We report the first transcriptome-wide study that compares the response to spinal cord injury in Xenopus regenerative (R-) and non-regenerative (NR-) stages. While previous transcriptome studies in spinal cord injury models have been reported, they have only been performed in either mammals with very limited regenerative capabilities[36–43] or in models that regenerate throughout their lifespan, such as the zebrafish and urodele amphibians[44–47], and have used microarrays. Our study, presented here, uses high-throughput sequencing, and provides a unique experimental paradigm, whereby differences in the response to spinal cord injury between these two stages can be identified, which could then explain the difference in regenerative ability.
Early morphological differences in the response to spinal cord injury between regenerative and non-regenerative stages
We first performed immunofluorescence assays for a comparative characterization of the response to spinal cord transection of these two stages. The presence of R- and NR-stages during Xenopus development has previously been characterized. Sims described in 1962 that stage 56 was the latest stage at which animals could survive spinal cord transection. By 1990, Beattie and co-workers had been able to characterize the response to spinal cord transection histologically and using subjective functional recovery observations in stage 49 to stage 62 animals. We have previously reported a decrease in functional recovery after spinal cord transection in animals from stages 50, 54, 58 and 66, observing a progressive decrease in regenerative ability as metamorphosis proceeded, using a qualitative method to evaluate recovery.
Here, we present a comparative analysis of axonal growth after spinal cord transection in stage 50 (R-stage) and stage 66 (NR-stage) using immunodetection of acetylated tubulin. We observed the first differences at the axonal growth level within the first 2 days after transection, where axons wrapped around the stumps in the R-stage but not in the NR-stage. By 6 days after transection, axons started to extend their tips into the lesion site, as previously described for the newt. R-stage animals, therefore, show a response akin to newts, which can regenerate throughout their lifespan.
Therefore, the first histological differences between R- and NR-stages in their response to spinal cord transection were observed within the first few days after the injury. These early differences allowed us to select experimental time points for the transcriptome-wide profiling.
The transcriptome deployed in response to spinal cord injury shows extensive differences between regenerative and non-regenerative stages
According to the early differences observed in the response to spinal cord injury in R- and NR-stages, we selected 1, 2, and 6 dpt for high-throughput RNA sequencing. Our experimental design allowed us to perform pairwise comparisons not only between sham and transected animals, but also between R- and NR-stages, thus allowing us to identify: (1) transcripts that responded to damage to the spinal cord only, leaving out damage to other tissues, and (2) those that showed a different response to injury when comparing these stages. The latter group is the key advantage to our experimental model, as it represents the differences in the response to injury between R- and NR-stages, and could therefore explain their differences in regenerative capability.A global picture of the results obtained revealed that the transcriptome in response to spinal cord transection displayed the following key differences when comparing the R-stage with the NR-stage. First, the number of differentially expressed transcripts was higher at 1 dpt in the R-stage, decreasing progressively towards 6 dpt, while the NR-stage showed the highest number of differentially expressed transcripts at 6 dpt (Figure 2b). Second, out of all differentially expressed transcripts detected in both stages, only 19% were regulated in both stages, while the remaining 81% were regulated in either stage (Figure 3a). Third, genes involved in neurogenesis and axonal regeneration, both categories directly related to spinal cord injury and regeneration, showed very different expression profiles when comparing R- and NR-stages. Finally, gene ontology analysis showed that most enriched biological processes regulated in each stage were either unique to each stage, or showed a different timing in their enrichment or the length of it (Figures 5,6,7). In addition, transcripts from categories regulated in both stages were not the same in R- and NR-stages, and we were able to validate their expression changes in biological replicates using RT-qPCR, further supporting the robustness of our results. Therefore, at a global level, we observed important differences in the timing of the transcriptional response, and in the repertoire of regulated transcripts and biological processes after spinal cord injury when comparing R- and NR-stages.
Differential regulation of transcripts from gene ontologies directly related to spinal cord regeneration
As mentioned previously, key differences between the responses to injury in R- and NR-stages included the gene ontologies ‘neurogenesis’ and ‘axonal growth cone’, predicted to be related to spinal cord regeneration. The cellular and molecular differences between regenerative and non-regenerative organisms, which allow amphibians and teleost fish to regenerate, or those inhibiting regeneration in mammals, remain unknown. However, neurogenesis and axonal regeneration have been proposed to contribute to the regenerative process after a massive loss of neurons and glia due to injury, in addition to the interruption of axonal tracts[6, 7].
Constitutive neurogenesis occurs in the spinal cord of Xenopus R-stages. After spinal cord hemisection in stage 56 animals, there is regeneration of supraspinal axonal tracts, including reticular nuclei. In mammals, however, although ependymal cells have been shown to have neural stem or progenitor activity after injury, they only give rise to astrocytes and oligodendrocytes, but not to new neurons. There is no evidence for axonal regeneration in the adult mammalian central nervous system.
We found that in the ‘neurogenesis’ category, transcripts for neurod4, MGC83023, and ascl1 were exclusively upregulated in the R-stage (Figure 4a). Neurod4 (Neurogenic differentiation 4) and ascl1 (Achaete-scute complex homolog 1) are both proneural transcription factors[52, 53], and while MGC83023 has not been characterized previously, it has a 97.3% sequence identity to the Xenopus tropicalis transcript for achaete-scute homolog 1-like. This specific upregulation in the R-stage of proneural transcription factors suggests a differential regulation of the neurogenic process during regeneration. Nevertheless, although we expected a similar specific upregulation of more transcripts from this category, a group of them showed an early upregulation in the R-stage at 1 and 2 dpt, and a delayed upregulation in the NR-stage at 6 dpt (Figure 4, II). Upregulation of these transcripts supports that neurogenesis could be taking place after injury in the R-stage, and that timely activation of the neurogenic program is required for successful spinal cord regeneration.
Another gene ontology category regulated differentially included transcripts that belong to the axonal growth cone group. More than half of these transcripts showed a strong downregulation at 6 dpt in the NR-stage, which was not observed in the R-stage. We found several factors that could explain failure of axonal growth cone extension. For example, external cues like semaphorins and reelin, and receptors that respond to extracellular cues like EPH receptors, ngfr(p75) and the netrin-1 receptor dcc (Figure 4, III, and Table 4). The latter has been shown to mediate a turning response in retinal ganglion cell growth cones in Xenopus. Conversely, the intracellular response machinery to extracellular cues was also altered, including downregulation of the cdk5r1 activator of CDK5 and microtubule-associated proteins. All of these molecules were mainly deregulated in the NR-stage and are in agreement with the degeneration of the distal part of severed axons and the lack of neurite plasticity events, such as sprouting or axonal regeneration, which could, in part, explain the lack of functional recovery in NR-stage animals.
Therefore, transcripts related to neurogenesis and axonal regeneration were differentially regulated, with specific upregulation of three proneural transcription factors in the R-stage and considerable downregulation of growth cone transcripts and deregulation of axonal guidance cues in the NR-stage. Furthermore, the fact that these known processes were regulated differentially in R- and NR-stages supports this comparative experimental paradigm in Xenopus laevis as a model to identify the molecular mechanisms that allow regeneration in the R-stage or inhibit it in the NR-stage.
Different biological processes were regulated in regenerative and non-regenerative stages
Another key difference in the response to spinal cord transection was observed at a global level in the gene ontology enrichment analysis. Differentially expressed transcripts in R- and NR-stages had different global profiles of enriched processes, and they could be arranged into two main groups: (1) processes related to stem or progenitor cell maintenance and differentiation, and (2) processes involved in providing an either permissive or non-permissive environment for regeneration.
In the first group, processes related to stem or progenitor cell maintenance and differentiation, we identified the following biological processes: metabolic processes, cell cycle and developmental processes. We found that a remarkable predominance of metabolic processes were enriched in the R-stage (Figures 5,6; purple), and seven out of ten top regulated transcripts belonged to genes involved in metabolism. This suggests a high regulation of metabolic processes after spinal cord injury in the R-stage. Recently, there have been several reports on stem cell metabolism and the role this plays in pluripotency maintenance[55–58]. These associate a highly glycolytic metabolism with ‘stemness’ and cell proliferation, whereas the switch towards oxidative metabolism causes a shift towards differentiation. Furthermore, recent work from the Daley group showed that Lin28 enhances tissue repair through reprogramming of cellular metabolism in different injury models in mice. The predominance in our results of transcripts related to metabolic processes regulated after injury in the R-stage supports the notion of metabolism as a key regulator of endogenous stem cells and their capacity to differentiate during neurogenesis.In addition, enrichment of upregulated cell cycle genes at 1 and 2 dpt in the R-stage (Figure 5, blue) and a concomitant transient (only 1 and 2 dpt) downregulation of developmental genes (Figure 6, orange) suggests a stem cell proliferation phase followed by a differentiation phase in which cell proliferation ceases, developmental processes are no longer repressed anymore, allowing differentiation.
Processes involved in promoting a permissive or non-permissive environment for regeneration were the immune response and inflammation, oxidation and reduction, and response to stress. The main differences between R- and NR-stages were that amongst upregulated transcripts, immune response and inflammation was only enriched in the NR-stage (Figure 5, green); amongst downregulated transcripts oxidation and reduction was only enriched in the R-stage (Figure 6, yellow); while the response to stress was enriched transiently (1 and 2 dpt) in the R-stage (Figure 5, pink), but enriched at all time points in the NR-stage (Figures 5,6, pink). It has been proposed that the mammalian spinal cord provides a non-permissive environment for neurogenesis because neural stem and progenitor cells (NSPCs) are present (ependymal cells), but they only give rise to glial cells, not to new neurons. However, when spinal cord NSPCs isolated by stimulation with FGF2 in mammals are transplanted into a neurogenic niche in the brain (for example, dentate gyrus), they give rise to new neurons. This raises the question of which are the extrinsic cues that are permissive for neuronal differentiation, and which are non-permissive, like those present in the mammalian spinal cord. Regarding this, while both positive and negative effects have been associated with the immune and inflammatory responses after spinal cord injury in mammals, regenerative models have been shown to have a less developed or more controlled response[60–63].
The immune response is also associated with oxidation-reduction processes, as leukocytes can have iNOS activity and therefore be involved in the production of reactive oxygen species. A higher infiltration of iNOS positive leukocytes in metamorphic stage Xenopus tadpoles (described as non-regenerative in this study) has been correlated to the lack of tail regeneration. The enrichment pattern for immune response and inflammation and for oxidation-reduction processes in our results support this evidence. Finally, a sustained response to stress could also be associated with the maintenance of a non-regenerative permissive environment.
While neurogenesis and NSPC activity, and the presence of either permissive or non-permissive environments had been proposed to be key factors during spinal cord regeneration, our results shed light on which factors could be regulating these processes. The results obtained here not only suggest metabolic, developmental and proliferation processes as differentially regulated in response to injury, but also provide an integrative view of how they could interact during spinal cord regeneration, and which genes in particular are being regulated. The same applies to the role of the immune response and inflammation, response to stress and oxidation-reduction and their effect as extrinsic factors affecting regeneration. This is valuable information as it is this knowledge of how processes predicted to contribute to spinal cord regeneration can be modulated that will finally provide new strategies to promote spinal cord regeneration in mammals.