Mechanisms of regeneration: to what extent do they recapitulate development?

Regenerative research is traditionally marked by Abraham Trembley's seminal work on the Hydra in the 18th century (Trembley, 1744). The subsequent inclusion of a variety of organisms, notably salamanders (see Glossary, Box 1), broadened investigations into the evolutionary roots of regenerative capability (Spallanzani and Spallanzani, 1768; Morgan, 1901). The 20th century brought a deeper dive into embryogenesis and organogenesis, leading to discussions on whether regeneration (see Glossary, Box 1) represents a recapitulation of development (see Glossary, Box 1) (Muneoka and Bryant, 1982). This hypothesis has been detailed since the early 21st century, with research efforts converging on the similarities of cell fates and individual gene functions during both development and regeneration. Although there have been numerous studies demonstrating differences between development and regeneration (discussed throughout this Review), the common perception has remained that regeneration mirrors development. However, in recent years, technical breakthroughs in genome editing, single-cell -omics and imaging, along with the integration of an extensive array of emerging experimental model organisms, have started to position regeneration research as a gateway to unveiling a hidden biology that is activated by injury and may diverge from developmental processes [which can be likened to the analogy of (re)building a house; see Box 2]. Accordingly, recent findings have begun to reveal additional molecular and cellular divergences between the two, and have highlighted further regeneration-specific mechanisms. Beyond being a fundamental biological debate, the discovery that regenerative processes might not simply echo developmental sequences may present opportunities for the development of innovative methods to reconstruct lost or damaged organs or structures, potentially contributing to and revolutionizing therapeutic strategies.

In this Review, I posit regeneration as a biological response to injury that orchestrates complex morphogenetic processes, including the restoration of morphological features and functionality of the original structure that was compromised or lost. The study of development and regeneration parallelism, both historically and more recently, has been investigated primarily through amphibian structural regeneration, such as limb or tail regrowth (Muneoka and Bryant, 1982; Suzuki et al., 2024). Additionally, discussions in this debate have included the pluripotency of certain cell types in planarians (see Glossary, Box 1) (Alvarado, 2000). Therefore, this Review focuses primarily on these areas, while discussing some of the evolving considerations that currently constrain a more nuanced understanding of this captivating discussion.

Many discussions on the parallels between regeneration and development have revolved around gaining a deeper insight into the blastema (see Glossary, Box 1): a structure that forms at the amputation plane during regeneration or in the animal body during development to give rise to complex structures. This tissue is mainly identified by its morphological features, which resemble a pool of undifferentiated cells. Although the term blastema is commonly associated with regeneration, it has also been used to describe developmental mechanisms throughout the 20th century and beyond, where it refers to tissues or group of cells in developing animals or plants, at sites where new organs will form, or to clusters of embryonic cells (please see detailed discussion by Holland, 2021). In early studies, morphological and proliferative features of cells in a blastema were found to be similar to those of embryonic cells, and the ability of these cells to differentiate into different lineages suggested a possible pluripotency for both conditions (Umanskiĭ, 1938; Wallace and Wallace, 1973; Morrison et al., 2006). However, the definitions of cellular plasticity states (see Glossary, Box 1), such as multipotency (see Glossary, Box 1) or pluripotency, were less clear at the time, which has also influenced their association with injury or regeneration.

Nonetheless, the properties of regenerative blastemas, coupled with ambiguous definitions of cellular plasticity, have not only bridged the understanding of development and regeneration but also influenced cross-species comparative studies without taxonomic limitations. In particular, flatworm planarians can regenerate their entire body, and the planarian blastema contains cell populations called neoblasts that appear to have pluripotent-like properties to mediate regeneration (Ivankovic et al., 2019). This pluripotent property has further propelled the notion of a widespread similarity of blastema during both development and regeneration across different animals, serving as an entry point to decipher the evolutionary underpinnings of regenerative capabilities (Alvarado, 2000).

Additionally, the concept of the blastema is associated with Thomas Morgan Hunt's highly influential categorization of types of regeneration. These types included so-called epimorphosis and morphallaxis (see Glossary, Box 1), in which cells remaining after amputations either proliferate or rearrange to drive regeneration, respectively (Morgan, 1901). Throughout the years, epimorphosis became the concept commonly associated with the blastema in a wide range of regeneration scenarios (Agata et al., 2007; Elchaninov et al., 2021). Meanwhile, morphallaxis research focused almost exclusively on Hydra, while mostly excluding them from the debate over parallels between development and regeneration. These discussions were, at their core, sparked and driven by morphological observations.

Morphological analysis continues to play a role in defining complex biological structures, yet the advent of advanced molecular tools, alongside exhaustive single-cell and genome-wide evaluations, is reshaping our understanding of cell types of regeneration and the blastema. The enhanced precision offered by modern stem cell research and developmental biology also contributes to refining these concepts. The notion of pluripotent cells (see Glossary, Box 1) in planarian blastema remains supported and may extend to similar regenerative processes in other flatworms (Kimura et al., 2022). Meanwhile, contemporary research using molecular tools, especially in vertebrates, reveals that the concept of a universally applicable pluripotent blastema across all animals and anatomical structures is not substantiated (Aztekin, 2021a). A series of recent discoveries, discussed below, indicate that regenerative cells, such as those found in a blastema, are highly context-dependent. Moreover, in vertebrates, the blastema is increasingly regarded as an assembly of lineage-restricted stem and progenitor cells, which may or may not bear varying degrees of resemblance to cells involved in development.

During the development of organs, multipotent lineage-restricted progenitor and stem cells work to form complex structures. Take lung progenitors as an example: they generate cells for the lung, such as alveolar cells, but not muscle or brain cells. Early regeneration research using lineage-tracing methods (see Glossary, Box 1) indicated a certain plasticity in regenerative cells that was not typical in development. For example, electroporation-based lineage tracing in the axolotl spinal cord suggested a potential transdifferentiation (see Glossary, Box 1) where spinal cord cells could produce muscle and cartilage (Echeverri, 2002). However, genetically stable lineage tracing techniques with minimal side effects later revealed a stricter germ layer and lineage fidelity in regenerative cells, which mirrored developmental cell type competencies, such that the muscle lineage reforms muscle and the epidermal lineage reforms the epidermis (Kragl et al., 2009). Indeed, advanced in situ lineage-tracing examinations, as well as single-cell RNA sequencing (scRNA-seq)-based studies (see Glossary, Box 1) with multiple regenerative model systems, have collectively established that, although regenerative cells may possess certain expanded potencies, mostly in line with their lineage, they do not necessarily exhibit dual lineage identities or display pluripotency [e.g., Xenopus tail (Gargioli, 2004; Aztekin et al., 2019; Kakebeen et al., 2020), salamander limbs (Kragl et al., 2009; Gerber et al., 2018), zebrafish fins (Tu and Johnson, 2011) and mouse digits (Rinkevich et al., 2011; Johnson et al., 2020)].

One of the well-established regeneration-specific cell fate mechanisms involves muscles; studies based on morphological changes in newt limb regeneration revealed a possible reversal of cell fate – a stark contrast to typical developmental processes. During development, multinucleated muscle cells form from mononucleated Pax7⁺ embryonic satellite cells (Chal and Pourquié, 2017). Meanwhile, during regeneration, multinucleated muscle fibers in newts can revert to mononucleated cells, suggesting a morphological and functional shift back to a more primitive cell state, often interpreted as dedifferentiation (Hay, 1959) (for a discussion on the definition of dedifferentiation, see Box 3). In contrast, Xenopus tail and axolotl limb regeneration seem to depend on Pax7⁺ satellite cells, as demonstrated by various lineage-tracing approaches, including tissue transplantation and CRISPR-based (see Glossary, Box 1) knock-in strategies (Chen et al., 2006; Sandoval-Guzmán et al., 2014; Fei et al., 2017). However, whether these cells function like embryonic Pax7⁺ satellite cells, and are therefore equivalent, remains unclear. Nevertheless, the muscle lineage demonstrates the potential use of both developmental- and regeneration-specific strategies to rebuild muscle tissue after amputations. This diversity further emphasizes that, even within closely related species, multiple strategies can be used, in contrast to the developmental setting, where a single strategy is employed.

Another regenerative phenomenon that has long intrigued scientists is the similarities between the regenerative amphibian limb blastema and the developing limb bud mesoderm (see Glossary, Box 1), where the limb bud mesoderm lineage, particularly fibroblasts (see Glossary, Box 1), has been postulated to be responsible for the formation of the blastema (Muneoka et al., 1986; Sessions and Bryant, 1988). Subsequent research in axolotls, employing molecular tools for lineage tracing (Kragl et al., 2009; Gerber et al., 2018) and, more recently, scRNA-seq (Gerber et al., 2018; Leigh et al., 2018), has confirmed that fibroblast populations significantly contribute to blastema formation, and can differentiate into both soft (fibroblastic cell types) and hard (skeletal system) connective tissue lineages, exhibiting the lineage-restricted multipotency of limb bud mesoderm (Logan et al., 2002). Moreover, scRNA-seq analyses on axolotl limbs have also revealed significant overlaps and differences between such regeneration-associated cells with limb bud mesoderm properties (Gerber et al., 2018; Leigh et al., 2018) (Fig. 1A). However, ongoing single-cell investigations in a diverse range of species, such as in frogs and mice, suggests that the limb bud mesoderm does not contain a stable singular transcriptional program (e.g. Feregrino et al., 2019; Aztekin et al., 2021; Bastide et al., 2022; Markman et al., 2023; Zhang et al., 2023). This unresolved variability potentially introduces limb bud mesoderm as a continuous cell identity across a spectrum, instead of a discrete transcriptional program, and further complicates comparisons with the blastema. Moreover, although numerous cellular and molecular players impacting the blastema have been revealed (reviewed by Bassat and Tanaka, 2021; Min and Whited, 2023), further examinations are required to understand how the previously identified mechanisms reflect on this evolving concept of cell fates in a blastema.

Also crucial for blastema formation is the apical epithelial cap (AEC; see Glossary, Box 1) (Christensen and Tassava, 2000). The AEC is formed from a migrated skin covering an amputated area during, for example, amphibian limb or zebrafish fin regeneration, which in turn enables blastema formation by providing necessary signals, including extracellular matrix factors, as well as mitogenic and chemotactic cues (Haas and Whited, 2017; Aztekin, 2021b). Historically, the AEC was thought to be analogous to the apical ectodermal ridge (AER; see Glossary, Box 1), a developmental tissue that forms temporarily and is integral to limb morphogenesis (see Glossary, Box 1) (discussed by Aztekin, 2021b). This analogy was based on a potentially similar function (since their elimination can halt development in amniotes or regeneration in salamanders) and shared gene expressions for a select number of genes, therefore suggesting that the AER might reform to act as the AEC (Christensen and Tassava, 2000). However, this hypothesis, originally proposed based on observations in salamanders, led to some confusion, since salamanders were also considered to not have an AER during their development (Tank et al., 1977; Purushothaman et al., 2019; Glotzer et al., 2022). Unlike salamanders, Xenopus tadpoles are known to have an AER during their development and AEC during their regeneration (Keenan and Beck, 2016). Leveraging this, scRNA-seq-based findings in Xenopus provided direct evidence that limb regeneration-associated AEC uses cells very similar to the AER (Aztekin et al., 2021) (Fig. 1B). Furthermore, as the previous conclusions in salamanders were based on tissue morphology or low-throughput gene examinations, revisiting salamanders with scRNA-seq-based cross-species analyses has recently identified AER-like cells in axolotl developing limbs (Zhong et al., 2023) (Fig. 1B). After this identification, it was shown that AER-like cells, in fact, do not fully reform during axolotl regeneration, unlike in Xenopus. Moreover, axolotl blastemal cells, but not Xenopus tadpole blastemal cells, specifically adopt parts of AER cell identity during regeneration (Zhong et al., 2023), suggesting distinct regenerative mechanisms that differ from developmental processes.

Despite the above discussed findings related to the amphibian limb blastema and the AEC, it remains unclear how the developmental state of amphibians influences their regenerative mechanisms. Xenopus tadpoles can regenerate their limbs only during their development (Dent, 1962), raising the possibility that their juvenile state may more readily reform developmental cell types during regeneration, thereby reusing developmental mechanisms. Likewise, axolotls are neotenic animals (see Glossary, Box 1), retaining juvenile features into adulthood, and the majority of axolotl studies in the literature examine pre-metamorphic animals that are still growing. Thus, axolotls, with their juvenile features, might be also employing mechanisms that share significant similarities to their developmental features, although they may have alternative strategies in their post-metamorphic stages. Further comparisons in animals at different developmental stages, such as between larval and post-metamorphic salamanders, could reveal multiple routes to regeneration within the same species.

In contrast to vertebrates, extensively studied whole-body regeneration-competent species, such as the planarian Schmidtea mediterranea and the fresh-water polyp Hydra vulgaris, provide a more-complex comparison point. These organisms can reproduce either asexually, using resident stem cell sources, or sexually, providing two different points of comparison for development and regeneration (Ivankovic et al., 2019; Vogg et al., 2019). Certain components during asexual reproduction of these species have been proposed to recapitulate processes that are used during regeneration (see discussions by Ivankovic et al., 2019 and Vogg et al., 2019). However, whether their sexual reproduction and subsequent embryogenesis and development (similar to those of vertebrates) mirror regeneration is less well understood.

Planarians use piwi-1⁺ pluripotent neoblasts, which reside throughout the animal body, to maintain homeostasis and form a blastema during regeneration (Ivankovic et al., 2019). Remarkably, the transplantation of a single adult neoblast into a planarian lacking stem cells can restore and reconstitute the organismal homeostasis, showcasing their long-known pluripotency (Wagner et al., 2011; Zeng et al., 2018); however, it remains unclear whether this pluripotent nature is similar to that of vertebrate embryonic pluripotent stem cells (see Glossary, Box 1). Recent research has expanded on this knowledge, revealing that the planarian blastema contains not only pluripotent neoblasts but also lineage-primed progenitors (Reddien, 2013; van Wolfswinkel et al., 2014). Nonetheless, the majority of planaria studies focus on asexually reproduced clonal strains, which are not suitable for studying sexual reproduction and embryogenesis. Leveraging a different planaria clonal line, one study delved into the differences between planarian embryonic cells, which form after sexual reproduction, and adult neoblasts (Davies et al., 2017). In this work, piwi-1⁺ cells were shown to be identifiable early in embryonic development and capable of differentiating into temporary and terminally differentiated populations, including lineage-primed progenitors relevant during regeneration. Nonetheless, planarian early embryonic transcriptional programs did not mirror those in adult neoblasts, nor did these early embryonic cells exhibit the same functional capabilities as neoblasts, based on grafting studies (Davies et al., 2017), underscoring the distinct properties between cells involved in development and regeneration. Unlike planaria, Hydra mainly use three lineage-restricted stem cell sources: unipotent epidermal, unipotent gastrodermal and multipotent interstitial stem cells, during their asexual reproduction and regenerative processes (Vogg et al., 2019). However, similar to planarians, commonly used Hydra also present practical challenges to studying their sexual reproduction (e.g. difficulties in distinguishing which animals underwent asexual or sexual reproduction, and variations in animal sizes), leaving it unclear whether their developmental and regenerative mechanisms are similar.

The cellular landscapes of both planarians and Hydra regeneration have been mapped at the single-cell level (Benham-Pyle et al., 2021; Nuninger et al., 2024 preprint), enabling a detailed comparison of their developmental processes, which awaits further elucidation at a similar level of resolution. In contrast to these commonly used model organisms, the examination of related species could provide a more feasible comparison. For example, the regenerative panther worm Hofstenia (Kimura et al., 2022), as well as cnidarians Nematostella (Röttinger, 2021) and Hydractinia (Bradshaw et al., 2015), are more readily available for sexual reproduction and embryonic studies. Thus, the examination of such lesser-studied regenerative species using single-cell methods promises to reveal cell fates that emerge during embryogenesis, and to determine which reproductive mode and subsequent cell fates more closely reflect their regenerative mechanisms.

Morphogenesis during regeneration may represent a unique process that deviates from the standard developmental patterns for tissue shape, patterning and scale. Some of these deviations may have no significant impact on regeneration. For example, the generation of somites, which are physical segments of mesodermal cells that serve as the foundation for the vertebrae, is a case in point. These structures form during the development of the body, including the tail, but when the amphibian tail regrows after amputation, the distinct physically separate somites are not observed, despite the presence of some related cell types (Gargioli and Slack, 2004). Meanwhile, some deviations from developmental patterns have been associated with a failure to perform complete regeneration. A notable variation is observed in lizard tail regeneration; instead of regenerating an ossified skeletal system, lizards generate a cartilaginous tube, which has led lizard tail regeneration to be considered an ‘imperfect’ regeneration scenario (Vonk et al., 2023). Interestingly, a similar situation is also seen in axolotls, where amputated tails containing notochord do not regenerate as notochord, but as cartilage (Echeverri et al., 2001). These examples also highlight another component of the house analogy (Box 2), where parts of the original structure have been replaced with another material, but the ultimately reformed structure retains its function.

Another example is seen in Xenopus post-metamorphic froglets, where amputations result in a cartilage-containing spike formation (Dent, 1962), which is associated with improper Hox gene activations that are related to setting correct limb positional information (Satoh et al., 2006; Lin et al., 2021). These findings emphasize that a significant departure from the typical developmental process may or may not correlate with failure to generate an exact replicate of the initial structure. Moreover, such identified changes point to deeper layers of complexity in the molecular and cellular underpinnings of regenerative morphogenesis.

Contemporary research showcases distinct regenerative mechanisms influencing morphogenesis. A recent intriguing study found that, in newts, the deletion of fibroblast growth factor 10 (Fgf10) – a well-studied crucial gene influencing limb formation – leads to the incomplete development of hindlimbs (Suzuki et al., 2024). However, incredibly, when these malformed hindlimbs are amputated, some of them regenerate into almost perfect limbs (Suzuki et al., 2024) (Fig. 2A). Similarly, another recent study examined the role of the somite segmentation regulator Hes7 during axolotl tail development. Although disruption of Hes7 during development resulted in vertebral fusions, the same perturbed tails, when amputated and regenerated, had nearly identical vertebral morphology to non-perturbed controls (Masselink et al., 2024 preprint). This stark contrast between initial development and subsequent regeneration highlights the distinct roles that Fgf10 and Hes7 play in these processes. In contrast, deletion of a combination of newt Hox genes shows a more consistent phenotype, with mutations affecting limb development also affecting limb regeneration (Takeuchi et al., 2022) (Fig. 2B). In another line of investigation, differences between how skeletons form during development versus during regeneration have been examined (Kaucka et al., 2022). In contrast to developmental skeletogenesis, where bone growth and hardening occur simultaneously, in regenerating newt limbs, ossification is delayed until the limb is fully formed, and it is coupled with an expansion of the cartilage in a transverse direction, leading to a bulkier skeletal structure (Kaucka et al., 2022). Altogether, these examples demonstrate that both molecular and cellular mechanisms can exhibit significant divergences during development and regeneration.

Patterning programs illustrate the various overlaps and distinctions between development and regeneration, and accentuate the plastic nature of regenerative mechanisms. For example, sequential activation of Hox clusters, which mark distinct proximal-distal limb segments during development, is well established to be recapitulated during amphibian limb regeneration (Gardiner et al., 1995; Roensch et al., 2013). Recently, developmental and post-amputation-specific regulatory regions for Hox genes have been identified in the limbs of the axolotl, suggesting a divergent regulatory landscape that guides proximal-distal axis formation during development and regeneration (Kawaguchi et al., 2024). Meanwhile, the dorsal-ventral axis specification related genes Lmx1b and En1 were examined during mouse digit tip regeneration. Interestingly, after digits have been formed, deletion of these genes does not significantly disrupt the regenerating fully patterned digit tips (Johnson et al., 2022). Conversely, ablating Lmx1b⁺ cells during digit regeneration impairs regeneration (Mahmud et al., 2022), indicating that genes in these cells other than Lmx1b might play a more crucial role. Moreover, generating En1 or Lmx1b mutants that contain double-ventral or double-dorsal digits, respectively, and amputating these digits, showed that double-dorsal digits can still regenerate, while double-ventral digits can form small blastemas that fail to rebuild any structure (Castilla-Ibeas et al., 2023). Collectively, these findings indicate that establishment of the dorsal-ventral axis is not a prerequisite for proper digit tip regeneration, in contrast to its developmental counterpart. Last, a recent study in the axolotl limbs has shown that anterior-posterior axis information is mediated by a positive-feedback loop that centralizes posterior Hand2⁺ cells, which are first formed during limb development and maintained in later stages of the axolotl life cycle (Otsuki et al., 2023 preprint). Moreover, mosaic deletion of Hand2 results in limb development with fewer digits, and amputation of these limbs leads to regenerates with even fewer digits and, in some cases, additional digits (Otsuki et al., 2023 preprint) (Fig. 2C), highlighting yet another distinction between developmental and regenerative requirements.

Development and regeneration often unfold at vastly different scales, which may require the involvement of additional elements for animal size-appropriate morphogenesis. Notably, regenerating organs can significantly surpass the size of their embryonic counterparts (Wells et al., 2021), profoundly influencing biological phenomena, including mechanical properties and morphogen (see Glossary, Box 1) gradients that rely on diffusion. Indeed, limb blastema transplantation between axolotls of different sizes showed that the final size of regenerated limbs is regulated by animal size, but not by the size of the transplanted blastema (Wells et al., 2021). Moreover, this animal size-appropriate regeneration was suggested to be influenced by the number of neuronal connections and the type of neurons (Wells et al., 2021). Furthermore, the amplification of developmental mechanisms may also play a part in the scaling of regenerative processes. For example, the interplay between posterior expression of sonic hedgehog (SHH) and anterior expression of FGF8 is crucial for limb morphogenesis across different sizes of axolotl, and the extent of these expression domains scales with the overall size of the axolotl (Furukawa et al., 2022 preprint) (Fig. 3A). These discoveries underscore the diversity of regenerative mechanisms, from integrating new components and adapting to size differences, to modifying existing developmental pathways to animal size-appropriate morphogenesis.

Different dynamic processes are observed during regeneration and development. Regenerative species are known to adapt to the level of amputation. For example, whether the amputation is at the upper or lower arm of a newt, regeneration completes at a consistent timepoint (Goss, 1969) (Fig. 3B). Moreover, more proximal amputations in newt limbs, zebrafish fin or Xenopus tails lead to faster growth compared with distal amputations (Iten and Bryant, 1973; Lee et al., 2005; Hayashi et al., 2014) (Fig. 3B), demonstrating that growth dynamics, and hence regeneration speed, are not fixed but instead can adapt to the position of injury. This flexibility is in contrast to the more rigid developmental growth rates, which are typically consistent within a species. The factors driving these differences remain elusive, although research on Xenopus tails suggests that the Hippo signaling pathway, potentially along with the tissue mechanical properties, might influence position-related growth rates (Hayashi et al., 2014). Furthermore, limb regeneration in salamanders has been found to cause dynamic changes in translation machinery (Subramanian et al., 2023; Zhulyn et al., 2023), although it remains unclear whether these changes in translation are relevant to limb development.

Such dynamic aspects could also have implications for the broader gene regulatory networks. Indeed, cell type-specific enhancer sequences in the zebrafish fin and heart have been shown to be activated only after injuries (Kang et al., 2016; Wang et al., 2020). Moreover, investigations into leg development and regeneration in the crustacean Parhyale hawaiensis have shown that similar gene expression programs are active in both scenarios, yet they differ in their dynamic activation times, and the variation in gene expression program is greater across individuals during regeneration (Sinigaglia et al., 2022) (Fig. 3C). Another study in the sponge Sycon cilatum found distinct gene expression patterns for early regenerative processes that then mirror developmental gene expression in the late stages of regeneration (Soubigou et al., 2020). Meanwhile, studies on zebrafish heart regeneration did not find such parallels between development in gene expression patterns, and showed distinct enhancer activations (Weinberger et al., 2024).

The interplay between these various mechanisms remains complex and undefined, but together they suggest that the dynamic nature of regeneration offers a variety of alternative strategies that depart from developmental pathways. As with the house analogy (Box 2), these examples show that the reconstruction process may require alternative timescales, a new blueprint and novel construction methodology.

Although there may be overlaps, some aspects of regeneration distinctly diverge from developmental processes. Regeneration begins in response to injury (which under normal circumstances is not encountered during development), which results in impaired structures, cell death and extracellular matrix disorganization (Chera et al., 2009; Aztekin et al., 2020; Ahmad et al., 2023). This disruption is thought to be partly remediated by the immune response. In particular, many vertebrate regeneration processes are known to require controlled activation of the innate immune system (Godwin et al., 2013; Petrie et al., 2014; Simkin et al., 2017; Aztekin et al., 2020). Considering that organisms with compromised immune systems can develop normally, it suggests that the involvement of immune system function may bear regeneration-specific mechanisms. Indeed, macrophages, for example, can directly promote muscle regeneration after injury (Ratnayake et al., 2021) or can contribute to collagen deposition to promote zebrafish heart regeneration (Simões et al., 2020). In addition, another myeloid lineage cell type, osteoclasts, have been shown to be important for stump bone degradation, which is needed for blastema formation and limb segment integration during regeneration (Riquelme-Guzmán et al., 2022). A second non-developmental distinction is evident in the initial response of the skin to injury. Amputations create an open area that must be sealed via wound healing, a process not encountered within development. Moreover, highly regenerative species perform such a wound healing response rapidly without scar formation (Godwin and Rosenthal, 2014), further emphasizing the divergent immune response, as well as the possibility that skin cells can be wired to differentially activate upon mechanical perturbations. Indeed, it has been recently recognized that repeated injury to skin cells causes changes in the chromatin landscape, priming them to counter inflammation (Naik et al., 2017; Larsen et al., 2021). Thus, the immune and skin responses to closing wounds are clearly divergent, accentuating a specialized non-developmental aspect of regenerative mechanisms.

Neuronal and muscular lineages propose specific regenerative mechanisms, spurring long-standing debates in regeneration research. Although neuronal involvement is often deemed essential for regeneration, this requirement is not universal across species and for development. For example, research on Xenopus tadpole limbs (Filoni et al., 1999), deer antlers (Suttie and Fennessy, 1985) and mouse digits (Johnston et al., 2016; Dolan et al., 2022a) has shown neurons are not indispensable for successful regeneration, unlike in salamander limbs (Singer, 1978) or fish fins (Simões et al., 2014). Intriguingly, limb development in salamanders also does not require neural connections, but if the limbs have previously connected to nerves, they then become dependent on neural signals for regeneration (Popiela, 1976; Kumar et al., 2011). These observations suggest that the necessity for neural involvement in regeneration is highly context dependent and can deviate from patterns seen in development. Similarly, limb muscle formation in salamanders is not crucial for limb development, and the absence of muscles has been shown to affect limb regeneration marginally (Hu et al., 2022b). Collectively, these insights underscore that some lineages, although not essential during development, may have intricate and variable roles during regeneration.

Recent research has identified additional factors in regeneration that are typically not involved in development. Notably, cellular senescence (see Glossary, Box 1), commonly linked to aging, has been shown to positively contribute to regeneration. Contrary to its non-proliferative nature, senescence following injury appears to influence nearby cell proliferation in axolotl limbs through Wnt pathway activation (Yu et al., 2023). Strikingly, senescent cells in the cnidarian Hydractinia have been demonstrated to reprogram terminally differentiated cells to progenitor states (Salinas-Saavedra et al., 2023), a trait that counters the general association of senescence with terminal differentiation and aging. Although developmental senescence has been implicated in tissue remodeling and influencing patterning gene expression (Muñoz-Espín et al., 2013; Storer et al., 2013), whether it can alter cell fate during development remains ambiguous.

Transient cell states are known to be significant contributors to regeneration. Comprehensive time-course scRNA-seq datasets during planarian whole-body regeneration have revealed that amputations induce the formation of transient cell populations in the muscle, epidermis and intestine (Benham-Pyle et al., 2021). Although it remains unclear whether these transient cell states are required for regeneration, they express genes involved in proliferation, tissue remodeling and tissue polarity, all of which are necessary for successful regeneration (Benham-Pyle et al., 2021). The importance of transient cell populations is not limited to planarians, but has also been demonstrated in other regenerative processes, e.g. in epicardial progenitor cells and fibroblasts during zebrafish heart regeneration (Hu et al., 2022a; Xia et al., 2022), the epidermis during Xenopus tail regeneration (Radek et al., 2023 preprint), and the basal epidermis during killifish tail fin regeneration (Granillo et al., 2024 preprint). Whether similar transient populations are formed and play important roles during development is an unexplored topic.

The concept of organism-wide systemic responses and inter-organ communication activated by injury has come to the forefront in contemporary research (Sun and Poss, 2023). Injuries such as amputations in amphibian limbs (Payzin-Dogru et al., 2023 preprint) or mouse digits (Dolan et al., 2022b) are now understood to elicit changes throughout the body (e.g. inducing proliferation), likely mediated by the pre-established nervous and humoral systems. Some of these phenotypes have also been linked to specific enhancer activations (Sun et al., 2022). These systemic effects have been proposed to both facilitate and hinder regenerative potential, depending on the context. For example, amputating an axolotl limb triggers cell division in distant organs (Johnson et al., 2018) and also allows subsequent amputations of contralateral limbs to form a blastema in a shorter period (Payzin-Dogru et al., 2023 preprint). In contrast, amputating a mouse digit tip and then amputating other digits after some time results in reduced regeneration (Dolan et al., 2022b). Mechanisms guiding these processes and their relationship to developmental processes have yet to be clarified.

Cellular metabolism is also considered to undergo dynamic changes during regeneration (discussed in detail in Mahmoud, 2023). Briefly, it is generally accepted that stem and progenitor populations use glycolysis as the primary energy source, whereas more differentiated cells rely more on oxidative phosphorylation. This raises the possibility that regeneration may rewire cells to use glycolysis for their cellular energy needs, consistent with the idea that differentiated cells revert to their progenitor state (Box 3). Indeed, when a salamander limb (Peadon and Singer, 1966) or a mouse digit tip (Fernando et al., 2011) blastema forms, these tissues lack significant numbers of endothelial cells that would have provided oxygen to leverage oxidative phosphorylation. In alignment with this observation, it has been suggested that the blastema leverages glycolysis during regeneration of the whole body of planaria (Osuma et al., 2018), zebrafish tail (Sinclair et al., 2021; Brandão et al., 2022; Scott et al., 2022), zebrafish heart (Honkoop et al., 2019) and Xenopus tail (Love et al., 2014; Patel et al., 2022). However, rewiring to glycolysis during regeneration, and thus developmental traits, may not be the only route for regeneration. Recently, a spatial transcriptomics-based study suggested that amputations of mouse digit tips in young and old animals leads to different metabolic gene expression, with older mice showing higher glycolysis and oxidative phosphorylation at the same time compared with younger animals, although these old animals were also suggested to show less regeneration fidelity (Tower et al., 2022). Furthermore, although it has been documented that salamander limbs show specific signs of glycolysis during regeneration (Wolfe and Cohen, 1963), more detailed analysis, including mass spectrometry, have failed to detect a complete rewiring for glycolysis during axolotl limb regeneration (Rao et al., 2009; 2014; Sibai et al., 2020; Varela-Rodríguez et al., 2020). Collectively, these findings indicate metabolic rewiring as a component that can recapitulate developmental settings for specific species, although its generalizable role requires further investigation.

Consequently, specific facets of regeneration are intricately linked to injury responses, and further establish new non-developmental players or injury-induced functions during regenerative mechanisms. As in the house analogy (Box 2), building or repairing a house may require new tools that were not originally used during construction.

Important conceptual considerations have emerged based on recent technological advances, identifying regenerative and developmental cell types, and their associated transcriptional profiles. It is now recognized that determination of cell types extends beyond morphology to encompass genome-wide quantitative gene expression analyses (McKinley et al., 2020). Based on this, recent findings suggest that different species and structures possess distinct cellular landscapes (Aztekin, 2021a), leaving the broad applicability of regenerative mechanisms unclear. Furthermore, to illuminate the parallelism debate, we need more direct discrimination of what exactly a regenerative response and an injury response are. Regeneration is initiated with an injury, but not all injuries and their subsequent effects on cells result in regeneration. Ideally, to discriminate injury responses and identify true regenerative events, the effect of injury must be isolated from regenerative responses. This could be done by performing similar injuries, such as amputation, where one condition results in regeneration and another condition results in no regeneration. As the injury remains the same in this comparison, the changes observed in both would classify as an injury response. If this examination is carried out in the same species, the list of changes associated with regeneration can be further narrowed down, also eliminating species-specific responses. Therefore, focusing on amputation responses in species with persistent regenerative capacity, such as salamanders or zebrafish, cannot easily distinguish between regeneration and injury responses induced by amputation alone. This situation could obscure the broader spectrum of injury responses and potentially misidentify them as regenerative mechanisms. Examining species and systems that exhibit different regenerative capabilities, such as the Xenopus brain (Endo et al., 2007), spinal cord (Edwards-Faret et al., 2017), tail (Beck et al., 2003) or limb (Dent, 1962), zebrafish pectoral fins (Yoshida et al., 2020), or the mouse (neonatal) heart (Porrello et al., 2011) or digit tip (Borgens, 1982), and comparing their regeneration-competent and -incompetent conditions, may provide a more rigorous approach (Aztekin and Storer, 2022). By recognizing the context in which regeneration occurs, we can refine our evaluations across species and biological systems, and improve our understanding of the extent to which regeneration mirrors developmental processes.

With more comprehensive and quantitative definitions of regenerative and developmental cell types, and thus a formal understanding of the players in morphogenesis, can we finally begin to see the full picture. One of the key milestones in our understanding of development and regeneration will be to decipher how the identified cellular players interact in time and space, and to define the possibilities of morphogenesis scenarios. During development, animals follow a relatively similar morphogenesis trajectory through a series of cell-cell interactions with limited variability. Because of this highly reproducible nature of development, embryogenesis and model organisms have been chosen as excellent specimens for understanding biology. In contrast, regeneration must occur in much more diverse situations and is performed within significantly larger regenerative variability space, as seen by examples discussed throughout this Review. Additionally, injury that initiates regeneration can be inflicted with different tools, such as teeth from different animals (e.g. Bothe et al., 2020), a set of scalpels applying different pressures or chemicals producing different mechanical and chemical stresses. The position or angle of the injury (Wolfe et al., 2000) determines the area of the injury to be populated by cell types of different abundance and spatial organization, thus directing cell-cell interactions. Regeneration can also occur under varying environmental conditions, such as fluctuating oxygen levels or temperatures (Schmidt and Jasch, 1971), or different microbiomes (e.g. Chapman et al., 2022), which alter the animal physiology without causing lethality. Such variances, as well as the metabolic, immune or hormonal state (Hirose et al., 2019) in which the animal received the initial insult and how these parameters changed over the course of the regeneration, also introduce variability. In contrast, embryonic development occurs in a more stable condition (although they can also survive certain environmental constraints) or results in abnormalities or lethality. Consequently, the number of morphogenetic scenarios that can be recorded within regeneration exceeds the possibilities seen in development (Fig. 4). This increased flexibility positions regeneration as a process with greater variability, noise and adaptability in morphogenesis compared with development. However, this variability space also raises the likelihood of regenerative failure. As with any complex system influenced by many variables, the chance of a mismatch between parameters or a failure in one parameter impacting the entire system increases.

The debate between development and regeneration is a long-lasting discussion. The previous century's research revealed that although development can inform regenerative processes in some aspects, cellular capabilities for morphogenesis are multifaceted and new mechanisms can be unveiled after injury. Since the morphological evaluations that once enabled cross-species comparisons, and different regeneration scenarios (e.g. limb or whole-body) through studying the ‘blastema’, a myriad of new methods have ushered in a more-refined understanding of this discussion, pointing to the need to describe mechanisms specific to development or regeneration.

To establish the limits of the molecular and cellular space in which regenerative morphogenesis can occur, advances in methods, as well as the examination of lesser studied species, are needed. First, determining the spatial configuration and abundance of cell populations within tissues and their interactome remain hard to examine, limiting progress in understanding the geometry of development and regeneration. Likewise, how different lineage trajectories and system-wide coordination of differentiation occur and are timed remains unclear. The integration of morphogen gradients and mechanical properties in these processes demands sophisticated spatial-omics and 3D imaging techniques that are capable of capturing and quantifying entire regenerative systems. Studying regeneration and accounting for its variable nature would require processing large numbers of samples, which may pose practical challenges with current microscopy limitations. Beyond these approaches, developmental biology research has expanded by benefiting from in vitro simplified models such as stem cell-based systems, organoids and explants. Meanwhile, regeneration studies are predominantly reliant on in vivo research, with a few exceptions (Wang et al., 2015; Aztekin et al., 2021). Establishing simplified regenerative systems could provide a more practical and manipulable environment to further unravel mechanisms of regeneration. These deconstructed methods can play a pivotal role in integrating tissue, cell and molecular scale examinations. Last, the investigation of previously overlooked species through advanced -omics approaches could yield more comprehensive and measurable cross-species comparisons, offering insights into the evolution of regeneration, as well as how they diverge from their developmental processes. Altogether, establishing more comprehensive, quantitative and integrative methodologies, as well as embracing the variability in which regeneration occurs, can further reveal variances in morphogenesis during development and regeneration.

In summary, development typically adheres to a predetermined morphogenetic strategy, whereas regeneration is more dynamic, adjusting and responding to environmental conditions with a variety of tactics. This adaptability suggests the existence of previously unreported interactions between biological systems, untapped cellular and molecular mechanisms, and alternative routes that veer from the developmental script. Exploring such variability through this long-standing discussion between development and regeneration has the potential for discovering new principles of biology and inspiring therapies.

I thank the members of the Aztekin Lab and Celina Juliano for discussions and their critical reading of the manuscript. I also thank Jixing Zhong for her initial input on the manuscript and comments on fibroblast cell identity, Jochen Rink for discussions on cell fates in planaria, and Kelly Hu for discussions, particularly on regeneration variability space.