Model systems for regeneration: Xenopus

The goal of regenerative medicine is to identify novel therapies aimed at stimulating a full regenerative response in humans following injury, disease or during ageing. As such, uncovering the mechanisms employed by organisms with high regenerative capacity may lead to new therapies for humans, which generally have poor regenerative capabilities. As originally discovered by Lazzaro Spallanzani in the mid-18th century, the anuran (‘tailless’) amphibians, which include Xenopus laevis (X. laevis) and Xenopus tropicalis (X. tropicalis), are one such group of animals with high regenerative capacity (Dinsmore, 1991). Originally introduced as a model organism for endocrinology and physiology, but later adopted for cell and developmental biology (Gurdon and Hopwood, 2000), the African clawed frog (X. laevis) has been increasingly appreciated over the past 50 years for its regenerative capacity (Deuchar, 1975). Although this species has many advantages, its allotetraploid genome and long generation time make it less attractive for genetic studies. For this reason, its diploid relative, the Western clawed frog (X. tropicalis) was introduced as an experimental organism in the late 1990s to mitigate these disadvantages (Amaya et al., 1998).

In this Primer, we give an overview of the life cycle, evolutionary position and regenerative capacity of X. laevis and X. tropicalis. We then summarise the tools and techniques that are available to study regeneration in these species and review how studies of Xenopus have informed our current understanding of the cellular and molecular mechanisms of regeneration. Finally, we discuss the major benefits and limitations of using Xenopus as a model system to understand regeneration.

Although X. laevis and X. tropicalis diverged between 30 million and 90 million years ago (Evans et al., 2004; Session et al., 2016), there are many similarities between the two species (Fig. 1). Both follow the same morphological changes during development and can be staged using the same developmental staging system (Nieuwkoop and Faber, 1994; Zahn et al., 2017). Embryos can be generated in large numbers following a simple hormone injection and the rate of Xenopus development can be manipulated by modifying temperature, which greatly facilitates experimental manipulation during early development. It was this ease of generation of embryos that led to their adoption as models for developmental biology (Amaya et al., 1998; Gurdon and Hopwood, 2000). Although X. tropicalis embryos proceed slightly faster through gastrulation and neurulation than X. laevis embryos, and optimal housing temperatures differ between the two species (Ishibashi et al., 2017; Khokha et al., 2002), the timing and levels of orthologous gene expression are remarkably similar at comparable developmental stages, particularly after gastrulation (Yanai et al., 2011). Both species reach tadpole stages by 3-4 days post-fertilisation and metamorphosis within 2-3 months, although X. tropicalis commonly reaches sexual maturity quicker (Ishibashi et al., 2017) and is therefore the favoured species for genetic studies (Amaya et al., 1998; Tandon et al., 2017). Importantly, both species are supported by advanced genomic resources, including annotated genomes (Amaya, 2005; Hellsten et al., 2010; Session et al., 2016), and genetic approaches, such as transgenesis and gene-editing tools (Blitz et al., 2013; Hartley et al., 2002; Ishibashi et al., 2012; Kroll and Amaya, 1996; Love et al., 2011b; Ogino et al., 2006; Tandon et al., 2017).

Xenopus species occupy a phylogenetic position between urodele amphibians and amniotes and in concordance with this position their regenerative capacity is intermediate between the two (Fig. 2). Whereas urodeles display life-long regenerative capacity of many organs and appendages (Joven et al., 2019), amniotes and mammals generally display only embryonic or neonatally restricted regeneration, particularly of the central nervous system (CNS) (Bernstein and Stelzner, 1983; Boulland et al., 2013; Chen et al., 1997; Gaillard et al., 2007; Karl and Reh, 2010) and heart (Price et al., 2019). Notably, amniotes fail to regenerate their limbs or tailbuds even during embryogenesis (Schoenwolf, 1977, 1978; Summerbell, 1974). In contrast, Xenopus has high regenerative capacity in the pre-metamorphic larval stages, but this capacity is largely lost following metamorphosis, with the exception of a few tissues, such as the retina (see Table 1). Thus, X. laevis and X. tropicalis are in a useful position to understand the evolutionary transition between organisms with life-long regenerative capacity and those with limited regenerative capacity (a question of phylogeny) and to investigate the mechanisms that underlie the changes from regenerative capacity to non-regenerative capacity during the lifetime of an organism (a question of ontogeny). For brevity, hereafter we focus on regeneration in the context of tissues for which recent findings have led to new insights, namely the appendages (tail/limb), CNS and heart. Where similar regenerative processes have been reported in both X. laevis and X. tropicalis, we will refer to these species collectively as ‘Xenopus’.

Xenopus tadpoles can regenerate their tails, including all of the associated tissues, such as the spinal cord, major blood vessels, muscles and fin (Fig. 3A), from the early tailbud stages (Deuchar, 1975) through to pre-metamorphosis [Nieuwkoop and Faber stage (NF) 53] (Fukazawa et al., 2009; Gaete et al., 2012; Love et al., 2011a; Nieuwkoop and Faber, 1994). This ability is transiently lost at the start of feeding in X. laevis (the ‘refractory period’; NF45-47) (Beck et al., 2003). Regeneration encompasses three broad phases (Fig. 3B): (1) an initial wound closure and inflammatory response phase at 0-1 days post-amputation (dpa); (2) the formation of a blastema-like regenerative bud at 1-2 dpa; and (3) an outgrowth and overt regeneration phase, starting from 2-3 dpa (Love et al., 2011a). Although the newly formed tail does not regain the entirety of its original tissue patterning, and exhibits differences in muscle, vasculature and axonal organisation, the regenerated tail is fully functional by 7 dpa (Deuchar, 1975; Love et al., 2011a).

A fundamental question in regenerative biology is what drives the process of regeneration? When addressing this question, attention has primarily focused on the ‘blastema’ – a group of mesenchymal and proliferative cells located at the tip of regenerating tissues. At the cellular level, the Xenopus tail blastema was initially postulated to contain de- or trans-differentiating cells, based on early experiments in urodeles (Brockes and Kumar, 2002; Echeverri and Tanaka, 2002; Okada, 1980). However, grafting experiments in the X. laevis tail showed that the regenerating muscle, spinal cord and notochord are derived from their respective tissue lineages without de- or trans-differentiation (Gargioli and Slack, 2004; Ryffel et al., 2003). Thus, for a time, the regenerative structure was termed a ‘regeneration bud’ to distinguish it from groups of de-differentiated blastema cells containing pluripotent stem or progenitor cells (Gargioli and Slack, 2004; Beck et al., 2006; Love et al., 2011a). Subsequently, the presence of lineage-restricted progenitors was also found during limb regeneration in axolotls (Kragl et al., 2009) and digit regeneration in mice (Lehoczky et al., 2011; Rinkevich et al., 2011), suggesting that the existence of lineage-restricted progenitors may be a general property of regenerative tissues. In the X. laevis tail, the regeneration bud is now known to contain a number of lineage-restricted progenitors, such as Sox2/3⁺ neural progenitor cells (NPCs) (Gaete et al., 2012) and Pax7⁺ satellite cells (Chen et al., 2006).

Where studies of Xenopus appendage regeneration have been particularly instrumental is in promoting our understanding of the molecular mechanisms required to initiate a regenerative programme. Early assessment of drivers of regeneration focused on known developmental signalling pathways. These studies, alongside more recent genome-wide approaches, confirmed that developmental pathways, such as the Notch, bone morphogenetic protein (BMP), Wnt, fibroblast growth factor (FGF), transforming growth factor β (TGFβ) and sonic hedgehog (Shh) pathways, become reactivated during tail regeneration (Aztekin et al., 2019; Beck et al., 2003, 2006; Ho and Whitman, 2008; Lin and Slack, 2008; Lin et al., 2012; Taniguchi et al., 2014). Further studies have also highlighted the importance of additional cellular mechanisms and factors in tail regeneration, such as apoptosis (Tseng et al., 2007) and the extracellular matrix (ECM) (Contreras et al., 2009). Additionally, a subpopulation of epidermal cells, named ‘regeneration-organising cells’ (ROCs), have been reported to migrate to the amputation edge and release signals that promote the proliferation of the axial tissues (Aztekin et al., 2019). Using genome-wide approaches to assess changes in the transcriptome during the different phases of tail regeneration in X. tropicalis, two previously unappreciated processes unfolded (Love et al., 2011a). One was that the expression of many genes associated with metabolism change during regeneration, suggesting a potential role for metabolic reprogramming during regeneration (Love et al., 2014). In addition, genes implicated in the production of reactive oxygen species (ROS) were upregulated during regeneration. A follow-up study revealed that tail regeneration is associated with a sustained increase in ROS levels and that this is required for Xenopus tail regeneration to proceed (Love et al., 2013). It has since become apparent that sustained ROS production is a common feature of tissue and appendage regeneration in both invertebrates and vertebrates (Gauron et al., 2013; Khan et al., 2017; Pirotte et al., 2015; Zhang et al., 2016), suggesting that a role for ROS in regeneration is ancient and conserved. The current challenges are to ascertain how ROS levels are regulated during regeneration and to identify the downstream targets of ROS production that promote cell proliferation, growth factor signalling and metabolic reprogramming.

In addition to elevated ROS production, a requirement for bioelectrical changes has been observed following tail amputation in X. laevis. This includes the repolarisation of cellular membrane potential (Vm) (Adams et al., 2007; Tseng et al., 2010), the transient activation of a voltage-gated sodium channel, Na_V1.2 (Tseng et al., 2010), and the reversal of electric current densities (J_I) driven by changes in trans-epithelial potential (TEP) (Ferreira et al., 2016). In addition, depletion of ROS in regeneration-competent animals using the NADPH oxidase (Nox) inhibitor DPI impairs TEP and J_I dynamics, suggesting that bioelectrical signals are downstream of ROS (Ferreira et al., 2018). An indirect link between extracellular oxygen (O₂) influx and ROS production has also been proposed (Ferreira et al., 2018). Elevated O₂ influx correlates with closure of the wound epithelium and formation of the blastema; a similar increase is also observed following skin wounding in mice, suggesting this process is evolutionarily conserved (Ferreira et al., 2018). This process also appears to be downstream of ROS, as DPI-treated tadpoles show significant decreases in O₂ influx. It should be noted, however, that stabilization of a key mediator of hypoxia, hypoxia-inducible factor α (Hif1α), is not sufficient to rescue regeneration in DPI-treated tadpoles, even though Hif1α is itself necessary and sufficient for tail regeneration, via modulation of J_I current reversal. This suggests that both ROS and Hif1α are independently required for regeneration (Ferreira et al., 2018).

An interesting conundrum is how and why regenerative processes change with age. The ability to study regenerative and non-regenerative phases within the same species makes Xenopus a useful model to tackle this question. Although the presence of a ‘refractory period’ has not yet been documented in X. tropicalis, assessment of the refractory period in NF45-47 X. laevis tadpoles has confirmed the absence of some pro-regenerative mechanisms during the refractory period. For example, it has been shown that ROS levels, Na_v1.2 expression and TEP/J_I dynamics are impaired during the refractory stages in X. laevis tadpoles (Ferreira et al., 2018; Tseng et al., 2010), confirming the importance of ROS production and changes in bioelectrical dynamics for establishment of a full regenerative programme. At the cellular level, ROCs fail to migrate to the amputation site during the refractory stages (Aztekin et al., 2019), although the reason for this is currently unknown. Studies of the refractory period have also implicated the importance of inflammatory processes during tail regeneration. In regeneration-competent X. tropicalis tadpoles, inflammatory cells are recruited to the wound site over the first 6 h post-amputation (hpa) (Love et al., 2011a), in line with genome-wide studies in Xenopus, which demonstrated an enrichment of inflammation-associated genes during the early phase of regeneration (Aztekin et al., 2019; Chang et al., 2017; Love et al., 2011a). In the refractory period, however, inappropriate activation of the immune system may impair tail regeneration, as immunosuppression improves regeneration in these animals (Fukazawa et al., 2009). Contradicting this finding, the presence of bacteria and the activation of complexes downstream of the NF-kB pathway have recently been suggested to promote regeneration in the refractory period (Bishop and Beck, 2019 preprint). Further studies are required to fully elucidate the involvement of the immune system, but it is likely that tight regulation of the inflammatory response is required for efficient tail regeneration.

Xenopus tadpoles can regenerate their hindlimbs at the pre-metamorphic stages, especially during the early limb bud stages (NF50-52), but regenerative capacity diminishes steadily as limb development progresses (NF53 and onwards) (Keenan and Beck, 2016). In X. laevis, the number of digits formed at the completion of regeneration decreases as development proceeds until only a cartilaginous, hypomorphic spike is formed in pro- and post-metamorphic animals (Dent, 1962; Goss and Holt, 1992). Owing to its enclosed development, regeneration of the forelimb has only been studied in post-metamorphic frogs, when amputation also results in the regeneration of a hypomorphic spike (Endo et al., 2000). Xenopus therefore offers several advantages as a model system for investigating limb regeneration as one can study: (1) the successful regeneration during early hindlimb development; (2) the mechanism(s) responsible for progressive loss of regenerative capacity of the limb; and (3) the mechanisms that promote successful limb regeneration and patterning in the limb when regeneration is normally limiting.

Differential expression of genes involved in tissue patterning, inflammation, ECM composition and metabolism (Grow et al., 2006; King et al., 2003; Mescher et al., 2013), and the expression of developmental genes (e.g. fgf8, fgf10, msx1, sall4 and hoxa13) (Christen and Slack, 1997; Endo et al., 2000; Neff et al., 2011; Yokoyama et al., 2001) have all been reported during limb regeneration. Interestingly, it has been found that the developmental stage of the limb mesenchyme determines regenerative capacity far more than the developmental stage of the overlying epidermis (Yokoyama et al., 2000). Recent re-assessment of X. laevis limb development has led to the postulation that, like the chick, Xenopus has specific regions within the mesenchyme (i.e. the zone of polarizing activity; ZPA) and epidermis (i.e. the apical epithelial ridge; AER) that direct anterior-posterior patterning during development (Christen and Slack, 1997; Christen et al., 1998; Endo et al., 1997; Yokoyama, 2007; for a review, see Keenan and Beck, 2016). As development proceeds, the expression of ZPA and AER markers (shh and fgf8, respectively) decreases, correlating with the loss of regenerative capacity at ∼NF55-57 (Endo et al., 1997; Wang and Beck, 2014). Chromatin immunoprecipitation has also shown that histone modifications within the limb blastema are unchanged compared with those in the developing NF52-53 limb bud (Hayashi et al., 2015), suggesting that regenerative cells at this stage of development seem to retain memory of what they were prior to amputation. It is plausible that, as the mesenchyme differentiates into subpopulations as the limb develops (Tschumi, 1957), these cells are less able to acquire different identities following amputation, leading to inappropriate tissue patterning. However, other work has suggested distinct regeneration-specific mechanisms, including some regeneration-specific genes (grem1 and hsp70) (Pearl et al., 2008) and the differential involvement of micro-RNAs in regeneration-competent limb buds and regeneration-incompetent limbs (Zhang et al., 2018).

In the adult limb, studies have predominantly focused on assessing treatments that may promote successful regeneration or tissue patterning. This includes the modulation of bioelectricity (Herrera-Rincon et al., 2018; Tseng and Levin, 2013), the stimulation of regeneration with exogenous growth factors and forced gene expression, or cell transplantation (Lin et al., 2013; Wang et al., 2015). Although exogenous application of Fgf10 is sufficient to stimulate X. laevis hindlimb regeneration (Yokoyama et al., 2001), forced expression of Shh and Fgf10 is insufficient to enhance forelimb regeneration (Lin et al., 2013). However, regeneration of multiple digits can be stimulated by transplanting larval limb cells with activated Wnt/β-catenin signalling, when combined with a cocktail of Shh, Fgf10 and thymosin-β4 (Lin et al., 2013). Interestingly, the survival of the transplanted grafts relies on host thymectomy prior to metamorphosis, suggesting that immune suppression is required for donor cell survival. However, given that contralateral limbs − which were amputated, but not treated with the cell transplant or growth factor cocktail − fail to regenerate, it is clear that immune suppression alone is not sufficient to promote regeneration in the adult forelimb. Although complex treatment was required in this instance to stimulate regeneration, short-term modulation of bioelectricity has shown similar levels of success. For instance, chemical induction of sodium influx for 1 h is sufficient to induce regeneration of toes and toenails at 45 dpa (Tseng and Levin, 2013), and 24-h treatment with a wearable progesterone-containing bioreactor restores locomotor function to near-control levels by 7.5 months post-amputation, despite the animals failing to regenerate an appropriately patterned limb (Herrera-Rincon et al., 2018). Treated animals also display increased major blood vessel vascularization, increased innervation and re-organisation of bone patterning. This suggests that short-term treatment aimed at modulating bioelectricity is able to stimulate a long-term regenerative programme, which can promote functional regeneration, in spite of differences in morphology compared with intact appendages. Intriguingly, bioelectrical changes following injury can act over a long range, via a process termed bioelectric injury mirroring (Busse et al., 2018). It would be interesting to assess whether longer-term bioreactor treatment would induce improved patterning of the regenerate.

Overall, Xenopus appendages have provided useful mechanistic insights into the general properties of a core regenerative programme and suggest that the blastema of both the tail and limb are strikingly similar with regards to the production and pro-regenerative nature of ROS and bioelectric gradients. A recent study has also outlined a previously unknown role for melanocortin signalling in the production of ROS in the regenerating limb bud, showing that the application of melanocortin-stimulating hormone rescues regeneration in denervated limbs (Zhang et al., 2018); reduced ROS levels in denervated limbs further suggests a possible role of nerve signalling in the production of ROS in the blastema. Considering that both tail and limb regeneration are nerve dependent (Cannata et al., 2001; Filoni and Paglialunga, 1990; Suzuki et al., 2005; Taniguchi et al., 2008), it would be interesting to assess whether stimulation of the CNS is a general source of ROS production in regenerating tissues. In the future, it will be important to understand the differential regulation of signalling pathways during development and regeneration, the mechanism underlying the relationship between ROS, bioelectricity and nerve dependency, and the contribution of cellular memory of cells in the blastema to regeneration competency.

The regenerative potential of the mammalian CNS is very limited. As such, injuries to the brain or spinal cord commonly lead to devastating consequences. This can include motor, sensory and autonomic dysfunction, as well as the development of a chronic inflammatory state (Allison and Ditor, 2015; Sun et al., 2016). Research in mammalian models of CNS injury have focused on improving the ability of axons to grow through the injury site, or transplantation of exogenous stem cells, but these interventions have only shown limited success (Anderson et al., 2018; Dias et al., 2018; Jessberger, 2016). In X. laevis tadpoles, however, a crucial feature of the regenerative process in the CNS appears to be the ability to activate endogenous NPCs to proliferate and generate new neurons (Fig. 4A) (Bernardini et al., 2010; Gaete et al., 2012; Muñoz et al., 2015; Yoshino and Tochinai, 2004). Note that, although we term these cells ‘NPCs’, they are sometimes described in the literature as radial glia (RG) or ependymal cells (Chernoff et al., 2018; Edwards-Faret et al., 2018; McKeown et al., 2013).

The X. laevis telencephalon and mesencephalon can regenerate up to pre-metamorphic stages (NF47-54). Injury can be achieved by partial removal of the olfactory bulb and cerebrum (Yoshino and Tochinai, 2004) or via vacuum aspiration of neurons in the optic tectum (McKeown et al., 2013). Both injuries induce the proliferation of NPCs and the generation of newborn neurons (McKeown et al., 2013; Yoshino and Tochinai, 2004). In the mesencephalon, recovery, as measured by visual avoidance behaviour, is delayed by treatment with cell cycle inhibitors hydroxyurea and aphidicolin (McKeown et al., 2013), showing that proliferation is a key process required for functional regeneration. Newborn neurons integrate into the regenerating tectal circuit; integration and behavioural recovery is significantly enhanced at 24 h post-injury by visual stimulation (Gambrill et al., 2019), suggesting that experience-dependent plasticity can increase the rate of regeneration. Complete regeneration of the olfactory bulb following injury to the telencephalon requires reconnection to the olfactory nerve, consistent with an activity-dependent regenerative mechanism also in the telencephalon. However, the cerebrum can regenerate independently of this connection, suggesting that this may be a unique feature of sensory brain regions (Yoshino and Tochinai, 2006).

Like the brain, the Xenopus spinal cord can regenerate up to pre-metamorphosis following tail amputation (Gaete et al., 2012; Love et al., 2011a). An X. laevis transection model has also been published (Gaete et al., 2012; Muñoz et al., 2015). In X. laevis, both injuries induce the upregulation of Sox2/3 expression and the proliferation of Sox2/3-positive cells (Gaete et al., 2012; Muñoz et al., 2015). Spinal cord-specific manipulation of Sox2 expression, either by morpholino oligonucleotide knockdown or the expression of a dominant-negative form of Sox2, prevents spinal cord regeneration (Gaete et al., 2012; Muñoz et al., 2015). In non-regenerative (NR) post-metamorphic froglets, Sox2/3 expression is very low in the intact spinal cord and, although Sox2/3 expression increases upon transection, these cells do not proliferate and the spinal cord does not recover its integrity, leading to persistent paralysis (Fig. 4A) (Muñoz et al., 2015). These studies suggest that, like in the brain, the proliferation of NPCs is a key feature of successful spinal cord regeneration.

The ability to use two different injury models to study spinal cord regeneration is a key advantage of Xenopus. Whereas transection requires the bridging of two stumps, tail amputation requires the growth of multiple tissues in an organised fashion from a single wound (Fig. 4B). Studying how each of these processes is controlled enables researchers to address different questions. For example, transection allows the study of a purely spinal injury and, as this can be assessed in both regenerative (R) and NR stages, key factors that promote successful regeneration in NR stages can be elucidated. For instance, regeneration in NR stages can be improved by transplanting dissociated spinal cord cells from R, but not NR, animals (Méndez-Olivos et al., 2017). In this study, the grafted cells were first treated with a cocktail of epidermal growth factor (EGF), FGF2 and brain-derived neurotrophic factor (BDNF) to aid survival, before retention in a fibrinogen/thrombin matrix. Grafted Sox2/3⁺ cells from R animals self-organised into rosette-like structures and proliferated, and by 60 days post-transection had differentiated into neurons that projected axons into the host spinal cord. In turn, host axons were also able to grow into the graft, suggesting that the transplanted cells support a permissive environment for axonal growth. By contrast, transplanted cells from an NR host in the same paradigm are unable to survive, form rosette structures or differentiate into neurons. This suggests an intrinsic cellular difference between cells from R and NR stages, rather than a non-permissive environment in the injured NR spinal cord. This study also highlights the possibility of using transection in NR animals to assess therapeutics that can promote the regeneration of host axons.

On the other hand, tail amputation allows one to study mechanisms of nerve dependency, reparative organogenesis and tissue-to-tissue interactions. For instance, it has been shown that surgical removal of the spinal cord before amputation prevents the tail from regenerating, in part due to impaired cell proliferation in the notochord (Taniguchi et al., 2008). Conversely, it has been reported that notochord-derived Shh promotes proliferation in the spinal cord during tail regeneration (Taniguchi et al., 2014). These examples highlight the importance of communication between different tissues and cell types during epimorphic regeneration. Furthermore, the requirement for neuronal innervation is a common but poorly understood phenomenon of successful regeneration (Boilly et al., 2017).

Unlike the brain and spinal cord, the Xenopus optic nerve and retina retain regenerative capacity throughout life. Interestingly, retinal regeneration is achieved by a different cellular mechanism in X. laevis compared with X. tropicalis. In adult X. laevis, precursors of the ciliary marginal zone (CMZ) repopulate the injured retina following partial removal; however, after full removal of the retina, cells of the retinal pigmented epithelium (RPE) transdifferentiate to form the new neural retina (Yoshii et al., 2007). By contrast, in adult X. tropicalis, cells of the CMZ rather than the RPE appear to regenerate the retina following total removal (Miyake and Araki, 2014), similar to retina regeneration in zebrafish (Raymond et al., 2006). Following on from research in non-mammalian vertebrates, a population of retinal progenitors has recently been identified in the developing mouse CMZ (Bélanger et al., 2017), although whether these cells are retained in adulthood, and how they may respond to injury, is unknown.

Unlike other tissues in the CNS, optic nerve regeneration following crush is not predominantly due to the proliferation of existing cells or new neurogenesis (Beaver et al., 2001). Instead, X. laevis retinal ganglion cells (RGCs) are protected from death following axonal injury, and these cells regrow their axons to reconnect the retina to the optic tectum by 15 days post-crush (dpc) (Gaze, 1959; Zhao and Szaro, 1994). Regeneration follows two key phases: (1) inflammation and degeneration of injured axons up to 5 dpc; and (2) regeneration of RGC axons across the injury site from 5 dpc (Wilson et al., 1992; Zhao and Szaro, 1994). Translational ribosomal affinity purification has been used to assess transcriptomic changes specifically in regenerating RGCs at 1, 3, 7 and 11 dpc (Whitworth et al., 2017). Interestingly, although axonal growth-associated genes such as klf6 and gap43 are upregulated, as well as responses to endoplasmic reticulum stress, oxidative stress, hypoxia and inflammation, the majority of differentially expressed genes are down regulated, including genes involved in axonogenesis, axonal guidance and synaptic communication. This suggests that regenerating RGCs switch from a mature ‘neuronal’ gene expression profile to a programme that prioritises cell survival (Whitworth et al., 2017).

In summary, Xenopus tadpoles use a range of different mechanisms to regenerate their nervous system, from the activation of tissue resident cells in the spinal cord to transdifferentiation in the retina. Unlike mammals, the X. laevis optic nerve and retina can regenerate throughout life. The ability to regenerate the brain and spinal cord changes during the life cycle of Xenopus, from being near perfect before metamorphosis (as in axolotl or zebrafish), to being very limited after metamorphosis (as in mammals). This offers an opportunity to assess experimentally the mechanisms underlying this transition within the same species. Indeed, datasets using transcriptomic and proteomic approaches are starting to provide clues relating to the differences in response to injury in R and NR spinal cords (Gibbs et al., 2011; Lee-Liu et al., 2014, 2018), such as upregulation of metabolism, cell cycle control genes and the immune response in R, but not NR, stages. The challenge now is to identify the important players that will allow us to promote regeneration in non-regenerative animals such as mammals.

Heart disease is the leading cause of death worldwide. It is therefore understandable why so many organisms have been investigated for their cardiac regenerative capacity, in the hope of applying lessons learned from those organisms to advance regenerative approaches in humans. Although cardiac regeneration work has largely been dominated by studies in zebrafish and mice, heart regeneration has also recently been evaluated in Xenopus (Marshall et al., 2017). Heart regeneration in adult X. laevis frogs (>5 years old) is absent 1 year after endoscopy-based resection of the heart apex, which is calculated to remove approximately 4% of cardiac tissue (Marshall et al., 2017). In their study, the authors found evidence of persistent fibrosis, cardiac hypertrophy and deterioration of cardiomyocyte sarcomere structure, with no change in proliferation as evidenced by mRNA expression of cell cycle-related genes (Fig. 5A). They therefore concluded that the adult Xenopus heart responds similarly to the mammalian adult heart following injury. By contrast, a separate study (Liao et al., 2017) found that 12-month-old X. tropicalis frogs are capable of near scar-free regeneration by 30-60 days following approximately 10% apical resection of the heart (Liao et al., 2017) (Fig. 5B). There are notable differences in methodologies between these two studies, including the injury technique, the techniques used to assess proliferation, and the age of the animals, all of which might influence the observed regenerative outcomes (Liao et al., 2018; Marshall et al., 2018).

Given that Xenopus is well known to exhibit a diminishing regenerative capacity from larval to adult stages, a more detailed follow-up study assessed heart regeneration in X. laevis across different life stages, including metamorphic onset (NF57-58), climax (NF61-62) and end (NF66), as well as a few weeks after completion of metamorphosis (froglet) and in 6-month-old juvenile frogs (Marshall et al., 2019). Although cardiomyocyte proliferation was investigated, as is often the focus in heart regeneration studies, changes relating to the ECM were also studied. Specifically, it was shown that whereas R animals (injured before metamorphosis) are able to clear deposited fibronectin and collagen and successfully complete cardiac regeneration, NR animals (injured after metamorphosis) display persistent fibrotic scars and fail to complete regeneration (Marshall et al., 2019). It therefore appears that cardiac regeneration may be stage and species dependent. However, it is unknown whether pro-metamorphic X. tropicalis tadpoles display increased regenerative capacity relative to their X. laevis cousins, or what mechanisms allow heart regeneration to proceed better in adult X. tropicalis. Similar differences have been observed between the teleost fish zebrafish and medaka (Lai et al., 2017) and in surface-dwelling versus cave-living Mexican cavefish (Stockdale et al., 2018).

But what might underlie the change in regenerative capacity as metamorphosis proceeds? It is known that neonatal mice lose cardiac regenerative capacity at around 1 week after birth (Porrello et al., 2011), a time that is also associated with a peak in serum thyroid hormone (TH) levels (Hadj-Sahraoui et al., 2000). It is also notable that the regeneration of other Xenopus tissues is diminished during and after TH-regulated metamorphosis (Beck et al., 2009). This led Marshall and colleagues to ask whether TH levels or availability might be responsible for the switch between regenerative and non-regenerative states in X. laevis. They showed that both TH excess and deprivation in tadpoles impair fibrotic clearance and reduce regeneration efficiency. This was associated with increased fibrotic scar extent, a reduction in complete myocardial wall closure, altered ECM gene expression, and changes in the kinetics of tenacin-C deposition and elimination. These observations suggest that ontogenic loss of regenerative capacity is not simply due to increased TH levels, as both too little and too much TH have negative impacts on regeneration. Two plausible hypotheses for this exist: (1) that causing an animal to develop too quickly (i.e. with excess TH) or stunting its development (via TH deprivation) will have a knock-on effect on its ability to regenerate, or (2) that the exact level and/or timing of TH availability and function is key to maintaining a full regenerative response.

In summary, the regenerative capacity of its heart makes Xenopus an attractive model for heart regeneration studies. In the future, we expect Xenopus to be used more widely and in comparative studies with the mouse model. This should reveal similarities and differences with regard to permitting or inhibiting heart regeneration, findings that will no doubt provide essential information that could aid regenerative medicine approaches following heart injury in the clinic.

A major advantage of using Xenopus as a model species for regeneration comes from the fact that it harbours life stages during which it is regeneration competent and others when it is regeneration incompetent. As these stages are experimentally accessible, one can compare the key mechanisms that permit the switch between regeneration competence and incompetence in the same organism. Xenopus also provides an accessible testing ground for exploring pre-clinical interventions that may enhance regeneration in NR stages, prior to exploring related approaches in other models that are poor at regenerating, such as mammals.

Of course, as with any model organism, Xenopus has its limitations. In particular, the generation of transgenic and mutant lines is time consuming, primarily because of the time required for Xenopus to reach sexual maturity, which can range from 6 months to over a year in X. tropicalis and X. laevis, respectively. In addition, although Xenopus is becoming more widely used as a model organism for regeneration studies, it is relatively new compared with other models, and therefore regeneration research in Xenopus, especially X. tropicalis, is largely in its infancy. Nonetheless, many general principles of regeneration were first found in Xenopus, including the lineage-restricted nature of blastemal stem/progenitor cells, the reactivation of developmental signalling pathways during regeneration, and the key role of ROS and bioelectricity for successful regeneration outcomes.

An additional key advantage of Xenopus is that it is not only an excellent model for regeneration, but also for developmental studies. Hundreds of animals per fertilisation can be grown cheaply and whole-animal or targeted manipulations can be easily performed. This includes the ability to genetically modify only the left or right side of an embryo, leaving a contralateral control side, which is a unique advantage of Xenopus. We and others have also shown remarkable parallels between the mechanisms underpinning tissue regeneration and development. For example, sustained ROS production is required both for appendage regeneration and for early development (Han et al., 2018; Love et al., 2013). Xenopus provides an excellent model to study the shared mechanisms, and any interplay between them, that are responsible for both tissue formation and regeneration.

Finally, despite the time required for the generation of transgenic and mutant Xenopus lines, there is an ever-growing body of methodologies and resources available for Xenopus, including centralised information resources, and reagent and animal suppliers, such as Xenbase, the National Xenopus Resource and the European Xenopus Resource Centre, that facilitate access to transgenic and mutant lines, but also where researchers can learn techniques and tools especially relevant to Xenopus (Gilchrist et al., 2004; Grant et al., 2015; James-Zorn et al., 2013; Pearl et al., 2012). For these reasons, we anticipate that continued studies in X. laevis and X. tropicalis will continue to provide important new insights relevant to regenerative medicine for decades to come.

We apologise to authors whose work could not be included in this Primer owing to space constraints.