autofluorescent cell debris, we also immunostained a selection of transplanted spinal cord sections with an anti-GFP antibody. Fig. 8D shows that the EGFP-positive cells were colabeled with the anti-GFP antibody and confirms that they were more abundant in mice that received transplants of Tβ4as-NPCs.

By performing immunofluorescence with an antibody specific for Tβ4, we also confirmed that the Tβ4as-NPCs maintain a downregulation of the peptide in vivo. As can be seen in Fig. 8E, transplanted Tβ4as-NPCs displayed only the green fluorescence, whereas transplanted EV-NPCs showed colabeling of the EGFP with Tβ4 staining.

To investigate the phenotype of the cells derived from the grafted NPCs, spinal cord sections were immunostained with anti-GFAP antibody for astrocytes, NG2 for oligodendrocyte precursors (Yang et al., 2006), and anti-β-tubulin III or MAP2 or Smi32 antibody to stage neuronal differentiation. It is interesting that, whereas NPCs expressed nestin at the time of transplantation (supplementary material Fig. S3), nestin immunofluorescence was no longer detectable 8 weeks after the grafts (data not shown), indicating that NPCs entered a differentiation program in vivo. By contrast, confirming earlier reports (Frisen et al., 1995; Namiki and Tator, 1999; Shibuya et al., 2002; Sieber-Blum et al., 2006), intense nestin fluorescence was present in the host tissue near the lesion site (data not shown).

In injured mice transplanted with the EV-NPCs, EGFP-positive cells showed a rounded shape, never displayed a neuronal-like morphology and did not express neuronal antigens (Fig. 9A,B). The cellular morphology of the Tβ4as-NPCs engrafted into injured spinal cord showed that a large proportion of these cells remained undifferentiated, maintaining a rounded shape and forming clusters (Fig. 8C,D). Interestingly, some Tβ4as-NPCs exhibited multipolar extended processes resembling neuronal cells and were generally found in the area surrounding the lesion (Fig. 9A). To verify their neuronal differentiation, we then performed immunofluorescence with specific markers. This analysis showed that whereas grafted EV-NPCs remained rounded and undifferentiated, a small number of the Tβ4as-NPCs (0.1-0.2%) displayed a neuronal-like morphology and expressed either β-tubulin III, MAP2 or Smi 32 (Fig. 9B). By contrast, although NG2 expression was upregulated in the lesioned area (Jones et al., 2002), we failed to identify grafted NPCs by the expression of the NG2 proteoglycan (data not shown).

These data suggest that a low percentage of Tβ4as-NPCs are capable of terminal differentiation along a neuronal lineage in an inhospitable environment such as the injured spinal cord.

The small number of Tβ4as transplanted cells differentiated into neurons was insufficient to account for the functional recovery observed in grafted and injured mice. We therefore explored the hypothesis that engrafted Tβ4as-NPCs might provide an environment conducive to the attachment and growth of endogenous neural and neuronal cells.

Tβ4 plays a pivotal role in regulation of actin dynamics in neurons and is involved in cell survival and neurite elongation by influencing cytoskeleton changes and the redistribution of cell adhesion molecules (Yang et al., 2008). In particular, it has been proposed that Tβ4 exerts its neuropromoting effects, at least partly, via mechanisms that involve activation of the cell adhesion molecule L1. The neuronal recognition molecule L1 has been shown to favor axonal growth in an inhibitory environment (Castellani et al., 2002; Chen, J. et al., 2005; Chen et al., 2007; Roonprapunt et al., 2003; Xu et al., 2004; Zhang et al., 2005). On the basis of these observations, we investigated whether the functional improvement of mice transplanted with Tβ4as-NPCs might be due to increased L1 expression. In Fig. 9C, a series of confocal images shows that Tβ4as-grafted NPCs overexpress L1 and create an environment that is conducive to host neurite regrowth. Numerous anti-β-tubulin-III-positive processes were found to extend into areas rich in transplanted Tβ4as-NPCs. On close inspection, it appeared that host neurites were attracted to and made contact with Tβ4as-grafted NPCs in the lesioned area (Fig. 10A; arrows), most probably due to the ability of these cells to express L1.

Furthermore, as shown by double immunofluorescence (Fig. 10B), β-tubulin-III-positive fibers were also immunostained for GAP43 (neuromodulin), which specifically marks enhanced axon growth status and axonal regeneration (Schreyer and Skene, 1993). This finding shows that the microenvironment after transplantation of the Tβ4as-NPCs is more hospitable to axon regeneration.

Given the fact that serotonin (5-HT) fibers play important roles in locomotion (Barbeau and Rossignol, 1991) and in recovery after injury (Ribotta et al., 2000), we examined whether Tβ4as-NPCs overexpressing L1 promote sprouting of 5-HT axons. By immunofluorescence, we found a high density of 5-HT fibers localized around Tβ4as-NPCs in the proximity of the lesion site (Fig. 10C). Because 5-HT fibers provide a diffuse innervation and positively correlate with a degree of functional recovery (Pearse et al., 2004; Ribotta et al., 2000), we suggest that a more vigorous regeneration and sprouting of host fibers might explain the enhanced improvement in gait.

In this study, we examined for the first time the effects of downregulation of Tβ4 levels in mouse embryonic NPCs and the potential of grafted Tβ4as-NPCs to promote repair in a mouse model of spinal cord contusion injury.

Downregulation of Tβ4 levels in NPCs cultured in differentiating conditions resulted in twice the number of neurons, which showed a higher number of prominent neurites and a significantly enhanced neurite outgrowth. Importantly, when grafted onto the injured spinal cord, Tβ4as-NPCs promoted functional recovery. This which might be due to (i) enhanced survival of transplanted cells; (ii) differentiation towards the neuronal phenotype; and/or (iii) enhanced regeneration and sprouting of host fibers close to the lesion site.

Neurite elongation is responsible for neuronal patterning and for the formation of connections crucial to the development of the nervous system and for nerve regrowth following injury. The establishment of neuronal networks is regulated by actin cytoskeletal dynamics. Tβ4 is enriched in developing neurite processes where actin is required (Choe et al., 2005). However, conflicting results have been obtained by the overexpression or downregulation of Tβs on neurite-promoting activity. In zebrafish, in vitro overexpression of Tβ in regenerating retinal ganglion cells results in alterations to neurite shape and excessive branching (Roth et al., 1999b). In cultured cortical and hippocampal neurons, overexpression of Tβ15 enhances neurite branch formation through its G-actin-sequestering activity (Choe et al., 2005), whereas in cultured Lymnaea pedal neurons, Tβ downregulation leads to a significant increase in neurite outgrowth (van Kesteren et al., 2006). Controversy also surrounds the effect on the actin cytoskeleton generated by modulation of the Tβ4 concentration. Tβ4 is generally believed to sequester monomeric G-actin, thus facilitating depolymerization of the actin filaments (Sanger et al., 1995). However, exceptions to this action have been reported, depending on the cell type and on the levels of expression of Tβ4 and of other actin-binding partners within the cell (Golla et al., 1997; Fan et al., 2009). Further studies are required to clarify the type of reorganization of the actin cytoskeletal (depolymerization or polymerization) that occurs during downregulation of Tβ4 in NPCs. However, it is conceivable that in embryonic NPCs the effects of Tβ4as are mediated by increasing filamentous actin turnover in growing neurites, thereby enhancing microtubule-based extension. In this respect, we observed that β-tubulin III immunoreactivity was generally greater within growing neurites of Tβ4as-transduced neurons than in EV neurons.

Alternatively, Tβ4 might directly influence the various aspects of neuronal differentiation of NPCs by antagonizing the commitment of progenitors to the neuronal lineage. Indeed, Tβ4 has been detected in the nucleus of cells, where it might alter the expression of different genes directly involved in the determination of neuronal fate (Huff et al., 2004; Moon et al., 2006). In this respect, our study also shows that a reduction in Tβ4 facilitates neuronal differentiation of NPCs without increasing the proliferation of neural progenitors, probably by enhancing exit from the cell cycle and having an instructive differentiating effect.

Changes in actin dynamics induced by downregulation of Tβ4 levels could result in a different anchoring and distribution of cell adhesion molecules that are needed during neurite growth (Theriot, 1994; Yu and Malenka, 2003). L1 and N-cadherin are cell adhesion molecules that promote axonal growth during development (Walsh and Doherty, 1997). N-cadherin, complexed with catenins (Yap et al., 1997), provides traction forces and triggers intracellular signaling cascades that are required for neurite extension (Kiryushko et al., 2004). Indeed, we found that the downregulation of Tβ4 in NPCs induces an increase in the expression of both N-cadherin and β-catenin. Because increases in this adhesion complex have been shown to generate neurons with higher neurite output, the observed different morphologies of neurons derived from Tβ4as-NPCs could be due to the upregulation of the adhesion complex (Chen, Q. et al., 2005; Otero et al., 2004). It is known that N-cadherin can activate ERKs and induce neurite outgrowth (Perron and Bixby, 1999). In addition, pharmacological inhibition of ERK activation strongly inhibits the ability of this adhesion protein to promote neurite growth (Pang et al., 1995). Moreover, ERKs play a significant role in neuronal differentiation, initiation of neurite outgrowth and rearrangement of neurites (Sweatt, 2001). Taken together, these reports enable us to hypothesize that the phenotype found in Tβ4as-transduced neurons is due to the increase in N-cadherin, which in turn supports the activation of ERKs. However, we cannot exclude the possibility that Tβ4as-induced remodeling of the actin cytoskeleton first activated ERKs, and that the sustained activation of ERKs influenced neurite outgrowth, promoted neuronal fate determination, and increased expression of the N-cadherin–β-catenin adhesion complex.

Finally, we found that downregulation of Tβ4 increases AMPA receptor expression in neurons derived from Tβ4as-NPCs. Interestingly, it has recently been demonstrated that N-cadherin is associated with AMPA receptors and increases the level of AMPA surface expression in neurons (Nuriya and Huganir, 2006). This result confirms that actin-binding proteins are able to affect the distribution of ion channels and regulate receptor trafficking in neurons. (Redell et al., 2007; Yamoah et al., 2005). Moreover, the role of the actin network in regulating the localization of ion channels is important in establishing the electrical properties of neurons (Hattan et al., 2002; Petrecca et al., 2000).

Although further studies will be required, it is noteworthy that our study has not only indicated the involvement of Tβ4 in neurite outgrowth and in neuronal differentiation, but has for the first time identified possible molecular candidates through which Tβ4 might exert its function.

After transplantation into an inhospitable environment, such as the injured spinal cord, Tβ4as-NPCs survived and retained their differentiation capability, promoting the recovery of locomotion in injured mice. Because it seems unlikely that the functional recovery of injured mice was attributable to the low number of Tβ4as-NPCs that differentiated into neurons, we investigated alternative mechanisms to explain this, including changes in the properties of the extracellular matrix of grafted Tβ4as-NPCs.

The adhesion molecule L1 favors axonal growth in an inhibitory environment (Castellani et al., 2002; Chen, J. et al., 2005; Dong et al., 2002; Fransen et al., 1998; Lemmon et al., 1989; Roonprapunt et al., 2003; Xu et al., 2004; Zhang et al., 2005), promotes neurite outgrowth, and displays survival-promoting effects on cultured central nervous system neurons (Chen et al., 1999; Dong et al., 2002; Dong et al., 2003; Lemmon et al., 1989; Lindner et al., 1983; Rathjen and Rutishauser, 1984). We found that Tβ4as-NPCs overexpress L1 and survive better in the lesioned area, suggesting that L1 positively influences NPC survival, even after grafting into a hostile environment.

Embryonic stem cells overexpressing L1 support the regrowth of corticospinal tract axons and survive better than non-transfected stem cells in the injured spinal cord of adult mice (Chen, J. et al., 2005). Similarly, functional recovery and positive effects on damaged 5-HT and corticospinal axons of adult injured mice were reported after injection of an adenovirus expressing human L1 protein (Chen et al., 2007). We also observed an increased number of β-tubulin-III-positive fibers traveling close to the grafted Tβ4as-NPCs in the lesioned area. The observation that β-tubulin-III-positive fibers were found positive for GAP43, a universal indicator of axonal growth status, underlines the suggestion that enhanced axonal growth and regeneration occurred after Tβ4as-NPC graft. The robust serotonergic sprouting also demonstrates the potential of grafted Tβ4as-NPCs in promoting regeneration of spared host fibers. Thus, it is possible that Tβ4as-NPCs facilitate axonal regeneration by providing a growth-permitting guiding substrate through stimulation of the production of L1. Interestingly, a direct link between L1 and Tβ4 has recently been shown in which Tβ4 enhances L1 expression in a dose-dependent manner, and L1 mediates Tβ4-induced neurite outgrowth and survival in neurons in vitro (Yang et al., 2008).

Although our understanding of the exact mechanisms underlying the overexpression of L1 by Tβ4 will require further investigation, the present study indicates that Tβ4 might exert promoting effects on neuronal cells by generally altering the expression of adhesion molecules. One possible mechanism for regulation of the distribution of adhesive molecules rapidly and locally would be for Tβ4 to control the processes of endocytosis and exocytosis, which are in fact highly dependent on the actin cytoskeleton (Kamiguchi et al., 1998; Lee et al., 2008; Roth et al., 1999b).

In conclusion, taken together our data show that Tβ4 has a pivotal role in controlling different aspects of neuronal differentiation, including distribution of membrane channels, and thus in determining the intrinsic properties of neurons in vitro. The ability of Tβ4as-NPCs to foster a more permissive and/or hospitable environment for fiber regeneration and tissue repair might have important implications for therapeutic intervention to improve outcome after spinal cord injury.

Procedures involving animals and their care were conducted in strict accordance with the Policy on Ethics approved by the Society for Neuroscience, and with the European Communities Council Directive for Experimental Procedures. Every effort was made to minimize the number of animals used and their suffering. Telencephalic regions from embryonic day E14 wild-type CD1 mice were dissected and cultured as previously described (Ricci-Vitiani et al., 2004; Pallini et al., 2005).

Immunofluorescence was performed as previously described (Mollinari et al., 2005) using the following primary antibodies: rabbit anti-GFAP (1:1000; Chemicon), mouse anti-β-tubulin-III (1:500; Chemicon), mouse anti-MAP2 (1:500; Sigma-Aldrich), mouse anti-GalC (1:250; Chemicon). Coverslips were counterstained with Hoechst 33342 (Sigma-Aldrich), mounted with an antifading glycerol medium and observed with a confocal microscope (Olympus FluoView FV1000). In every experiment, at least 500 cells were counted in ten different fields to calculate the percentage of neurons. For the length of neurites, 20 major neurites per day from three independent experiments were measured from the cell body to the tip of the longest process using ImageJ (NIH, Bethesda, MD). Major neurites were defined as those with the longest length.

HA-Tβ4 cDNA was subcloned in the antisense orientation under the CMV promoter of a lentiviral vector that carried the EGFP reporter gene under the PGK promoter. Recombinant lentiviruses were derived by the combined transfection of different plasmids as described (Ricci-Vitiani et al., 2004). After infection, cells were sorted for their fluorescence (FACS Vantage, Becton and Dickinson) until a virtually pure population of transduced cells expressing EGFP alone (EV) or the Tβ4as was obtained. The protocol of lentiviral infection was then repeated three times in order to obtain different transduced-NPC clones.

Relative quantitative and single cell RT-PCR were performed as previously described (Merlo et al., 2007; Carunchio et al., 2008). The primer sequence information can be found in supplementary material Table S1.

Neurospheres were mechanically dissociated and then incubated with diluted trypsin to obtain a cell suspension that was fixed in 2% paraformaldeyde, washed in PBS and suspended in a citrate solution containing propidium iodide (Sigma-Aldrich) and RNAse (Sigma-Aldrich) (Andreassen et al., 2001). Cell-cycle analysis was performed by FACS (FACS Calibur, Beckton and Dickinson) counting 30,000 events per experiment.

Western blot analysis was generally performed as previously described (Merlo et al., 2007). For Tβ4 detection, the acrylamide gel was washed several times in PBS and incubated in 10% glutaraldehyde (Sigma-Aldrich) for 40 minutes. Details on antibodies can be supplied on request.

Electrical activity of the cell soma was recorded in the whole-cell configuration of the patch-clamp method (Hamill et al., 1991) in neural stem cells from 1-7 days after plating. Recordings were performed at room temperature (22-24°C) using borosilicate glass patch pipettes pulled with PP-83 (Narishige, Japan). The electrodes with resistance of 4-5 MΩ were connected to an Axopatch 200B amplifier (Axon Instruments, Union City, CA). pCLAMP 8 software was utilized for the data acquisition system (Axon Instruments). The signals were filtered at 2-10 KHz and digitized at 20-50 KHz. After the establishment of a gigaseal, the pipette resistance and capacitance were compensated electronically and the cells were accepted for study only if these parameters remained stable (Zona et al., 2006). P/4 subtraction was used to eliminate capacitive transients and leak currents whenever possible. Control and agonist- or antagonist-containing solutions were applied with a gravity-driven system (SF-77B Perfusion Fast Step Warner Instruments, Hamden, CT). Details of solution composition can be supplied on request.

Adult female Swiss (CD1) mice weighing 27-30 g were used (Catholic University Breeding Laboratory, Rome, Italy). Anesthetized animals were subjected to a T7-T8 laminectomy. A modified aneurysmal clip (80 g/mm²) was then applied for 1 second over the dura mater. Immediately after injury, mice received homotransplants (10⁵ cells in 4 μl) of either EV-NPCs (n=8), Tβ4as-NPCs (n=8) or vehicle medium (n=8) via a glass pipette with a sharp beveled tip 100 μm in diameter that was connected to a Hamilton microdrive syringe. The NPCs or vehicle medium were slowly injected 1-2 mm rostrally into the lesion at 0.2-0.3 μl steps over 10 minutes to prevent loss of fluid along the needle tract.

Gait abnormalities in mice with contusion injury of the spinal cord were assessed weekly by footprint analysis using the CatWalk system (Noldus, Wageningen, The Netherlands) (Hamers et al., 2001; Hamers et al., 2006). Briefly, the animals traverse a walkway in a dark room with a glass floor through which light is beamed from the long edge. Light is reflected completely internally. Only when the paw touches the floor light is the light deflected and exits the glass, so that only the contact area is visible. The intensity of the signal depends on the pressure exerted, so that the spot will appear brighter when more weight is put on the paw. Animals crossing the walkway are videotaped using a computer-assisted setup, and digitized data are thresholded in order to extract the paw-floor contact areas and remove background. At least three runs per animal were performed in each session. Labels were then assigned to the prints (left and right, fore and hind paws) and several parameters were measured by the Catwalk software, including: (i) base of support (distance between the central pads of the hindfeet); (ii) intensity (mean brightness, in arbitrary units, of all pixels of the print at max contact; this is a measure of weight support of the different paws); (iii) max area (maximum area of a paw, in pixels, that comes into contact with the glass plate); (vi) print area (surface area, in pixels, of the complete print); (v) print length (length of the complete print); (vi) print width (width of the complete paw print); (vii) stride length (distance between two consecutive prints on each side); and (viii) angle of rotation (angle formed by the intersection of lines from the left and right prints) (results not shown). Each mouse ran across the CatWalk several times before surgery to establish baseline locomotor parameters.

Cryostat sagittal sections (40 μm) were obtained from 8 week grafted animals, and confocal immunofluorescence was performed as previously described (Merlo et al., 2007). Details on antibodies can be supplied on request. For quantitation of cell survival, the total number of EGFP-expressing and DAPI-positive cells was counted. To quantify the differentiation pattern of transplanted cells, we used confocal microscopy to count the number of EGFP-positive cells that were double-labeled with a different neuronal marker. We then counted the number of EGFP-positive cells that were double-labeled with the neuronal marker in three random fields per section. On average, 100-200 cells were counted per field.

Measurements are expressed as mean ± s.e.m. or mean ± s.d. for behavioral analyses. Statistical analysis was carried out using the Student's t-test or the one-way ANOVA test (MedCalc software, version 9.5.0) for behavioral experiments. Data were considered statistically significant if P<0.05.