Functional and molecular clues reveal precursor-like cells and immature neurones in the turtle spinal cord

Abstract

In lower vertebrates, some cells contacting the central canal (CC) retain the ability to proliferate, leading the reconstruction of the spinal cord after injury. A better understanding about the nature of these cells could contribute to the development of novel strategies for spinal cord repair. Here, by combining light and electron microscopy, immunocytochemistry and patch-clamp recordings, we provide evidence supporting the presence of precursor-like cells and immature neurones contacting the CC of juvenile turtles. A class of cells expressed the ependymal and glial cell marker S100 and displayed morphological and electrophysiological features of radial glia: relatively low input resistance, high resting potential, lack of active membrane properties and extensive dye-coupling. A second class of S100 reactive cells were characterized by a higher input resistance and outward rectification. Finally, some CC-contacting cells expressed HuC/D – a marker of immature neurones – and fired action potentials. The coexistence of cells with functional properties of precursor-like cells and immature neurones suggests that the region surrounding the CC is a site of active neurogenesis. It remains to be demonstrated by lineage analysis whether, as in the embryonic cerebral cortex, radial glia are the progenitor cells in the turtle spinal cord.

The region surrounding the central canal (CC), termed central gelatinosa (CG) (Lenhossék, 1895), is a dynamic area displaying a remarkable degree of plasticity. Ependymal cells in the adult spinal cord have an evolutionary conserved origin in the ventral neuroepithelium and seem to retain the proliferative capacity of neural stem cells (Fu et al. 2003). Indeed, in some anamniotes, the ependymal and peri-ependymal cells proliferate and generate neurones and glia. In the eel, the cells lining the CC orchestrate the reconnection of the spinal cord after complete transection (Dervan & Roberts, 2003). In mammals, after spinal cord contusion there is proliferation of CG cells which mediates limited endogenous repair (Beattie et al. 1997).

The CG of young turtles is a site of cell proliferation where some cells express neurotransmitters and receive synaptic contacts suggesting preservation of neurogenesis after birth (Fernández et al. 2002). However, detailed functional and molecular evidence supporting the neuronal nature of these cells and the presence of their putative precursors is still lacking. Here, we address this issue by combining morphological, immunocytochemical and electrophysiological techniques. We found two functionally distinct types of cells that synthesize proteins typical of ependymal and glial cells. In addition, another group of CC-contacting cells exhibited molecular and functional characteristics of immature neurones. These various classes of cells may represent, as in the embryo, dynamic differentiation stages in a lineage-related cell population (Fishell & Kriegstein, 2003).

Methods

Juvenile turtles (Chrysemys d'orbigny, 5–6 cm carapace length) were used following the norms established by our local Committee for Animal Research (UDELAR). The spinal cords of anaesthetized animals (5 mg kg⁻¹ sodium methohexitone, Brietal, Lilly, i.p.) were fixed by perfusion. For immunohistochemistry (n = 30), we used 10% paraformaldehyde in 0.1 m phosphate buffer (PB, pH 7.4), and for transmission electron microscopy (TEM, n = 10) 4% paraformaldehyde and 1% glutaraldehyde in 0.1 m PB (pH 7.4). Tissues were sectioned with a vibrating microtome (40–60 μm). TEM studies were performed in postfixed tissues (1% OsO₄ in 0.1 m PB, pH 7.4) that were epoxy-resin embedded. Ultrathin sections were contrasted with uranyl acetate and lead citrate. Golgi-impregnated cords were processed as described elsewhere (Fernández et al. 1993).

Immunocytochemistry

We assayed the following primary antibodies from Chemicon: mouse anti-neuronal nuclei (NeuN, 1/500); rabbit anti-neurofilament M (1/200); rabbit anti-glial fibrillary acidic protein (GFAP, 1/500); mouse anti–GAD (1/500) and mouse anti-Vimentin (1/400). We also used anti-human neuronal protein (HuC/D, 1/30, Molecular Probes); rabbit anti-S100 protein (S100, 1/100, Sigma) and rabbit anti-GABA (1/5000, ImmunoStar). Sections were blocked with 0.5% bovine serum albumin (BSA) in PB (1 h) and incubated with the primary antibody (in PB and 0.3% Triton X-100) overnight. After blocking, the tissues were incubated in secondary antibodies (1/200, 1 h in the dark) conjugated either with horseradish peroxidase (HRP) or fluorophores. HRP was visualized using-3′,3′-diaminobenzidine or other chromogens. Control experiments replacing antibodies by pre-immune serum were performed.

Slice preparation

Anaesthetized turtles were decapitated and the blood removed by intraventricular perfusion with Ringer solution (6–10°C). The cervical enlargement was dissected out and transverse 300 μm slices were cut, placed in a chamber (1 ml volume) and superfused (1 ml min⁻¹) with Ringer solution of the following composition (mm): NaCl 96.5, KCl, 2.6 NaHCO₃ 31.5, CaCl₂ 4, MgCl₂ 2 and glucose 10. The solution was saturated with 5% CO₂ and 95% O₂ (pH 7.6). All experiments were performed at room temperature (20–22°C).

Electrophysiology

Cells were visualized with differential interference contrast (DIC) optics (Leica DM LFS). Patch-clamp whole-cell recordings were obtained with electrodes filled with (mm): K-gluconate 122; Na₂-ATP 5; MgCl₂ 2.5; EGTA 1; Mg-gluconate 5.6; K-Hepes 5; H-Hepes 5 and biocytin 10; pH 7.4; 8–12 MΩ. In some experiments, biocytin conjugated with Alexa 488 (1%, Molecular Probes) was used. Current-clamp recordings were performed with an Axoclamp 2B (Axon Instruments) driven by a programmable stimulator (Master-8, AMPI). Liquid junction potentials were determined and corrected (Barry & Diamond, 1970). In some experiments, tetra-ethyl-ammonium (TEA), (–)-bicuculline methiodide, 6,7-dinitroquinoxaline-2,3(1H,4H)-dione (DNQX) and dl-2-amino-5-phosphono-valeric acid (AP5) from Sigma and tetrodotoxin (TTX) from Alomone were added to the bath.

Values are expressed as mean ± standard error of the mean (s.e.m.). Statistical significance (P < 0.05) was evaluated using the Mann–Whitney U-test.

Immunocytochemistry of electrophysiologically characterized cells

Slices were fixed in 4% paraformaldehyde in 0.1 m PB (pH 7.4). Cells were examined directly with an epifluorescent microscope (Olympus BX60) or by using streptavidin conjugated with different fluorophores or HRP. In selected cases, slices were treated with different primary and secondary antibodies as described above. The slices were whole-mounted and examined with confocal (Leica, TCS SP2) or epifluorescence microscopy.

Results

Specific cell markers revealed two main cell types

Within the CG we found cells that expressed the glial and ependymal cell protein S100, and cells that expressed proteins of the Hu/elav family found in immature neurones (Fig. 1A–C). When compared with Golgi-impregnated material, S100-positive (S100⁺) cells exhibited morphological features of radial glia (RG): a process contacting the CC, a cell body lying at variable distances from the CC and a long process reaching the pia (Fig. 1D). The HuC/D positive (HuC/D⁺) cells resembled the bipolar cells revealed by the Golgi procedure (Fig. 1E). HuC/D⁺ cells had a short apical process contacting the CC and basal processes ramifying within the peri-ependymal mantle. Although we observed cells in the grey matter expressing proteins typical of mature neurones, like NeuN (Fig. 1F, arrow) and neurofilament M (not shown), ependymal cells were not reactive to these markers. The study of 32 spinal cord sections showed no NeuN-positive cells within a circular area (30 μm in radius) centred on the CC (Fig. 1F). However, the same area showed many Huc/D⁺ cells. In addition, antibodies detecting proteins expressed in mature glia (GFAP) or against vimentin did not reveal reactive cells in the ependyma (not shown).

Figure 1. Cell phenotypes in the CG.

A–C, S100⁺ and HuC/D⁺ cells were found around the CC (asterisks). C represents the merged A + B images. Arrows indicate same cells in all panels. D, S100⁺ cells are identical to Golgi-stained RG. E, HuC/D⁺ cells are closely similar to Golgi-stained immature neurones. F, NeuN antibody failed to label cells in the ependyma, but some NeuN⁺ cells (arrow) are present in the peri-ependymal mantle.

Different electrophysiological phenotypes correlate with specific cell markers

Whole-cell patch-clamp recordings revealed three main electrophysiological phenotypes. A group of cells displayed passive responses to a series of current pulses, with a linear current–voltage relationship (Fig. 2Aa and b), relatively low input resistance (373.7 ± 65.2 MΩ; n = 16) and large resting potential (−85.9 ± 1.9 mV; n = 16). These cells had morphological features of RG and displayed extensive dye coupling, forming cell clusters (Fig. 2Ac). Engulfed by the cluster there were unstained profiles presumably corresponding to HuC/D⁺ cells (Fig. 2Ac, asterisks). In 2 out 10 cases, we recovered isolated cells with the electrophysiological and morphological phenotype of RG (not shown). Clustered cells expressed S100 proteins (Fig. 2Ac, panels on the right).

Figure 2. Two types of non-spiking cells in the CG.

A, responses to a series of current pulses of a cell close to the CC (a) that had a linear current–voltage relationship (b). The recorded cell was difficult to identify due to extensive dye coupling (c). Notice the ghosts of uncoupled cells (asterisks). Panels on the right show corresponding confocal optical sections of a cell (from a different cluster) that expressed S100. Scale bar: 5 μm. B, a different cell type characterized by a high input resistance and an outward rectification. C, the sag in response to depolarizing current pulses was blocked by 10 mm TEA (a). The current–voltage relationships before and after TEA are shown in b. D, a representative biocytin-filled cell (a, arrow) with a high input resistance and outward rectification (a, upper inset; calibration bars: 20 mV, 50 pA, 200 ms). Immunocytochemistry for S100 in the same section is shown in b. The arrow points to the recorded cell and the asterisks indicate S100-negative profiles. The merged images (a + b) are shown in c. Main panels are 17 confocal stacked optical sections. The insets (lower right) show corresponding 1 μm-thick optical sections of the cell (arrowhead).

We found a second group of cells with high input resistance (3.5 ± 0.55 GΩ; n = 19) and resting potential (−68.3 ± 3.7 mV; n = 19), that were significantly different (P < 0.01) from those of RG. These cells showed a delayed rectification in response to depolarizing current pulses (Fig. 2B). Addition of TEA (10 mm, n = 5) to the bath eliminated the outward rectification (Fig. 2Ca and b). The morphology and electrophysiology (inset) of a representative cell is shown in Fig. 2Da. The cell had a bipolar shape with a main process contacting the CC and a thin one entering the parenchyma (Fig. 2Da, arrow; arrowhead in inset). Immunocytochemistry revealed that the recorded cell expressed S100 and was surrounded by other S100⁺ cells (Fig. 2Db and c, arrow).

Finally, a third group of cells was distinguished by their ability to fire action potentials. A subpopulation (n = 15) fired a single full spike in response to sustained depolarization (Fig. 3A) whereas others (n = 31) fired repetitively (Fig. 3B). The resting potentials of single-spiking cells (−50 ± 2.4 mV; n = 11) and repetitive-spiking cells (−57 ± 1.1 mV; n = 28) were significantly different (P < 0.01). However, their input resistances (single-spiking: 5.5 ± 0.7 GΩ; n = 14; repetitive-spiking: 6 ± 0.3 GΩ; n = 31) did not differ significantly. Thirty-three per cent of single-spiking and 74% of repetitive-spiking cells showed spontaneous firing at their resting potential. The robustness of repetitive firing varied among cells. As shown in Fig. 3Bb, some cells fired for several seconds without spike amplitude attenuation. Other cells (13 out of 31), showed spike amplitude attenuation and eventually stopped firing after 1–2 s. As shown by the instantaneous frequency plot, repetitive-spiking cells displayed a slow developing frequency adaptation (Fig. 3Bb). Single- and repetitive-spiking cells had morphological characteristics resembling those of HuC/D⁺ cells (Fig. 3Bc).

Figure 3. HuC/D⁺ cells fire action potentials.

A, electrophysiological properties of a cell that fired a single full spike in response to sustained depolarization. B, electrical properties of a repetitive spiking cell (a). This cell fired to prolonged depolarization (5 s) with little spike amplitude attenuation (b). The instantaneous frequency plot evidenced a slow-developing frequency adaptation. The morphology of the cell is shown in c. C, action potential waveforms of a single spiking (a) and a repetitive spiking (b) cell. D, action potentials (a) were blocked by 1 μm TTX (b). E, DIC image of a cell (arrowhead) which fired repetitively (a, inset). Confocal corresponding images of the same cell are shown in b revealed with Alexa 596, and in c after immunocytochemistry for HuC/D. The merged images (b + c) are shown in d. Calibration bars in inset: 20 mV, 100 ms.

We found conspicuous differences in the spike waveforms of single- and repetitive-spiking cells. The spike amplitude (36.7 ± 4 mV; n = 14) and half amplitude duration (5 ± 0.41 ms; n = 14) of single-spiking cells were significantly different from those of repetitive-spiking cells (69.5 ± 2.2 mV and 3.4 ± 0.2 ms; n = 31). However, the most noticeable difference was the slow after-hyperpolarization (sAHP) observed in repetitive-spiking cells (Fig. 3Cb), but absent in single-spiking cells (Fig. 3Ca). In all cells tested (n = 16), the action potentials were completely abolished by TTX (1 μm, Fig. 3D).

As suggested by their morphology, spiking cells expressed the early neuronal marker HuC/D. Figure 3Ea shows the DIC image of a repetitive firing cell (Fig. 3Ea, inset) and its confocal image after revealing biocytin with Alexa 596 (Fig. 3Eb). The recorded cell was HuC/D⁺ as confirmed by the confocal images (Fig. 3Ec and d).

Putative immature neurones receive functional synaptic contacts

Since neurotransmitters influence various aspects of neuronal maturation in the developing brain (Owens & Kriegstein, 2002; Spitzer et al. 2002), we searched for fine structure evidences of synaptic contacts on immature neurones. TEM revealed typical synapses on cells with the phenotype of immature neurones (Fig. 4A). Double-labelling experiments demonstrated that HuC/D⁺ cells receive GAD-positive terminals (Fig. 4B). Some of these synapses were functional since in all spiking cells we observed spontaneous postsynaptic potentials (Fig. 4C and D, also see Fig. 3A and B). Spontaneous inhibitory postsynaptic potentials (IPSPs; mean frequency 3.5 ± 1.6 IPSPs s⁻¹, n = 18) were detected in 19 cells (Fig. 4C, arrowheads; see also Fig. 3Bb). Figure 4 also shows a barrage of excitatory postsynaptic potentials (EPSPs; mean frequency 45.4 ± 3.9 EPSPs s⁻¹, n = 32) leading to spike firing (Fig. 4C, asterisks). In all tested cells (n = 16), TTX strongly reduced spontaneous events supporting their synaptic origin (Fig. 4Da). As suggested by immunocytochemistry (Fig. 4B), spontaneous IPSPs were mediated by GABA since they were blocked by bicuculline (Fig. 4Db; n = 3), a selective GABA_A receptor antagonist. In the cell shown in Fig. 4Db, the spontaneous IPSPs (upper trace, arrows) were abolished by bicuculline (20 μm), and conspicuous EPSPs were seen in the presence of the blocker (lower trace). The time course of spontaneous EPSPs in the presence of bicuculline (Fig. 4Db, lower trace; duration at half-amplitude: 27 ± 1.7 ms, n = 10) and in normal medium (duration at half-amplitude: 22 ± 2.6 ms, n = 10) did not differ significantly, and was similar to that of monosynaptic EPSPs in rat spinal interneurones (Yoshimura & Jessell, 1990) and turtle motoneurones (Delgado-Lezama et al. 2004). Most spontaneous EPSPs in CC-contacting cells were mediated by glutamate receptors. Figure 4Dc shows a cell with a high rate of spontaneous EPSPs in normal medium (control). Addition of DNQX (20 μm) and AP5 (100 μm) blocked most synaptic activity (Fig. 4Dc, n = 4). Since in this cell spontaneous IPSPs were very rare (0.03 IPSPs s⁻¹) even when the membrane potential was depolarized to facilitate the detection of hyperpolarizing events (not shown), the vast majority of synaptic activity observed in the presence of DNQX and AP5 was non-glutamatergic EPSPs (Fig. 4Dc, lower trace).

Figure 4. HuC/D⁺ cells receive functional synaptic contacts.

A, electronmicrograph of a synapse contacting an immature neurone. The upper inset shows synapse location followed in 10 serial sections (calibration bar: 1 μm). Asterisks indicate the synaptic bulb. Arrowheads mark the synapse portion enlarged in the lower inset. B, GAD immunocytochemistry revealed GABAergic terminals (arrows) upon HuC/D⁺ cells. C, spontaneous synaptic potentials recorded in a repetitive spiking cell. The arrowheads point to large IPSPs. A barrage of EPSPs (asterisks) led to spike firing. The action potentials are clipped off. D, the spontaneous synaptic potentials were abolished by 1 μm TTX (a). In a representative cell with a high spontaneous rate of IPSPs (b, arrows), bicuculline (20 μm) eliminated the IPSPs. DNQX (20 μm) and AP5 (100 μm) eliminated most spontaneous EPSPs (c).

Discussion

Based on electrophysiological, anatomical and molecular clues, we described here three main types of cells lining the CC. Strikingly, in addition to RG with functional properties of precursor-like cells, we present compelling evidence that some cells contacting the CC are immature neurones at different developmental stages. Our findings point to the turtle spinal cord as a unique model to study the generation, differentiation and integration of new neurones into operational spinal circuits.

S100⁺ cells: precursor-like cells in the CG?

An important number of cells lining the CC in the turtle were reactive to S100, a marker for ependymal and glial cells in different vertebrates (Didier et al. 1986; Martens et al. 2002; Dervan & Roberts, 2003). Electrophysiology revealed that S100⁺ cells constitute a heterogeneous population. Similar to embryonic cortex (Noctor et al. 2002), some cells in the CG displayed electrophysiological properties typical of RG, with relatively low input resistance, large negative resting potentials and lack of active responses. These cells had the morphology of RG and were usually found forming clusters of dye-coupled cells, suggesting the presence of gap-junctions. Interestingly, the properties of RG in the CG closely resemble those of cycling precursors in the embryonic cortical ventricular zone (Lo Turco & Kriegstein, 1991; Noctor et al. 2002). In a few cases, cells with electrophysiological and morphological features of RG were found uncoupled, suggesting a different functional state. In the developing cortex, RG couple and uncouple depending on the phase of the cell cycle (Bittman et al. 1997). Thus, the uncoupled RG in the CG may represent a subset of cells in a particular phase of the cell cycle (M phase, Bittman et al. 1997). In the CG, RG did not express vimentin, a marker usually linked to these cells. However, the expression of RG markers differs with brain regions, developmental stage and vertebrate species (Campbell & Götz, 2002). Therefore, our findings suggest that as in the brain of juvenile turtles (Weissman et al. 2003), RG persists in the postnatal spinal cord and display functional properties similar to precursor-like cells. Even though the properties of RG in the CG resemble those of neural progenitors in the ventricular zone of the mammalian brain, their ability to generate both glial cells and neurones remains to be confirmed by line age analysis.

A second type of S100⁺ cell had a higher input resistance, a TEA-sensitive outward rectification and was never found forming clusters. Voltage-dependent K⁺ channels have been implicated in the differentiation and migration of precursor cells (Hendriks et al. 1999). Therefore, this group of S100⁺ cells could be a very immature cell type differentiating to the glial or neuronal lineages. Alternatively, it may be a distinct type of mature glial cell contacting the CC. Indeed, in the early postnatal development of the rat spinal cord, several types of glia with a variety of voltage-dependent K⁺ conductances have been described in the grey matter (Chvátal et al. 1995).

HuC/D⁺ cells: immature neurones in the CG?

The presence of neurones contacting the CC has been proposed in a variety of species including mammals (Vigh & Vigh-Teichmann, 1998). Some of the evidence stems from the fact that these putative neurones contain GABA and receive synaptic contacts (Roberts et al. 1995; Fernández et al. 2002). However, these features are not unambiguously related to neurones since glia may also contain neurotransmitters (Araque et al. 2001) and receive synaptic contacts (Bergles et al. 2000). In the present study, we show the first functional evidence that some cells contacting the CC belong to the neuronal lineage, i.e. they fire action potentials (Carleton et al. 2003). Spiking cells expressed the early neuronal marker HuC/D (Marusich & Weston, 1992), but not other proteins found in mature neurones. We have previously shown colocalization of BrdU and the neuronal marker TUC-4, suggesting that there is neurogenesis in the CG (Fernández et al. 2002). Therefore, spiking cells may be immature neurones generated in the CG after birth.

The spike of CC-contacting cells was brief and sensitive to TTX, indicating that similarly to rodents and chicks, it is mediated by Na⁺ channels even at the earliest differentiation stages (Spitzer et al. 2000). A higher density of Na⁺ channels may account for the larger action potential observed in repetitive-spiking cells. On the other hand, repetitive firing could indicate a more advanced stage of development of HuC/D⁺ cells. Indeed, the differences between the action potential waveform of single- and repetitive-spiking cells parallel those of the action potential of rat motoneurones at different stages of early postnatal development (Gao & Ziskind-Conhaim, 1998). In developing spinal neurones (Spitzer et al. 2000), repetitive firing may appear as a consequence of the development of Ca²⁺-dependent K⁺ channels, which are known to generate the sAHP needed to remove Na⁺ channel inactivation (Russo & Hounsgaard, 1999).

Our study showed anatomical and functional evidence that putative immature neurones are connected to spinal circuits. As in the embryo, the synaptically driven spiking observed in some cells may be important for the differentiation of excitability in CC-contacting cells (Spitzer et al. 2000, 2002). The fact that glutamate receptor antagonists failed to completely block spontaneous EPSPs indicates that transmitters other than glutamate also contribute to synaptic excitation. We also found a rich GABAergic plexus with terminals contacting some HuC/D⁺ cells. These GABAergic synapses were functional since we observed spontaneous IPSPs mediated by GABA_A receptors. It has been proposed that in early neurogenesis, GABA operates as a depolarizing signal modulating neuronal differentiation and migration (Owens & Kriegstein, 2002). Therefore, it may be possible that GABA could be an excitatory signal also for HuC/D⁺ cells. To test this possibility, a recording technique that does not change the internal Cl⁻ concentration should be used.

Summarizing, our finding that RG display properties of precursors suggests that they may be a source of new cells in the spinal cord of young turtles. Some newborn cells seem to be committed to the neuronal lineage as indicated by the presence of neurones at early maturational stages still connected to the CC. It is tempting to speculate that immature neurones in the CG migrate to integrate to existing circuits in the dorsal or ventral horns.