Regeneration of Xenopus laevis spinal cord requires Sox2/3 expressing cells

Abstract

Spinal cord regeneration is very inefficient in humans, causing paraplegia and quadriplegia. Studying model organisms that can regenerate the spinal cord in response to injury could be useful for understanding the cellular and molecular mechanisms that explain why this process fails in humans. Here, we use Xenopus laevis as a model organism to study spinal cord repair. Histological and functional analyses showed that larvae at pre-metamorphic stages restore anatomical continuity of the spinal cord and recover swimming after complete spinal cord transection. These regenerative capabilities decrease with onset of metamorphosis. The ability to study regenerative and non-regenerative stages in Xenopus laevis makes it a unique model system to study regeneration. We studied the response of Sox2/3 expressing cells to spinal cord injury and their function in the regenerative process. We found that cells expressing Sox2 and/or Sox3 are present in the ventricular zone of regenerative animals and decrease in non-regenerative froglets. Bromodeoxyuridine (BrdU) experiments and in vivo time-lapse imaging studies using green fluorescent protein (GFP) expression driven by the Sox3 promoter showed a rapid, transient and massive proliferation of Sox2/3⁺ cells in response to injury in the regenerative stages. The in vivo imaging also demonstrated that Sox2/3⁺ neural progenitor cells generate neurons in response to injury. In contrast, these cells showed a delayed and very limited response in non-regenerative froglets. Sox2 knockdown and overexpression of a dominant negative form of Sox2 disrupts locomotor and anatomical-histological recovery. We also found that neurogenesis markers increase in response to injury in regenerative but not in non-regenerative animals. We conclude that Sox2 is necessary for spinal cord regeneration and suggest a model whereby spinal cord injury activates proliferation of Sox2/3 expressing cells and their differentiation into neurons, a mechanism that is lost in non-regenerative froglets.

INTRODUCTION

In mammals, including humans, spinal cord injury (SCI) results in loss of motor and/or sensory function below the level of the injury, leading to paraplegia and quadriplegia. Approximately 130,000 individuals/year have an acute SCI, joining a total of 2.5 million patients living with chronic paralysis (reviewed in Thuret et al., 2006; Barnabe-Heider and Frisen, 2008). SCI results in a massive loss of neurons, oligodendrocytes and astrocytes and almost no tissue regeneration or functional recovery is observed (reviewed in Tuszynski and Steward, 2012; Sabelström et al., 2013). Several therapies tested in clinical settings have had only limited effects on functional recovery indicating that further understanding of the basic mechanisms underlying regeneration is required (reviewed in Tetzlaff et al., 2011).

Unlike mammals, teleost fish, amphibians such as adult urodeles (e.g. newts) and anuran larvae (e.g. Xenopus) can achieve functional recovery after spinal cord transection (Filoni et al., 1984; Becker et al., 1997; Tanaka and Ferretti, 2009; Díaz-Quiroz and Echeverri, 2013; Lee-Liu et al., 2013). In mammals, regeneration and tissue repair are restricted to the early stages of life and are progressively lost during development (reviewed in Ferretti et al., 2003; Godwin, 2014). Anuran amphibians such as Xenopus laevis represent an interesting case of stage-dependent spinal cord regeneration. Histological studies have shown that X. laevis larvae before or at the beginning of metamorphosis (stage 50–54) have strong regenerative capabilities, which are no longer present in post-metamorphic stage 66 froglets (Sims, 1962; Forehand and Farel, 1982; Filoni et al., 1984; Beattie et al., 1990; Gibbs et al., 2011; Gaete et al., 2012). Nevertheless, no detailed characterization of how these animals respond to injury nor a method to measure functional recovery has been reported and a comparison between the cellular and genetic mechanisms involved in spinal cord regeneration during these two different stages is largely incomplete.

The presence of regenerative and non-regenerative stages, its external development, its robustness for experimental manipulations, the easy generation of hundreds of larvae and froglets together with the recent advances in genomics and genetics makes Xenopus laevis a unique model organism to study regenerative biology (reviewed in Harland and Grainger, 2011). We recently performed a transcriptome-wide profile analysis comparing the response to injury between regenerative and non-regenerative stages, and demonstrated that at each stage a very different gene repertoire is deployed in response to injury (Lee-Liu et al., 2014). Among other gene ontology groups we have found that neurogenic genes respond differentially between these two stages suggesting that activation of neural stem and progenitor cells and their differentiation into neurons could play an important role in spinal cord regeneration.

Sox2 and Sox3 are members of the SoxB1 family of SRY-related transcription factors. Sox2 is expressed in early stages in the developing nervous system, in neural stem and progenitor cells and in most epidermal and ectodermal stem cells (Kamachi and Kondoh, 2013; Sarkar and Hochedlinger, 2013). Sox2 also plays a key role in early development, nervous system development, pluripotency of stem cell biology, reprogramming, adult neurogenesis and tissue homeostasis (Avilion et al., 2003; Ellis et al., 2004; Ferri et al., 2004; Masui et al., 2007; Pevny and Nicolis, 2010; Arnold et al., 2011; Kamachi and Kondoh, 2013; Thomson et al., 2011). Sox3 is also expressed in fish, frog and chick throughout the ectoderm before neural induction, and is then limited to the neuroectoderm (Brunelli et al., 2003; Koyano et al., 1997; Okuda et al., 2006; Rex et al., 1997; Rogers et al., 2008). Together, Sox3 and Sox2 are required for the maintenance of a neural stem and progenitor cell pool and proper development of the nervous system (Bylund et al., 2003; Sandberg et al., 2005; Mizuseki et al., 1998; Rogers et al., 2009). We and others have recently demonstrated that Sox2⁺ cells are also necessary for spinal cord regeneration after tail amputation in X. laevis and axolotl (Gaete et al., 2012; Fei et al., 2014).

Here we aimed to further characterize X. laevis to study SCI and to examine the role of Sox2/3 expressing cells in spinal cord regeneration. We provide a detailed histological characterization of the response to spinal cord transection at different stages before and during metamorphosis. We adapt a test to measure swimming as an indicator of functional restoration, allowing its comparative analysis with anatomical and histological recovery. We used immunofluorescence in fixed tissue and perform studies in live animals to show that Sox2/3 expressing cells are present in the ventricular layer of the spinal cord throughout metamorphosis, albeit with decreasing levels of Sox2/3 protein and the number of cells expressing these genes. In response to SCI, a massive and rapid activation of Sox2/3 expressing cells occurs in regenerative animals. In contrast, a very slow and partial activation of these cells is observed in non-regenerative animals. Our functional studies using morpholino knockdown and a dominant-negative Sox2 transgenic line demonstrate that the activation of Sox2/3 expressing cells is necessary for regeneration.

MATERIALS and METHODS

Growth and manipulation of Xenopus laevis

Frogs obtained from Nasco, were subjected to in vitro fertilization and cultured as previously described (Gaete et al., 2012) until they achieved Nieuwkoop and Faber stages 49 to 54 for regenerative stages (R-stages), and stage 56 to 66 for non-regenerative stages (NR-stages). All animal procedures were approved by the Committee on Bioethics and Biosafety from the Faculty of Biological Sciences, Pontificia Universidad Católica de Chile. For spinal cord transection, tadpoles and froglets were operated as described (Gaete et al., 2012; Lee-Liu et al., 2014).

Hematoxylin and eosin staining

Histological analyses were performed in paraffin embedded sections. For this, animals were fixed in 4% paraformaldehyde (PFA). Tissue was delimited and dehydrated in methanol solutions of increasing concentration (25 to 100%), followed by isopropanol and paraffin embedding. 10 μm longitudinal sections were obtained and placed in silanized slides. For hematoxylin and eosin staining, slides were deparaffinized with xylene followed by dehydration in ethanol solutions (100% to 50%), after which slides were incubated for 2 minutes in hematoxylin, subsequent ethanol solutions and counter stained in eosin for 30 seconds. These were followed by washes in ethanol and xylene. A coverslip was placed on the slide and sealed with entellan mounting medium.

In situ Hybridization and Immunofluorescence

Animals were fixed in 4% PFA for 2 hrs at room temperature or overnight at 4°C. Delimited tissue was processed for cryosections by embedding in sucrose solutions of increasing concentrations (5–10–20%), followed by optimal cutting temperature compound (OCT), frozen in liquid nitrogen and sectioned at 10 μm. Samples were mounted on silanized slides.

In situ hybridization of sections was performed as described (Lemaire and Gurdon, 1994; Sive et al., 2000), with minor modifications: cryosections were pre-heated on a heating plate at 65°C overnight before starting the protocol. Sections were incubated with proteinase K for 20 minutes. The neuroD and neurogenin3 probes were used at 1:200. Anti-digoxigenin AP Fab fragment was used (1:10000, Roche, 11093274910).

To perform immunofluoresence, samples were permeabilized in phosphate-buffered saline (PBS) + 0.2% Triton X-100 (PBST), blocked for 30 min in PBST + 10% goat serum, incubated with primary antibody overnight, and secondary antibodies for 2 hrs at room temperature in the same blocking solution, washed 3 times for 10 minutes after each antibody incubation with PBST, and followed by DNA staining and mounting with vectashield (Vector Laboratories, H-1000). For DNA staining, sections were incubated for 20 minutes in TOTO3 (1:1,000, Molecular Probes, T3604) in 1X PBS at room temperature. Antibodies use in this study are rabbit polyclonal anti-Sox2 (1:200, Cell Signalling Technology, 2748S), mouse anti-Sox2 (1:200, Cell Signaling Technology, 4900S), mouse monoclonal anti-acetylated tubulin (1:500, Sigma, T7451), anti-Doublecortin (Dcx, 1:500, Cell Signaling Technology, 4604S), AlexaFluor 488 or 555 (1:500) as a secondary antibody. For isolation and observation of the complete spinal cord (flat mount), spinal cords were dissected and immunofluorescence was performed as described above, except that 0.2% was replaced with 1% PBS/Triton X100.

Proliferation Assay

For proliferation assays, the thymidine analogue bromodeoxyuridine (BrdU) was added to the culture medium (0.1X Barth) to a final concentration of 400 μM for 16 hrs (Fig. 4) or 4 hrs (Fig. S3). Double labeling for Sox2 and BrdU was performed as described (Yoshino and Tochinai, 2004). Mouse monoclonal anti-BrdU was used (1:50, Roche, 11170376001). In experiments with the thymidine analogue chlorodeoxyuridine (CldU, MP Biomedicals, LLC, 105478), tadpoles were incubated with 3.8 mM CldU in 0.1X Barth for 4 hrs after electroporation with X.Tropicalis Sox3 promoter driving GFP (Sox3::GFP). Spinal cords were dissected, rinsed, embedded in gelatin and cut into 30 μm sagittal sections using a vibratome. For CldU labeling, spinal cord sections were blocked in 10% goat serum in PBST for 1 hr, then incubated with rabbit anti-GFP (1:500, Chemicon, AB3080) for 2 hr, rinsed in PBST, and treated with 2N HCl for 1 hr at 37°C, rinsed in PBST and blocked again in 10% goat serum in PBST for 1 hr before incubating overnight at 4°C with rat anti-CldU (1:400, OBT0030G). The CldU signal was enhanced with biotin tagged goat anti-Rat secondary antibody (1:500, Jackson ImmunoResearch) followed by detection with streptavidin-Alexa Fluor 546 (1:500, Invitrogen, S11225). The sections were mounted and imaged using confocal microscopy.

Spinal cord electroporation

Stage 50 tadpoles were anesthetized in 0.02% ethyl 3-aminobenzoate methanesulfonate (MS222), immobilized on a Sylgard elastomer platform (Dow Corning, Wiesbaden, Germany) and injected 5 or 10 times with 1 mM morpholinos (4 nl/injection) in the ependymal zone of the spinal cord using a pulled glass capillary. Morpholinos were electroporated into the spinal cord by applying five 40 V pulses (pulse length: 50 msec, frequency: 300 pps) in each polarity, across the back with home made platinum electrodes, using a SD9 Stimulator (Grass Tele-factor, USA) (Echeverri and Tanaka, 2003). Animals were transected and allowed to recover in 0.1X Barth containing antibiotics.

DNA electroporation and in vivo 2 Photon Imaging

Albino stage 50 animals were anesthetized in 0.02% MS222 and the plasmid Sox3pXt-GFP (4 μg/μl) was injected with a glass capillary into the central channel of the spinal cord. Voltage pulses were applied with a Grass SD9 stimulator across the back using platinum electrodes (5 pulses of 35V in each polarity). The next day or the second day after electroporation, animals were screened and anesthetized in 0.01 % MS222 for in vivo imaging once a day over the next 10 days by 2 photon microscopy (Javaherian and Cline, 2005; Bestman et al 2012). Tadpoles were placed in a custom-built chamber with the coverslip directly on the surface of their backs.

Spinal cord isolation

Spinal cord samples were dissected from tadpoles and froglets at 1, 2 or 6 days post-injury (dpt) or days post-sham operation (dps). In tadpoles, this segment went from the start of the spinal cord including the lesion site up to the middle of the tail (approximately of 9 mm long). In froglets we isolated the segment from the hindbrain, including the gap area, to the end of the spinal cord (approximate of 3 mm long). For RT-qPCR we isolated a fragment from the midpoint of the transection site (Lee-Liu et al., 2014).

RT-qPCR and Western blot

Total RNA was obtained from dissected spinal cord samples and analyzed by RT-qPCR as described (Lee-Liu et al, 2014). For western blotting spinal cords from stages 50 (n=10), 58 (n=5) and 66 (n=3) were isolated and homogenized in RIPA lysis buffer with protease inhibitors. Primary antibodies against Sox2 (1:2000, Cell Signaling Technology, 2748S), Dcx (1:2000, Cell Signaling Technology, 4604S) and tubulin (1:200000, Sigma, T8535) were used.

Morpholinos

Sox2-lissamine-morpholino (5′-GTACAGCATGATGGAGACCGAGCT-3′, Gene Tools, Philomath, OR; Van Raay, 2005), Sox2a-lissamine-morpholino (5′-GCCGCCTCGATGTACAGCATGATGG-3′, Gene Tools, Philomath, OR) Sox3-lissamine-morpholino (5′-GCTCCAAATGTATAGCATGTTGGAC-3′, Gene Tools, Philomath, OR) and standard control-lissamine morpholino (5′-CCTCTTACCTCAGTT-3′, Gene Tools, Philomath, OR) were used.

Transgene constructs and transgenesis

The pHSP70::SOX2BD(−)GRHA::EGFP (dnSox2) was generated from pISceI- HSP70::SOX2BD(−)GRHA and pISceI-HSP70::EGFP (Gaete et al., 2012). Transgenesis was performed using the Restriction Enzyme Mediated Integration method (REMI, Ishibashi et al., 2007a,b,c). One-cell stage embryos were injected with 50 pg of pHSP70::SOX2BD(−)GRHA::EGFP. Embryos were incubated until stage 42 at 18°C, and changed afterwards to 23°C. At stage 50 tadpoles were heat shocked (30 minutes at 34°C) to select the EGFP⁺ animals and they were allowed to develop for 1 year until sexual maturity. F1 tadpoles were generated through natural mating. Once they reached stage 50, they were heat shocked (30 minutes at 34°C) for 3 days (−1, 1 and 2 dpt) and treated with 10 μM dexamethasone dissolved in 0.1X Barth (see Fig. S6A). They were analyzed by video-tracking at 0, 10, 15 and 20 days post-injury. At 15 dpt two animals were taken per time point to perform immunofluorescence in flat-mounted spinal cords (Fig. S6B).

Swimming ability tracking

Swimming capability was evaluated by video-tracking free swimming during 5 minutes at 0, 5, 10, 20 and 30 dpt. Animals were placed in a 15 cm diameter plate filled with water and recorded with a video camera. ANY-maze program (Stoelting Co, Wood Dale, IL) was used to track swim paths (Guo et al., 2011). Mean lengths of the swim paths (“total distance”) of the three trials were used for graphical presentation and statistical analysis with Graph Pad PRISM.

Imaging and statistical analysis

Samples were photographed using a confocal microscope (FV-1000 Olympus Confocal Laser Scanning Microscope). Intensity was measured using Z- plot analysis and fire LUT pseudocoloring of ImageJ (NIH). Total cell counts, including quantification of colocalization, were determined using the ImageJ (NIH) cell counter plugin. Results for intensity were analyzed by two-way analysis of variance (ANOVA) and Bonferroni test, whereas cell counting was analyzed by one-way ANOVA and Tukey post-hoc test. Results for swim tracking were analyzed by 1 way ANOVA and Bonferroni Multiple Comparison Test. Error bars in all figures indicate the “standard error of the mean” (SEM). Differences were considered statistically significant at *P <0.05, **P <0.01, and ***P <0.001 of at least 3 independent experiments.

RESULTS

Characterization of regenerative and non-regenerative stages

Previous studies have described the decreased regenerative capacity of the spinal cord during metamorphosis (Sims, 1962; Michel and Reier, 1979; Filoni et al., 1984; Bettie et al., 1990, Gibbs et al., 2011; Gaete et al., 2012), however, the value of Xenopus laevis as a model organism to study spinal cord regeneration would be significantly enhanced by detailed characterization of the cellular response to injury and quantitative methods to measure locomotor recovery at regenerative and non-regenerative stages. We therefore carried out: i) a comparative histological analysis of the cellular response to spinal cord injury (SCI) at pre-metamorphic (stages 50, 54), pro-metamorphic (stage 56) and post-metamorphic (stage 66) stages, and ii) a quantitative measurement of the locomotor recovery after SCI. Results from these studies allow us to correlate functional with histological recovery.

As a model of SCI we performed complete spinal cord transection at the mid-thoracic level by completely disrupting the rostral to caudal continuity of the spinal cord (Filoni et al., 1984; Gaete et al., 2012). For histological analysis we performed hematoxylin and eosin staining of longitudinal sections from animals before and after injury. Pre-metamorphic animals have a small spinal cord (400–600 μm diameter) with an abundance of cells in the ventricular layer, all with a characteristic dark-blue staining. The cell layer has no complex stratification and axons with reddish staining are distributed through the white matter (Fig. 1A,F).

Fig. 1. Characterization of regenerative stages of Xenopus laevis.

Hematoxylin and eosin staining of longitudinal sections of spinal cord from uninjured animals and at 2, 6, 10 and 20 dpt in pre-metamorphic stage 50 (A, B, C, D, E) and 54 (F, G, H, I J) respectively. Magnifications shown in black box for stage 50 (A′, B′, C′, D′, E′) and 54 (F′, G′, H′, I′ J′). Samples of swim trajectories (top, single representative animals for each step) and graphs of the total time animals spent swimming in 5 minutes at stage 50 (K) and stage 54 (L) in sham-operated animals and in animals with spinal cord transection at 0, 5, 10, 20 and 30 dpt. Dotted line: injury site or ablation gap, yellow arrowheads: meningeal layer, arrow: cell clusters, vz: ventricular zone, svz: sub-ventricular zone. Scale bar (A–J) 100 μm; (A′–J′) 30 μm. * p ≤ 0.05.

In pre-metamorphic (stage 50) animals at 2 days post transection (dpt) the rostral stump of the spinal cord was sealed closed whereas the caudal stump remained open. Almost no cellular material was observed in the ablation gap and axonal tracts were completely disrupted (Fig. 1B,G dotted line indicates the injury site). Notably, cell clusters with staining to similar cells in the ventricular layer, together with some fibrilar material were seen in the ablation gap at 6 dpt in stage 50 animals (Fig. 1C and see arrow). At 10 dpt a cellular bridge formed across the ablation gap. Many axons crossed the injury site and a meningeal layer was reconstituted (Fig. 1D,I yellow arrowheads indicate the meningeal layer). At 20 dpt, at least 60% of the transected animals exhibited continuity of the ependymal canal and axon tracts, however the normal morphology was not restored and there were fewer cells in the ventricular layer (Fig. 1E). Animals that were injured at stage 54 demonstrated a similar initial response to injury as seen in stage 50 tadpoles, except that fewer cell clusters were detected at 6 dpt and most of them were associated with the rostral stump (Fig. 1H and see arrow). By 20 dpt many axons crossed the ablation gap, however a continuous ependymal canal had not formed even though several cells populated the ablation gap (Fig. 1J).

We analyzed locomotor recovery at different days after injury by video tracking the swimming distance achieved by tadpoles over 5 minutes. As expected, animals at stage 50 and 54 were unable to move at 0 dpt, but they gradually recovered the ability to swim. Functional restoration was statistically significant at 20 dpt when compared to animals immediately after transection (Fig. 1K and L). Of note, animals after recovery seem to have a different swimming behavior, most of the times swims mainly at the borders of the plate or in short circles suggesting that recovery of the neuronal circuitry is not perfect and that more detailed test are necessary to fully understand spinal cord regeneration. Based on these results we define stage 50 and 54 as regenerative-stages (R-stage), although the efficiency of the cellular and function recovery is lower in stage 54 compared to stage 50.

Similar histological and locomotor analyses were performed in pro-metamorphic (stage 56), metamorphic climax (stage 58) and post-metamorphic stages (stage 66). These animals have a larger spinal cord (800–1600 μm diameter) with more complex organization, composed of a stratified ventricular zone and many axonal tracts (Fig. 2A,F and see Supplementary Information Fig. S1A). In contrast to the response of R-stage animals, the stumps of the spinal cord did not seal closed at 2 dpt in these animals (Fig. 2B,G and see Supplementary Information Fig. S1B). Furthermore, at 6 and 10 dpt cell clusters were scarcely found or completely absent in the ablation gap. Instead, the ablation gap was filled with fibrillary material (Fig. 2C,D,H,I and see Supplementary Information Fig. S1C,D). In addition, in stage 66 froglets, other cells types infiltrated the ablation gap (Fig. 2I). The differences in the response to injury between the younger and older stages were even more evident at 20 dpt because continuity of the ependymal canal or axon tracts were not achieved. Fibrillary material remained in the ablation gap and closure of the stumps was not clearly detected (Fig. 2E,J and see Supplementary Information Fig. S1E). In some cases however, the meninges did appear to be reconstituted, especially in stage 56 and 58 animals (Fig. 2E, see Supplementary Information Fig. S1E).

Fig. 2. Characterization of non-regenerative stages of Xenopus laevis.

Hematoxylin and eosin staining of longitudinal sections of spinal cords from uninjured animals and at 2, 6, 10 and 20 dpt in pro-metamorphic stage 56 (A, B, C, D, E) and post-metamorphic stage 66 (F, G, H, I J) respectively. Black boxes indicate magnifications for stage 56 (A′, B′, C′, D′, E′) and 66 (F′, G′, H′, I′ J′). Samples of swim trajectories (top, single representative animals for each step) and graphs of the total time animals spent swimming in 5 minutes at stage 56 (K) and stage 66 (L) in sham-operated animals and in animals with spinal cord transection at 0, 5, 10, 20 and 30 dpt. Dotted line: injury site or ablation gap, yellow arrowheads: meningeal layer, arrow: cell clusters, vz: ventricular zone, svz: sub-ventricular zone. Scale bar (A–E) 100 μm; (F–J) 200 μm; (A′–J′) 30 μm.

No recovery of swimming ability was observed when the spinal cord was injured at these older stages (Fig. 2K and L), consistent with the anatomical and histological observations. Based on these observations we defined pro-metamorphic animals and froglets as non-regenerative-stage (NR-stage) animals

These studies clearly define R- and NR stages with regards to the ability of X. laevis to restore spinal cord morphology and function following spinal cord transection. This temporal distinction in the capacity for regeneration will facilitate investigation of the genetic and cellular underpinnings that allow spinal cord regeneration in larvae, and why this capability is lost after metamorphosis in froglets and adult X. laevis.

Expression of Sox2 and Sox3 during metamorphosis

The histological data presented above showed that cells in the ventricular layer react differently to injury in R- and NR-stages, suggesting that a response of cells in the ventricular layer is required for recovery from injury. The data further suggest that a differential response of ventricular layer cells to injury in R- and NR-stages may account for the different regenerative capability in these two stages. Our prior work showed that Sox2 is expressed in most cells of the ventricular layer in stage 50 animals (Gaete et al., 2012). The peptide used to generate the antibody against Sox2 is highly conserved between both human and X. laevis Sox2 and Sox3, raising the possibility that both proteins are detected in our experiments. For clarity, when we use this antibody we refer to detection of Sox2/3 to indicate that both Sox2 and Sox3 proteins may be detected. We therefore characterized the expression of Sox2/3 in ventricular layer cells at different stages during metamorphosis using indirect immunofluorescence in longitudinal cryosections from R-stages (stage 50 and 54) and NR-stages (stage 58 and 66).

We have found that at stage 50 Sox2/3 is expressed at high levels in the ventricular layer cells (Fig. 3A,B). Detailed analysis showed heterogeneity of the nuclear shape and size, although cells with an elongated and big nucleus, characteristic of neural stem and progenitor cells, were the most prominent immunolabeled cells seen at this stage (Fig. 3C, N and O). Similar results were observed at stage 54, although Sox2/3 levels were reduced and cells with elongated nuclei were less frequent compared to stage 50 (Fig. 3D–F, N and O).

Fig. 3. Analysis of Sox2/3 expression in the spinal cord in non-regenerative and regenerative stages.

(A–L) Longitudinal cryosections through the spinal cord of animals at stage 50 (A,B,C), 54 (D,E,F), 58 (G,H,I) and 66 (J,K,L), analyzed by immunostaining for Sox2/3 (green) and staining for TOTO3 (nuclei, blue). A,D,G,J. Double labeling for Sox2/3 and TOTO3. B,E,H,K. Sox2/3 only. (C,F,I,L) Higher magnification images of regions shown in white boxes in B,E,H,K. (M) Western blot showing immunolabeling for Sox2/3 in tadpoles (stage 50–58) and froglets (stage 66). α-tubulin immunolabeling is shown as a loading control. (N) Analysis of the signal intensity of Sox2/3 immunolabeling in of B,E,H,K. (O) Analysis of the area of the nuclei of C,F,I,L. cc: central channel, vz: ventricular zone and svz: sub-ventricular zone. Scale bars (A,B,D,E,G,H,J,K) 20 μm; (C,F,I,L) 2 μm. *P <0.05, **P <0.01, and ***P <0.001.

In non-regenerative tadpoles (stage 58) and froglets (stage 66) Sox2/3 immunolabeling was lower and labeled nuclei were round. No elongated nuclei were detected (Fig. 3G–L, N and O). In addition, the Sox2/3⁺ cells were located in more than one layer and, particularly in froglets, were mainly located in a sub-ventricular zone (Fig. 3J).

Sox2/3 protein levels were also analyzed by western blot of spinal cords isolated from R- and NR-stages. In agreement with the immunofluorescence analysis and our previous report (Gaete et al., 2012), higher levels of total Sox2/3 protein were present in the R-stage, although detectable levels of the protein were still found in NR-stage 66 (Fig. 3M). Intriguingly, qRT-PCR studies from isolated spinal cords showed that sox2 and sox3 mRNA levels remained unchanged and even are slightly increased during metamorphosis suggesting that the observed changes in protein level are regulated at a post-transcriptional level (see Supplementary information Fig. S2A, B).

In summary, we conclude that the number of Sox2/3⁺ cells and the levels of Sox2/3 protein, but not the corresponding mRNAs, are higher in the ventricular zone of the spinal cord in R-stages, and that levels decrease during metamorphosis. The shape and location of the Sox2/3+ cells also change during metamorphosis. These developmental decreases in Sox2/3⁺ cell numbers and protein levels correlate with the loss of regenerative capacity, suggesting that Sox2/3+ cells may play a role in the regenerative process.

Proliferative response of Sox2/3⁺ cells to spinal cord injury

The concomitant decrease in Sox2/3⁺ cell numbers and the loss of the regenerative capacity and the fact that Sox2/3⁺ cells are still present in NR-stages prompted us to study how these cells respond to spinal cord transection. To evaluate the effect of injury on the ability of Sox2/3+ cells to proliferate, we incubated animals at stage 50 and 66 with BrdU for 16 hours before injury (uninjured), and at 2, 6, 10 and 20 dpt, fixed and performed double immunofluorescence for BrdU and Sox2/3. Sox2/3⁺ cells incorporate BrdU in R-stage animals before injury (Fig. 4A,B and see Supplementary Information Fig. S2C), reflecting the active proliferative process occurring at this stage. This basal level of proliferation could be explained by the process of neurogenesis that happens during larval stages of development (Schlosser et al., 2002). Contrary to that, no BrdU incorporation was detected in non-injured stage 66 froglets (Fig. 4I, J).

Fig. 4. Proliferation of Sox2/3⁺ cells in response to spinal cord injury.

(A–P) Immunolabeling for BrdU and Sox2/3 in stage 50 or 66 animals at designated times after spinal cord transection show dramatic increases in proliferation in regenerative stages but not non-regenerative stages. Spinal cords were transected and animals were treated for 16 hrs with BrdU and fixed at 2, 6 or 10 dpt. Horizontal sections through the spinal cord were analyzed by double immunofluorescence BrdU (green) and Sox2/3 (red) from uninjured animals and at 2, 6 and 20 dpt in tadpoles stage 50 (A–H) and uninjured, 2, 6 and 10 dpt in froglets stage 66 (I–P). (B,D,F,H,J,L,N,P) Higher magnifications of regions shown in black boxes for A,C,E,G,I,K,M,O. Arrow and arrowheads in C, D and F: BrdU+Sox2/3⁺ cells and BrdU⁺ Sox2/3⁻ respectively. vz: ventricular zone and svz: sub-ventricular zone. Nuclei: TOTO3 (blue). Scale bars are as indicated.

In stage 50 animals at 2 dpt, we observed a massive proliferation of almost 50% of Sox2/3⁺ cells in the ventricular zone (Fig. 4C, D, see arrows and supplementary Fig. S3G). BrdU incorporation was higher in the dorsal side of the spinal cord, possibly because of the higher levels of Sox2/3 in the dorsal spinal cord (Gaete et al., 2012). BrdU incorporation was also more abundant rostral to the injury site compared to the ablation gap or caudal to the injury (see Supplementary information Fig S3A–F). No increase in BrdU incorporation was observed in uninjured animals (see Supplementary Fig. S3A). BrdU was also incorporated in Sox2/3 negative cells, especially in skin and muscle cells (Fig. 4C, see arrowheads).

This proliferation was transient and only detected at 2 dpt. At 6 dpt less than 10% of the Sox2/3⁺ cells were BrdU⁺, and no double-labeled cells were detected at 20 dpt (Fig. 4E–H and see supplementary information Fig. S3G). Notably, at 6 dpt many Sox2/3⁺ cells were found in the ablation gap as self-organized neural tube like structures (Fig. 4E,F and see Supplementary information Fig. S3H).

Similar experiments performed in non-regenerative froglets provided very different results. Most notably, no activation of Sox2/3⁺ cells was observed at 2 dpt, and only BrdU⁺ and Sox2/3 negative cells were found in the ablation gap (Fig. 4K and L). Only at 6 dpt were Sox2/3⁺ and BrdU⁺ cells detected, but the proliferation level was far below that observed in regenerative animals. In froglets, only 15% of Sox2/3⁺ cells were double-labeled with BrdU (Fig. 4M,N and see supplementary information Fig. S3G). At 20 dpt Sox2/3⁺/BrdU⁺ cells were not observed, but proliferation of other cell types persisted at the injury site (Fig. 4O and P). In froglets, Sox2/3⁺ cells were not detected in the ablation gap at any time-point after transection.

We conclude that although Sox2/3⁺ cells are present in both R- and NR-stages, their response to injury is very different. Sox 2/3+ cells exhibit a rapid, transient and massive response in regenerative stages in contrast to a delayed, and very limited response in non-regenerative froglets. In addition, Sox2/3⁺ cells were located in the ablation gap of stage 50 animals, suggesting that Sox 2/3+ cell migrate from the injured ventricular zone into the ablation gap.

In vivo analysis of the response to injury of Sox2/3 expressing cells

To demonstrate that X. laevis R-stage tadpoles are also amenable for in vivo imaging and to study the behavior of Sox2/3⁺ cells in response to injury we performed 2-photon microscopy. For this we used a plasmid containing a 2 kb fragment of the X. tropicalis Sox3 promoter driving GFP expression (pSox3::GFP). Although it is not possible to rule out the expression of this DNA in Sox2/3 negative cells the following evidence suggest that this promoter drives expression of GFP mainly in Sox2/3⁺ cells: i) transgenic tadpoles expressing pSox3::GFP showed GFP expression mainly in the neural plate during early development and in the CNS at later stages (see Supplementary Fig. S4A–D); ii) when this construct was electroporated into the spinal cord in stage 50 animals GFP was mainly detected in cells with radial glia morphology and co-localized with Sox2/3 (Fig. 5B), something that was not observed when we electroporated a construct that drives GFP expression under the control of a constitutively active promoter (data not shown) and iii), the GFP+ cells incorporated CldU in response to injury (Fig. 5C).

Fig. 5. In vivo time-lapse imaging of Sox2/3⁺ cells in response to spinal cord injury.

(A) Schematic representation of in vivo electroporation and imaging by 2 Photon Microscopy of pSox3::GFP in the spinal cord of stage 50 tadpoles. Double immunofluorescence in sagittal sections of the spinal cord for (B) GFP (green), Sox2 immunolabeling (red), (C) CldU (red). Co-localization analysis is shown in (B′ and C′) in a pseudo-color scale. (D–S) Z-stacks of in vivo 2 photon microscope images of GFP-expressing cells after electroporation, show a control sham animal at (D) 1 day after electroporation (dpe), (E) 2 dps, (F) 4 dps, (G) 6 dps. After SCI, the increase in total number of GFP+ cells is shown in (H) at 2 dpe, (I) 2 dpt, (J) 3 dpt, (K) 4 dpt, (H′–K′) magnifications of (H–K). (L–S) Migration of GFP⁺ cells to the gap injury site is shown in (L) at 1 dpe, (M) 1 dpt, (N) 3 dpt, (O) 7 dpt, magnifications of (L–O) in (L′–O′), and no migration in (P) at 1 dpe, (Q) 1 dpt, (R) 3 dpt, (S) 8dpt. Fold change of Sox3-GFP⁺ cells in sham control animals (T), after injury (U) and migration to the gap injury (V). Scale bar in (B–C′; D–G; H–K; L–O; P–S) 100 μm, (H′–K′; L′–O′) 50 μm.

To perform in vivo studies of Sox2/3⁺ cells, pSox3::GFP was injected in the ependymal canal, animals were electroporated and analyzed by in vivo time lapse 2-photon microscopy before and after spinal cord transection (Fig. 5A). As a control experiment, we imaged pSox3::GFP-expressing cells in sham-operated animals. GFP+ cells showed no gross morphological changes 2, 4 and 6 days post-sham operation (dps) compared to before injury. The number of GFP+ cells remained relatively constant after the sham operation (Fig. 5D–G and T).

In contrast, spinal cord transection induced a dramatic change in the pSox3::GFP expressing cells. The number of GFP+ cells increased in more than 50% of transected animals compared to the same animals before injury (Fig. 5H–K, and U). The variability in the magnitude of the changes in GFP+ cells is likely due to the variability in the initial number of electroporated cells and because some cells die after injury. The increased number of GFP⁺ cells was mainly detected in the region proximal to the rostral stump (Fig. 5J and K). Furthermore, GFP⁺ cells were detected in the ablation gap of most of the animals analyzed (Fig. 5K,O and V).

In addition to the increase in proliferation, we also observed the morphology of cells electroporated with pSox3::GFP. These cells maintain a radial glia morphology, with a long process extending from the central canal to the meninges. Some GFP+ cells were located in the ablation gap (Fig. 5M–O) or in the stumps (Fig. 5Q–S).

These experiments demonstrate directly that injury induces Sox2/3+ cells to proliferate and increase the number of Sox2/3⁺ cells. The data further suggest that proliferation of Sox2/3+ cells, especially close to the injury site, results in morphological changes in Sox2/3+ cells and their migration into the ablation gap. These events may be required for spinal cord regeneration.

Activation of Sox2/3⁺ cells is necessary for spinal cord regeneration

To test the function of Sox2/3 during spinal cord regeneration we performed knockdown experiments during regenerative stages. As a first approach we used a morpholino previously shown to specifically knockdown Sox2 in the eye field (van Raay et al., 2005). This morpholino has 5-mismatches with sox3 cDNA (see Supplementary Fig. S5), suggesting that the effects observed with this morpholino are specific to Sox 2 (Eisen and Smith, 2008).

We evaluated Sox2 knockdown using immunofluorescence in tissue sections and by western blot. Lissamine-labeled morpholinos were injected in the central canal of the spinal cord and electroporated into the ependymal layer. The spinal cord was transected after electroporation. Animals electroporated with the control morpholino (Co_MO) and the Sox2_MO were screened at 1 dpt to identify animals with similar electroporation efficiency in the ventricular layer based on lissamine labeling. We detected no difference in Sox2/3 protein levels on 1 dpt (Fig. 6A and B, see arrows A′-A‴ and B′-B‴), but by 2 dpt animals electroporated with the Sox2_MO had fewer cells labeled with lissamine and they had no detectable Sox2/3 protein. Furthermore, the number of Sox2/3⁺ cells was reduced (Fig. 6D-D‴). Importantly, at 2 dpt cells electroporated with Co_MO maintained the same levels of Sox2/3 protein as neighboring non-electroporated cells (Fig. 6C, see arrows C′-C‴). Western blots showed a clear reduction of Sox2 protein in animals electroporated with Sox2_MO compared to those with Co_MO (Fig. 6E), confirming that the morpholinos efficiently reduced Sox2 protein and the number of Sox2⁺ cells in stage 50 tadpoles.

Fig. 6. Sox2 is required for functional recovery after spinal cord injury.

(A–D‴) Tadpoles at stage 50 were electroporated with control lissamine-tagged morpholino (Co_MO, red) and Sox2 lissamine-tagged morpholino (Sox2_MO, red), transected, and analyzed by immunofluorescence against Sox2/3 (green) in longitudinal sections at 1 (A–B″) and 2 (C–D‴) dpt. (A′,A″,A‴), (B′,B″,B‴), (C′,C″,C‴) and (D′,D″,D‴) magnification of insets showed in A,B,C,D respectively. Arrows: co-localization of morpholino with Sox2/3 antibody. (E) Western blot of Sox2/3 in Co_MO and Sox2_MO. (F) Graph of the total distance swam by tadpoles electroporated with Co_MO (black circles) and Sox2_MO (red circles). (G–H) Immunofluorescence of acetylated tubulin in spinal cord longitudinal sections of electroporated Co_MO tadpoles (G) and Sox2_MO (H) after 15 days post electroporation and transection. Morpholino (red) acetylated tubulin (green) and nuclei (blue). Pictures are representative of the results observed in two independent experiments. (I) Graph of the total distance swam by transgenic tadpoles. All animals received a heat shock one day before transection and at 1 and 2 dpt. Red circles: animals that overexpress dnSox2 (EGFP⁺), black circles transgenic animals that do not overexpress the transgene (EGFP⁻) and empty circles correspond to control animals. m/5min: meters in 5 minutes. Scale bar in (A–D) 100 μm, (A′-A″; B′-B″; C′-C″; D′-D″; G–H) 20 μm. *P <0.05, **P <0.01, and ***P <0.001.

To determine the effect of Sox2 knockdown in functional recovery following spinal cord transfection we measured swimming ability with video recordings. Animals electroporated with the Co_MO showed full recovery 2–3 weeks after transection (Fig. 6F, black circles) similar to wild type animals (Fig. 1K). Importantly, tadpoles electroporated with Sox2_MO failed to show significant recovery of their swimming capability at any of the time points measured (Fig. 6F, red circles).

To evaluate the effect of Sox2 knockdown at the cellular level, animals were electroporated with Co_MO or Sox2_MO and fixed at 15 dpt. Immunolabeling for acetylated tubulin indicates that axons crossed through the injury site in animals treated with the Co_MO (Fig. 6G), but not in Sox2 knockdown animals (Fig. 6H). Together these results indicate that Sox2 is required for both functional recovery from injury and morphological recovery of axon trajectories after injury.

To study the function of Sox2 using a different approach we prepared a transgenic line expressing a dominant negative Sox2 (dnSox2) construct that is under the control of a heat shock promoter and requires dexamethasone to translocate to the nucleus. This construct is known to disrupt neurogenesis during early development and tail regeneration (Kishi et al., 2000, Gaete et al., 2012). F1 founders carrying the transgene were crossed to wild type animals, tadpoles were raised to stage 50, and heat-shocked one day before transection and at 1 and 2 dpt. Swimming ability and axonal tracts were measured as above. We found that animals expressing the transgene (EGFP⁺ animals) have a delayed functional recovery and axonal growth compared to animals that do not express the transgene (Fig. 6I and see Supplementary Fig. S6A, B). Control animals that received heat shock and dexamethasone showed full recovery of swimming (Fig. 6I, empty circles), ruling out the possibility that the effect observed was because of this treatment.

To determine the possible contribution of Sox3 to spinal cord regeneration we designed a second morpholino against sox2 (Sox2a_MO) with 10 mismatches compared to sox3 mRNA and a Sox3 morpholino (Sox3_MO) that has 9 mismatches compared to sox2 mRNA (see Supplementary Fig. S5). Electroporation of each morpholino alone or a mix of both has no significant effect on Sox2 protein levels and in swimming recovery (see Supplementary Fig. S7 and data not shown). We conclude from these results that Sox2 is required for both functional recovery from injury and morphological recovery of axon trajectories after injury, but we were unable to demonstrate or discard a possible role for Sox3 in these processes.

Neurogenesis occurs during spinal cord regeneration in Xenopus

The formation of new neurons in response to SCI (tail amputation and transection) has been demonstrated in zebrafish (Reimer et al., 2008; Briona and Dorsky, 2014) and urodele amphibians (Benraiss et al., 1999; Tanaka and Ferretti, 2009). To test whether neurogenesis increases in response to SCI in X. laevis, we evaluated the expression of several neurogenesis markers, including Doublecortin (Dcx), neuro D and neurogenin. Dcx has been used extensively to study neurogenesis (Xiong et al., 2008; Cai et al., 2009; Kutsuna et al., 2012; Klempin et al., 2011; Fan et al., 2014), it is expressed in neuroblasts and neuronal precursors during development of the central nervous system (des Portes et al., 1998; Francis et al., 1999; Gleeson et al., 1999; Meyer et al., 2002) and in the sub-ventricular zone and the subgranular zone of the hippocampus in the adult mammalian brain (Nacher et al., 2001). Neurogenin and neuroD are transcription factors of the neural basic helix loop helix (bHLH) class (Bertrand et al., 2002) that are essential for neuronal differentiation during development in the vertebrate embryo (Koyano-Nakagawa et al., 1999; Miyata et al., 1999; Liu et al., 2000). In Xenopus they are expressed in all primary neurons, neurogenic placodes, and retina strongly (Lee et al., 1995; Schlosser and Northcutt, 2000).

Immunofluorescence analysis of Dcx showed expression in cells located in the ablation gap at 2 dpt (Fig. 7A, B). Furthermore, western blots showed elevated levels of Dcx protein at 1 dpt that continued to increase to 6 dpt (Fig. 7C). Similarly, in situ hybridization demonstrated an increase in neuroD and neurogenin2 at 2 dpt in cells outside the ependymal layer (Fig. 7D–E′).

Fig. 7. SCI induces neurogenesis in regenerative stage animals.

(A–B″) Immunofluorescence of the Doublecortin (Dcx, red) in spinal cord transversal cryosections at stage 50 in sham (A, A″) and 2 dpt (B, B″) animals. Nuclei: Hoescht (blue). (C) Western blot of Dcx in sham and 1, 2, 6 dpt tadpoles (upper panel) and Tubulin (bottom panel) as loading control. (D–E′) In situ hybridization of neuroD and neurogenin2 in 2 dps (D,E) and 2 dpt (D′–E′) tadpoles. (F) Dynamic expression of neurogenin3 for RT-qPCR in tadpoles (R) and froglets (NR) in 1, 2 and 6 dpt. (G–L′) Z-stack of in vivo imaging of Sox3::GFP electroporated cells after spinal injury showing early differentiation to neuronal morphology in: animal I at (G) 2 dpe, (H) 2 dpt and (I) 4 dpt, and animal II at (J) 2 dpe, (K) 3 dpt and (L) 5 dpt. Magnifications of (G–I) and (J–L) are show in (G′–I′) and (J′–L′) respectively. vz: ventricular zone. Scale bar in (A–B″) 20 μm (G–L) 100 μm, (G′–L′) 50 μm.

Recently we reported a transcriptome-wide profile comparing the response to SCI between R- and NR-stages. We found that genes related to neurogenesis were differentially regulated between R- and NR-stages (Lee-Liu et al., 2014), strongly implicating neurogenesis as a mechanism that allows regeneration in stage 50 tadpoles. Here we further characterized the response of neurogenin3, one of the genes identified. neurogenin3 mRNA was transiently up-regulated at 2 dpt in R-stages but was not present or up-regulated in NR-stage animals (Fig. 7F) suggesting that injury-induced neurogenesis only occurs in regenerative stage.

The results depicted above suggested that neurogenesis is activated in response to injury. To test whether Sox2/3 cells contribute to injury-induced neurogenesis, we took advantage of the fact that the long half-life of GFP allows cell-fate studies. pSox3::GFP was electroporated into the ventricular layer of the spinal cord. After spinal cord transfection, we used in vivo time-lapse 2 photon imaging to determine the fates of GFP+ radial glial progenitor cells (Bestman et al., 2012). We found that GFP⁺ cells acquired neuronal morphology after transection including a long axon and complex dendritic arbor (Fig. 7G–L and see Supplementary Fig. S8). These results provide direct evidence that Sox2/3⁺ ventricular layer cells generate neuronal progeny in response to injury and demonstrate that neurogenesis is one of the mechanisms contributing to spinal cord regeneration.

DISCUSSION

Many vertebrates including rodents and the chick are able to regenerate their spinal cord at early stages of development, but this regenerative capacity is lost at later developmental stages. One strategy to understand the mechanisms that permit or restrict spinal cord regeneration would be to study these species at early and later developmental stages. Here we present an alternate strategy of using X. laevis, an experimental animal that has both regenerative and non-regenerative stages, but also has the advantage that its external development allows ready access to examine and manipulate the spinal cord throughout development. We provide a detailed histological analysis of spinal cord regeneration showing that Xenopus laevis larvae lose their regenerative capabilities concomitant with the progress of metamorphosis. We complement the histological studies with a simple quantitative locomotor test to measure recovery after SCI. We demonstrate that Sox2/3+ radial glial progenitor cells proliferate and generate neurons in response to injury. Finally, we use knockdown and dominant negative strategies to demonstrate that Sox2 is required for injury-induced neurogenesis.

This study advances the use of Xenopus as model system to study the cellular and molecular mechanisms of spinal cord regeneration by demonstrating the application of a variety of experimental strategies. We had previously demonstrated that Xenopus can be used as a genetic model organism to study regeneration (Lee-Liu et al., 2014) by performing a transcriptome-wide analysis of spinal cord regeneration in R- and NR-stages. Results of that study show that useful transcriptomic information can be obtained by comparing R- and NR-stages. To further demonstrate that Xenopus is amenable as a model system to study spinal cord regeneration, we devised a functional recovery test that evaluates regeneration through locomotor function (Guo et al., 2011). The differences in locomotor recovery shown by R- and NR-stages indicate that this test provides a simple method to evaluate, for example, gene function and the effects of modifying gene expression on spinal cord regeneration. We complement the behavioral analysis with classical histological methods, genetic manipulations, and state of the art in vivo time lapse imaging for analysis of neurogenesis and cell fate. We anticipate that application of such a multidisciplinary approach to study spinal cord injury in Xenopus will continue to provide mechanistic insights into mechanisms of recovery from injury.

Characterization of Sox2/3 cells during spinal cord regeneration

Here we analyzed Sox2/3 immunoreactive cells in Xenopus tadpole tail regeneration and in spinal cord regeneration, using an antibody that recognizes a highly conserved region of Sox2 and Sox3. We were therefore unable to distinguish between these two Soxb1 proteins. We found that Sox2/3 at stages 50–54 was expressed in most cells in the ventricular zone lining the central canal. These cells had large elongated nuclei, and were massively activated to proliferate in response to injury. Sox2/3 immunoreactivity was strongly reduced in non-regenerative animals. At these stages, the Sox2/3⁺ cells had small and round nuclei and their proliferation was slowly and poorly induced in response to injury. Cytoplasmic factors determine nuclear size to a greater extent than DNA content (Walters et al., 2012, Levy and Heald, 2010). Therefore the larger size of the stage 50 nuclei could indicate the presence of more lax chromatin and a greater tendency for cell division compared to cells in stage 66 froglets, where the nuclei are smaller and pyknotic. Our observations in froglets are very much like those reported in rodent and primate spinal cords, where Sox2⁺ cells are present in the adult spinal cord, but are poorly activated in response to SCI (Bylund et al., 2003; Sandberg et al., 2005; Mizuseki et al., 1998; Rogers et al., 2009; Mothe et al., 2011). Intriguingly, we found that the levels of Sox2 and Sox3 mRNA do not decrease during metamorphosis suggesting that the protein levels are regulated through a post-transcriptional mechanism. In line with this, recent publications suggest a crucial role of miRNAs in the regulation of Sox2 protein levels (Peng et al., 2012; Du et al., 2013; Georgi and Reh, 2010).

Sox2/3 expressing cells are necessary for spinal cord regeneration

Our previous work demonstrating that Sox2 is required for tail regeneration in Xenopus laevis tadpoles (Gaete et al., 2012), suggested that it may play a role in spinal cord regeneration. Here we show that Sox2/3 immunoreactive cells proliferate in response to injury in regenerative stages, but not in non-regenerative stages. Furthermore our in vivo time lapse imaging demonstrates that Sox2/3+ cells at the injury site proliferate and generate neurons. Our loss-of-function studies, in which we decrease Sox2 function using morpholino antisense oligonucleotides or transgenic animals expressing dominant negative Sox2, indicate that Sox2 is required for recovery of spinal cord structure and function after injury. This may occur because radial glial progenitor cells are unable to proliferate in the absence of Sox 2. Studies, on neural development in non-mammalian vertebrates, where the loss of any single SoxB1 resulted in minor phenotypes (Bylund et al., 2003, Rogers et al., 2009; Dee et al., 2008; Fei, et al., 2014), suggest functional redundancy of SoxB1 family members. Although with the current available technical resources we are unable to rule out the potential contribution of other SoxB1 family members, such as Sox3, to spinal cord regeneration. Despite these caveats, our data demonstrate that Sox2/3 expressing cells are key players in spinal cord regeneration and that decreasing Sox2 function leads to loss in regenerative capability.

Sox2/3 expressing cells: mechanisms of spinal cord regeneration

Our data suggest two potential mechanisms by which Sox2/3 may be required for SC regeneration. One is that Sox 2 plays a role in neurogenesis. This is supported by the observations that Sox2/3+ cells proliferate in response to injury and the correlated decrease in Sox2/3+ cells in non-degenerative stages. In addition, neurogenesis markers were up-regulated in response to injury, and Sox3-GFP expressing cells acquired a neuronal morphology, suggest that the formation of new neurons from activated Sox2/3⁺ cells is a possible mechanism to restore the neuronal circuits destroyed by injury. Ependymal cells in mouse spinal cords are activated in response to injury, but all are fated to glia and oligodendrocytes (Meletis et al., 2008; Barnabé-Heider and Frisen, 2010). This result is consistent with the lost of regenerative capacity in froglets, and confirms our observation that all ependymal cells are not equally capable of injury-induced neurogenesis. Of note, neurospheres isolated from injured spinal cords make neurons when transplanted in neurogenic environments (i.e. dentate gyrus), but not when transplanted in non-neurogenic tissue such as the spinal cord (Shihabuddin et al., 2000; Horner et al., 2000; Horky et al., 2006). Studying the fate of Sox2/3 cells in regenerative and non-regenerative stages could allow a better understanding on how these cells are fated to neurons or glia and identifying the intrinsic and extrinsic factors modulating these options.

Another possible mechanism for Sox2/3 cells is the generation of a permissive environment for axon regeneration. In newts, the migration and subsequent repopulation of the ablation gap is important for regeneration after SCI (Zukor et al., 2011). Moreover, ependymal cells migrate to the injury site during regeneration, and are thought to provide a permissive environment for axonal growth (Michel and Reier, 1979; Goldshmit et al., 2012). We show that regenerative stage 50 and 54 tadpoles seal the rostral stump before the caudal stump at the second day after injury, and that at 6 and 10 days after injury ependymal-like cells invade the ablation gap, followed by formation of a neural bridge and reestablishment of the rostrocaudal continuity of the spinal cord. By contrast, in stages 58 and 66, the injury stump does not close, fibrillar tissue forms in the ablation gap and ependymal cells are absent from the ablation gap. Our data showing that Sox2/3⁺ cells populate the ablation gap in regenerative stages suggests that a mechanism whereby these cells provide a permissive substrate for axon regeneration is possible. Analysis by electron microscopy has shown that axons growth through the ablation gap in direct contact with prolongations from ependymal cell (Michel and Reier, 1979). In addition, such role has been described for Sox2 in peripheral nerve regeneration giving further support to this possibility (Parrinello et al., 2010).

CONCLUSION

We demonstrate that X. laevis is a very useful model organism to study spinal cord regeneration and use the comparison between R- and NR-stages to study at the molecular, cellular and behavioral levels how both stages respond to SCI. We expect that this approach will aid in identifying the mechanisms that allow spinal cord regeneration in frogs and explain why this capacity is lost in mammals. To further develop X. laevis as a model organism we applied methods to study gene-function (i.e. morpholino electroporation, transgenic lines) and in vivo microscopy. We envision that comparing regenerative and non-regenerative stages of spinal cord injury in Xenopus laevis will provide novel insights into understanding why restoration of function is so inefficient and fails in mammals and humans.

Highlights.

• Xenopus laevis has regenerative and non-regenerative stages.

• Sox2/3 expressing cells respond to spinal cord injury.

• Sox2 is necessary for spinal cord regeneration.

• Neurogenic markers increase in response to injury in regenerative stages.