Hallmarks of Regeneration

Abstract

Regeneration is a heroic biological process that restores tissue architecture and function in the face of day-to-day cell loss or the aftershock of massive injury. Capacities and mechanisms for regeneration can vary widely among species, organs, and injury contexts. Here, we describe hallmarks of regeneration found in diverse settings of the animal kingdom, including activation of a cell source, initiation of regenerative programs in the source, interplay with supporting cell types, and control of tissue size and function. We discuss these hallmarks with an eye toward major challenges and applications of regenerative biology.

Short summary:

Poss and Tanaka provide a review of concepts and mechanisms in regenerative biology. They describe “hallmarks” that are broadly represented across species, tissues, and injury contexts, as a means to articulate structure for addressing knowledge gaps and potential applications.

Introduction

Regeneration, the restoration of lost cells, tissues, body parts, or whole animals, is a central feature of multicellular life. Development from embryonic to adult stages incurs cell death, stresses, and strains that provoke and guide morphogenesis, events akin to regeneration. Final form is not truly final -- cycles of renewal maintain critical tissues, such as blood and epithelial linings.

Mammals are vulnerable to heart attacks, spinal cord injury, limb amputation, and stroke, for which regenerative responses are inadequate. Molecular, cellular, and tissue damages accumulate as animals age, and the ability to resolve these insults is tempered. Regeneration can counter these damages and bring near-immortality to some species, such as flatworms and polyps. Other vertebrate species, among them salamanders and teleosts, show extraordinary capabilities to regenerate limbs, hearts, spinal cord, and more. Yet, the defense mechanism of regeneration inevitably falls short in vertebrates. Aging conquers regeneration in all scenarios.

Regeneration is in some ways similar to cancer, and attempts at comparisons have been frequent over the past decades. Most notably, each process commandeers the machinery of cell growth and patterning for its purposes. However, some may consider these two players to be diametrically opposed. DNA-based changes underlie cancer, and cancer cells eventually identify cheat codes to escape checks and balances, with insidious outcomes. In contrast, regeneration is initiated by an injury event, with an ostensible goal to achieve perfect form and/or function. Therapies to improve human regeneration are an inspirational objective of the 21^st century. Yet, both a hammer and scalpel are needed to launch morphogenesis in a controlled manner from a disabled organ or tissue stump, avoiding tumorigenesis and metastasis.

Regeneration is an historic and decorated field of biology, and many laboratory model systems are employed with their own spectra of regenerative capabilities. Discoveries in this field can inspire regenerative medicine. Here, we attempt to articulate a simple framework for understanding regeneration. Our aim is broad, and we skew our examples toward non-mammalian species, such as planarians, salamanders, fish, and the like – laboratory model systems with which many readers might wish to gain more familiarity. Principles of tissue organization, molecular pathways, and injury responses are conserved across phylogeny, and species with outsized regenerative abilities can serve as role models for mammals toward a potential explosion of the field if methodologies allow. We describe four “hallmarks” of regeneration that typically occur in response to acute trauma, covering what we have learned and must strive to learn about their defining features (Fig 1). Upon damage and initial healing, a competent cellular source(s) of lost tissues must be activated, programs for morphogenesis are initiated within and upon this source, productive cell interactions occur, and new and old cells functionally integrate and restore size and shape.

Fig 1.

Hallmarks of regeneration.

1. Activation of cell sources

Trauma leads to an initial healing response of some form in all tissues, which can involve events such as clotting and re-epithelialization. When healing is complete in a regeneration-competent tissue, renewal of structure and function begins. A complex tissue has many cell types, and each may have a different origin and/or mechanism of origin. A reserve of stem cells may be involved, or differentiated cells might be tweaked in a manner that relieves them of function and enables division. The lesson learned from experiments across many body parts, injury types, and animal species is that every possible way of recruiting source cells for regeneration – stem cell activation, dedifferentiation, transdifferentiation, hypertrophy and endoreplication – has been observed.

Stem cells.

Stem cells can self-renew and give rise to other cell types. Arguably the most familiar and elegant starter material for regeneration is the stem cell and its requisite niche (Fig 2A). Cell loss or tissue damage signals disrupt the serenity of the niche, and stem cells awaken tasked to restore function. Many classic and exceptionally well-studied examples of stem cell-based regeneration have been best studied in mammalian models, such as mice. These include blood, derived from hematopoietic stem cells in bone marrow niches, and skeletal muscle, derived from satellite cells located beneath the basal lamina of multinucleated myofibers. Skin and its appendages contain dynamic stem cell populations that regenerate the epithelial barrier (Fig 2B), as well as hair growth and color^1–3.

Fig 2. Activation of cell sources.

A. Stem cells reside in a quiescence-maintaining niche before they wake to renew and differentiate upon cell loss or damage.

B. Dynamic stem cell populations in the skin regenerate the epithelial barrier.

C. Pluripotent stem cells populate interorgan spaces in planarian flatworms and replace differentiated cells in the process of turnover.

D. Injury-induced partial dedifferentiation of cardiomyocytes leads to acquisition of embryonic features and division during zebrafish heart regeneration.

E. Iris pigmented epithelial cells transdifferentiate to lens cells in amphibians, by depigmenting, forming a vesicle, and initiating production of lens crystallin proteins.

F. Spared Drosophila pyloric cells endocycle in response to genetic ablation, forming a structure composed of a reduced number of large, polyploid cells with effective barrier function.

Other spectacular examples of regeneration involve pluripotent stem cell populations, for example in planarian flatworms (Fig 2C). Stem cells in the form of neoblasts populate interorgan spaces of planarians and mediate turnover. Some of these neoblasts are pluripotent, demonstrated by the ability of single transplanted cells to repopulate multiple tissue types and even save an animal from lethal irradiation⁴. Neoblasts have been classified by morphology and transcriptomes, and the surface marker TSPAN-1 emerged as an indicator of a pluripotent class⁵, with other pluripotent classes likely present⁶. Certain other neoblast classes are more committed to fates of the various tissues, for instance intestine, muscle, nervous system, or pharynx⁷. Key research goals for this system, and for virtually all stem cell-based systems, include further detailing the stem cell hierarchy and the signals and steps toward fate commitment, and determining how neoblasts sense the extent of injuries and adjudicate their response. Contemporary studies of stem cell biology have used multiplexed clonal analysis and barcoding approaches to tag related cells and track them animal-wide. Methods for transgenesis that enable direct, live visualization of neoblasts and their progeny in species such as acoel worms is an important future direction to complement inferential methods⁸.

Direct duplication and dedifferentiation.

Many tissues regenerate without using stem cells by duplicating parenchymal cells, and dedifferentiation is a process that can assist in this. In innate regenerative biology, dedifferentiation typically refers to the tempering of functional properties of a specialized adult cell. This process can involve the reduced expression of signature “differentiation” genes of that cell and/or increased presence of a more “embryonic” set of genes that is associated with the growth and function of the organ’s initial form. Many examples of dedifferentiation are evident in innate regeneration. For instance, zebrafish and other non-mammalian vertebrates can regenerate heart muscle after extreme loss through injury-activated division of cardiomyocytes⁹ (Fig. 1D). During zebrafish heart regeneration, transcriptomes of regenerating cardiomyocytes acquire similarities with embryonic cardiomyocytes, including expression of transcription factors and myosin forms associated with heart development, while also dissembling sarcomeres and acquiring a more glycolytic (versus oxidative) state^10,11. Importantly, hearts of neonatal mice and other mammalian neonates also display a regenerative response to injury, with limited scarring as an adult mammal would display^12–14. Neonatal heart regeneration also features cardiomyocyte division, dedifferentiation, and acquisition of glycolytic features^14–16.

Regeneration of an amputated salamander limb involves blood loss, major skin damage, and depletion of large regions of muscle and skeletal bone or cartilage. Thus, to achieve regeneration, the animal effectively requires everything, everywhere, all at once. Although each of these tissues has its own progenitor mechanism, cell source mechanisms for a specific tissue can vary among species, even those in the same vertebrate class or order, and depend on the developmental stage of the animal. Newts and axolotl limb regeneration each replenish limb skeletal muscle after amputation injury. Larval axolotls and newts rely on satellite cells for skeletal muscle regeneration, as do postmetamorphic axolotls, based on cell transplantation and genetic fate-mapping assays^17,18. However, adult newts do not rely on satellite cells during limb regeneration, but instead on a process of dedifferentiation¹⁸. Differentiated myofibers undergo fragmentation into a mononuclear cell population that regenerates lost skeletal muscle.

In mammalian lungs, airways and alveoli contain stem cells that promote vigorous renewal of compartment-specific epithelial cell types in response to inhaled insults¹⁹. Many genetic tools now exist to ablate specific cell populations in these structures, including the major stem cell and differentiated cell types, and studies employing these have revealed notable plasticity among lineages. For example, induced, extreme genetic ablation of large airway basal cells, by targeted diphtheria toxin A expression via a tamoxifen-regulated Cre transgene, enables dedifferentiation of detoxifying club cells to form multipotent basal cells that respond to additional injury²⁰.

Dedifferentiation has been an attractive target for changing regenerative capacity. Forced expression of classic reprogramming factors can affect the chromatin status and transcriptional readouts of differentiated cells, reverting cells at low frequency to a pluripotent state. Multiple recent studies have reported that transient expression of reprogramming factors (Oct4, Sox2, Klf4 and c-Myc; OSKM) alter the regenerative responses of mammalian tissues that normally display poor regenerative capacity^21–24. The effects of this type of partial reprogramming have been examined in many tissues, and have been reported, for instance, to improve the ability of mouse cardiomyocytes to divide and respond to injury stimuli^21–24. The concept of partial reprogramming brings interesting ideas to regenerative biology, although the tumorigenic effects of OSKM factors make it difficult to imagine therapeutic use without exceedingly sophisticated safety measures. Other manipulations have reported pro-regenerative effects of transcription factor overexpression on the ability of tissues, such as skeletal or heart muscle, to dedifferentiate^25,26. It is important to catalog manipulations that toggle differentiation in productive ways.

Transdifferentiation and ancillary sources.

Transdifferentiation is the transition from one differentiated cell type to another, and for the purpose of this review we can distinguish transdifferentiation from reprogramming as an event not involving experimentally forced expression of exogenous genes. Transdifferentiation is related to dedifferentiation in that it can involve (but does not require) progression of a cell through an undifferentiated state such as a stem cell. While a dedifferentiated cell remains undifferentiated or redifferentiates to its previous cell type, transdifferentiation takes the trajectory on a new path to a different cell type.

Prominent examples of innate transdifferentiation have been discovered in studies of regeneration of eye parts, which occurs effectively in nonmammalian models. For many years the archetype of transdifferentiation has been the lens of the eye in amphibians, such as species of newts. This tissue faithfully regenerates after complete extirpation in adult newts, involving a process of budding from the pigmented epithelial cells of the ventral portion of the iris²⁷. These cells depigment, form a vesicle, and begin to produce lens crystallin proteins to finalize the structure (Fig 2E). This phenomenon can be modeled ex vivo and also occurs in larval axolotls and tadpoles, although in these cases with origins of ventral or dorsal iris, and cornea, respectively^28–30. The eye retinae also regenerate to great extents in fish and amphibians after injuries, such as removal, photoablation or neurotoxic damage, occurring also by transdifferentiation ³¹. In salamanders and tadpoles, the retinal pigmented epithelium transdifferentiates to generate neurogenic precursors for all key cell types ^31,32, whereas structural Müller glial cells distributed throughout the retina repopulate neuron classes in teleosts, such as zebrafish³³.

Mammalian pancreas, stomach, and liver are capable of high degrees of transdifferentiation, but only in response to extreme cell disabling or depletion^34–36. Pancreatic islet beta cells have a limited capacity to divide in adult mammals and a poor proliferative response to injuries, such as streptozotocin-induced depletion. Yet, at a low rate and following near total genetic ablation of beta cells in mice, alpha or gamma cells transdifferentiate to acquire beta-cell features such as insulin production, in part through the release of inhibitory signals within islets^34,37,38. The digestive chief cells of the stomach undergo metaplasia to gland-like cells upon extreme injury³⁵. Liver regeneration invokes a classic model of direct duplication, wherein hepatocytes lost from a major resection injury will be replaced by division and/or hypertrophy of spared hepatocytes to replace lost tissue mass^36,39. However, liver cells show plasticity during regeneration when key cell types are genetically disabled. Hepatocytes can transdifferentiate to bile duct structures after genetic disruptions in Notch signaling perturb biliary development, events potentiated by Tgf signaling⁴⁰. Conversely, biliary epithelial cells initiate hepatocyte production under various conditions in which hepatocytes are burdened by genetic or toxic insults^40–43. Examples of mammalian transdifferentiation may be rare, but they afford inspirational glimpses of the potential of differentiated cells to repurpose themselves in vivo under selective pressure.

Endoreplication and hypertrophy.

Whereas most definitions of regeneration invoke the recovery of cell number as a means for restoring tissue mass, prominent examples of regeneration occur without hyperplasia playing a major role. Mammalian skeletal muscle regeneration, for instance, recovers muscle mass after a lesion through fusion of myoblasts generated from satellite cells, building new multinucleated myofibers and contributing new myonuclei to existing fibers. Dedicated fusogens expressed on each participating cell are responsible for the specificity of these fusion activities, eliminating the possibility of fusion with non-muscle tissue^44,45. Liver regeneration, which occurs through different source mechanisms, will in the least severe instances of tissue loss involve endoreplication to duplicate the genome and cause growth of hepatocytes, with classic cell division events being less frequent^36,39. This process may entail the formation of mononucleated or multinucleated polyploid hepatocytes, typically with larger cell volumes, derived either from diploid or polyploid hepatocytes. Functions of polyploid hepatocytes are measurably similar to diploid counterparts⁴⁶; thus, liver regeneration achieves functional goals although tissue architecture and cellularity may differ from the original.

Many mammalian cell types exist with elaborate structural and functional requirements, or restricted capacity for cytokinesis, such as cardiomyocytes, neurons, and chondrocytes. Cell cycle blocks limit the ability of cardiomyocytes to divide in postnatal mammals^47–49. Thus, mammalian cardiomyocytes often endoreplicate to increase cell and organ mass during animal growth. Polyploidy further limits the ability of cardiomyocytes to divide after an injury such as an ischemic infarct^50,51. This result is also substantiated by findings in zebrafish, in which experimental polyploidization of cardiomyocytes could inhibit injury-induced heart regeneration⁵².

Additional examples of injury-induced polyploidy have emerged in recent years that highlight benefits of a rapid mechanism to increase cell surface area. In Drosophila, a puncture wound to the abdominal epithelium leads to a combined response of endoreplication and fusion by epithelial cells surrounding the wound⁵³. These enlarged, polyploid epithelial cells efficiently heal large injuries. Flies display an analogous endocycling response when most pyloric cells of the hindgut are genetically destroyed, leading to formation of a pylorus with a reduced number of large, polyploid cells with effective barrier function^54,55 (Fig 2F). The recognition of polyploidy as a strategy for tissue regeneration is growing, and genetic and/or in vivo imaging tools such as those readily available in Drosophila should be informative for understanding how and why polyploidy occurs.

2. Initiation of regeneration programs

A simple equation for regeneration requires a competent cell source and an environment with signals to control its morphogenesis. Achieving this competency and environment seems far from simple, however. Thousands of genes sharply alter expression levels during regeneration, either elevating or declining, to effect change in participating cell types. Diverse signaling centers form from spared tissue and dying cells release influential factors. Gene expression levels change in impactful ways away from the injured tissues as well, and previous injuries can prime tissue for regeneration, invoking alerted stages and memory-based responses. It is a great challenge to interpret all inputs and variables.

Regeneration regulatory elements.

Many studies have now profiled chromatin-based indicators of active gene regulation, such as reduced nucleosome occupancy or increased presence of histone modifications, in a variety of species, cell types, and injury contexts^56–64. These studies indicate widespread changes in chromatin structure affiliated with regeneration, findings that are complemented by reports indicating roles for chromatin regulatory factors during regeneration^65,66. Thus, injury prepares the genome to institute changes in gene expression that enable the release of pro-regenerative stimuli, and the competency of multifarious cells to respond to these stimuli.

Transcriptional responses are afforded specificity by enhancer regulatory elements, during all life stages as well as disease. Regeneration programs are similarly controlled by contextual enhancers, whose sequences likely represent many of the genomic locations of the afore-mentioned changes in chromatin structure. These sequences, called tissue regeneration enhancer elements (TREEs) or similar, can in combination with a permissive promoter direct transient expression preferentially or specifically to regenerating tissue^67,68. Candidate TREEs from genomic profiling have been experimentally validated in many regeneration contexts. Whereas mechanosensation of injury might be all or part of the stimuli for transcriptional activation of some TREEs, evidence indicates that other TREEs receive control well after initial injury stress by additional or alternative aspects of regeneration such as inflammation, mitogens, hormones, and patterning factors^{69, 57}. Finding rules for TREEs and what binds them is essential and challenging given the complex in vivo context, and much remains to be discovered. Mutagenesis of enhancer transgenes has revealed regulation in part by common transcription factors, such as AP-1 and TEAD, in zebrafish fin and heart regeneration^59,67,70 (Fig 3A). Egr-mediated activation of regeneration-associated genes was demonstrated by bioinformatical predictions and gene knockdown in acoel worms⁶², among other examples. Silencers that reduce gene expression from homeostatic levels during regeneration can be more challenging to find and verify. Yet, a small number of studies indicate the presence of repressive modules within regulatory elements^71,72, or of bona fide silencers that form context-specific interactions with nearby genes and reduce their expression⁷³.

Fig 3. Initiation of regeneration programs.

A. Tissue regeneration enhancer elements (TREEs) control the expression of genes induced after injury and during regeneration.

B. Dying Drosophila wing disc cells secrete Wingless (Wnt) to promote compensatory growth of the surviving tissue. Senescent cells transiently appear in the salamander limb blastema and produce mitogenic Wnt ligand.

C. Skeletal muscle satellite cells in one limb respond to a contralateral limb injury by entering a G_alert stage, reported to be controlled by circulating Hepatocyte growth factor activator and requiring signaling through the Hedgehog effector Gli3.

As enhancers change rapidly during evolution, they may be targets for evolutionary changes in regenerative capacity^59,74. This is an interesting concept that requires further exploration of interspecies differences, but many examples exist in which a limited number of molecular manipulations improves regenerative capacity. For instance, blocking Wnt signaling imparts head regeneration in certain planarian species^75,76. One hypothesis is that injury-associated enhancer sequences or their regulation near Wnt inhibitors differ among flatworms. A recent study assessing inbred mouse strains with different regenerative capacities after ear punch injury revealed different allelic expression responses in F1 hybrids⁷⁷; cis regulatory element variation likely contributes to these allelic responses. These types of studies can be informative in identifying impactful enhancers and their regulators.

Signals from dying and senescent cells.

Regeneration represents not only renewal, but also the demise of cells that enter cell death or senescence programs. However, in their final act these cells can play critically pro-active roles. Under extreme stress and when entering the apoptotic program, Drosophila wing disc cells secrete Wingless (Wnt) and other regulatory factors to promote compensatory growth of surviving tissue⁷⁸ (Fig 3B). An analogous set of Wnt-producing cells participate in the regeneration of an amputated head in Hydra, and apoptotic cells are required for Xenopus tail regeneration^{79 80}.

Cell senescence is generally caused by stress and results in growth arrest and acquisition of a Senescence Associated Secretory Phenotype (SASP), characterized by the expression of several diagnostic markers such as interleukins, chemokines and growth factors⁸¹. Senescent cells accumulate and are cleared during regeneration by immune cells. They represent another class of transient participants that can help place injured tissue on a trajectory toward regeneration. As with other cells, such as macrophages (see below), timing is everything. Early and transient exposure to senescent cells and their secretomes supports stem cell proliferation and regeneration in multiple tissues. Lengthened exposure that occurs in aging will blunt stem cell activity and promote senescence⁸². OSKM-induced in vivo reprogramming into a stem cell state requires senescent cells generated in the process, in particular their secretion of the cytokine interleukin-6⁸³. In skin keratinocytes and liver hepatocytes, transient induction of senescence programs through experimental k-ras expression prompts neighboring cells to turn on stemness genes⁸². Moreover, brief treatment of keratinocytes with conditioned media from senescent cells leads to hair follicle neogenesis, whereas prolonged treatment of keratinocytes with the SASP triggers their own senescence, associated with expression of p16 and p19 cell cycle inhibitors.

Stem cell reprogramming through the presence of senescent cells is deeply evolutionarily conserved and already exists as an endogenous program that assists head regeneration in the invertebrate Hydractinia⁸⁴. In highly regenerative vertebrates, such as salamanders, senescent cells appear in the limb blastema but are efficiently cleared from the tissue⁸⁵. Interestingly, these transient senescent cells produce Wnt ligand that has mitogenic effects on neighboring blastema cells⁸⁶ (Fig 3B). Analogously, in regenerating zebrafish spinal cords, senescent cells are present at the beginning of regeneration but are eventually cleared, whereas in the mouse they persist⁸⁷. Experimental senolytic clearance of senescent cells from an injured mouse spinal cord improved the regeneration outcome. Distinct senescence factors are implicated in various injury, healing, and regeneration contexts, indicating that the SASP may be context-dependent and is likely a rich source of pro-regenerative factors that tissues have relegated toward their own renewal.

Memories of cell identity/developmental lineage or inflammation.

Multiple studies have demonstrated that regenerating tissues remember their lineage. After injury, cells do not return to a fully pluripotent state to cross germ layer boundaries but can step back to early progenitor cell states within their developmental history^88,89,90, an ability that limits the odds they will form cancerous growths, such as teratomas. For example, in the gut epithelium, facultative repressive histone modifications, such as H3K27me3, mark lineage-associated genes and restrict expression of developmental genes prior to regeneration. When crypt-associated Lgr5⁺ intestinal stem cells are genetically ablated, enterocytes and secretory cells dedifferentiate to replace this population, but this process relies on the expression of the transcription factor Ascl2⁹⁰. Intestinal cell types, such as stem cells that express Lgr5 or Bmi, secretory cells, and enterocytes, can have similar open chromatin domains, whereas H3K27me3 marks correspond to repressed gene states⁹¹. Genetic depletion of the Polycomb component Eed1 to inhibit H3K27me3 deposition stimulates cell division, which consequently dilutes H3K27me3. As a result, developmental genes and enhancers associated with specific lineages become active, implicating H3K27me3 repression as an important means to restrain expression of earlier developmental transcriptional states⁹².

Tissues remember not only their developmental history but also the trauma of past inflammatory events⁹³. Historical traces of open chromatin comprise one component of this memory, facilitating trained responses to future insults. The exposure of skin to inflammatory agents predictably leads to AIM2 inflammasome assembly, which triggers inflammatory responses. Yet, even after the tissue has recovered, persistent regions of open chromatin at the AIM2 sensor gene and downstream effectors (including caspase-1 that promotes IL18 and IL1b maturation) help accelerate subsequent injury responses⁹³. The persistence of the transcription factor Jun at the open chromatin sites underlies inflammatory recall⁹⁴.

Another pronounced example of epigenetic memory is displayed by hair follicle stem cells⁹⁵. Denuding skin of its epidermal and junctional zone cell layers exposes hair follicle stem cells in the bulge. Under these conditions hair follicle stem cells step in as a compensatory reserve for skin regeneration. These follicle cells employ a transcriptional program similar, if not near-identical, to that of normal epidermal wound healing. This activation of an ancillary source mechanism primes hair follicle-derived epidermal stem cells such that they heal a secondary mechanical wound at accelerated rates. Open chromatin memories appear to constitute the basis of this priming and not only influence inflammatory response genes, but also genes involved in cell migration and wound healing, such as growth factors FGF2, Tgfb2, and Tgfb2 receptor⁹⁵. This epigenetic priming can occur in regions distant from an injury site even in cells that had not participated in repairing the original wound⁹⁶. Taken altogether, stem and progenitor cells retain remnants of open chromatin corresponding to their life experiences, and these are tamped down by repressive factors, such as histone H3K27me3 modifications. These memories guide regeneration of a missing body part or enhance a regenerative response to future tissue damage.

Long-range regulation of regeneration.

A tissue-depleting injury is a potential animal-wide crisis that does not occur in isolation, and studies highlight communication systems between regenerating and remote tissues⁹⁷. For instance, circulating hormones, in some cases stage- or sex-specific factors, impact the regenerative capacity of organs and appendages. Classic experiments revealed that pituitary gland signals, such as growth hormone, are essential for limb regeneration in salamander species^98–100. In Xenopus, the hypothalamus coordinates the upregulation of aMSH in the early frog limb blastema, which signals through the Mc4r receptor. Interestingly, provision of aMSH to denervated frog limbs could at least partially rescue blastema formation, suggesting an early role for this pathway¹⁰¹. Exposure to or disrupting the presence of circulating androgens can alter the regenerative capacity of specific fins in platyfish and zebrafish^102,103, and affects the ability of deer antlers to cast and regenerate normally¹⁰⁴. Natural increases in thyroid hormone levels during development are associated with increased ploidy and reduced proliferative capacity of murine cardiomyocytes. Reducing cardiac thyroid receptor signaling during development can increase the diploid complement of cardiomyocytes and improve their proliferative response of the heart to injury⁵⁰.

It is well-known that responses to tissue damage can also affect the physiology of other organs. For instance, cardiac and renal injuries frequently coexist in mammals, and several pathways have been identified that mediate crosstalk between these tissues in injury contexts⁹⁷. Trauma and regeneration can control ontogenetic development in constructive ways as well. In Drosophila, trauma or genetic ablation of imaginal wing disc tissue in developing larvae leads to a suppression of ecdysone release that stalls growth of uninjured imaginal tissues^105,106. This relay, which is regulated by long-range action of the secreted factor relaxin on neurons that ultimately limit production of signals for ecdysone synthesis^107,108, enables regeneration of an injured tissue to “catch up” and synchronize with other developing structures. In zebrafish, damage to the heart leads to the expression of the transcription factor cepbd in remote tissues, such as the brain and kidney, under the control of a corticosteroid-responsive enhancer, a response that is required to retain physiological fluid dynamics during regeneration⁶⁹.

Tissue damage can also affect the proliferative responses of uninjured tissue. Limb amputation in salamanders can lead to activation of cell cycling across distant tissues, such as the contralateral limb, heart, and liver¹⁰⁹ (Fig 3C), an effect attributed in part to adrenergic signaling¹¹⁰. Amputation of a planarian head similarly induces body-wide cell cycling, after which proliferative events are localized at the injury site during the formation of a blastema^111,112. The initial response in this model is explained by a wave of Erk signaling activity that is proposed to be transmitted by muscles in the planarian body wall. Importantly, injury can ready uninjured tissue for enhanced regenerative responses¹¹³. Reported a decade ago was that skeletal muscle satellite cells in one limb could respond to a contralateral limb muscle injury by entering a G_alert stage with differential size, metabolic properties, and cycling dynamics, endowing them with a stronger response to a subsequent direct injury¹¹⁴. Studies have indicated that the G_alert state is controlled by circulating Hepatocyte growth factor activator, and that it also requires signaling through the Hedgehog effector Gli3^115,116. Muscle or bone injury can cause an analogous G₀ to G_alert change in other stem cell types, indicating that this is a generalizable phenomenon^114,117.

3. Interactions among supporting cell types

Source cells for regeneration need cell partners, not only to act as a contingency when they themselves are disabled, but also to produce pro-regenerative cues and for guidance. Every regenerating organ has a supporting cast of tissue-specific cell types that partake in its regeneration. For instance, the wound epidermis of a regenerating limb or fin, the epicardial mesothelium of the heart¹¹⁸, Paneth and other cells of the intestinal crypt¹¹⁹, or stromal cells for hemopoietic stem cells¹²⁰. Nerves, immune cells, vasculature, and fibroblasts are among several cell types that often participate in regeneration -- with common themes emerging from different regeneration contexts (Fig 4A).

Fig 4. Interactions among supporting cell types.

A. For regeneration to supersede scarring, several cell types play a supportive role. Immune cells such as pro-regenerative macrophages, nerve cells and fibroblasts can produce cues for source cells.

B. (Left) Spiny mice, but not house mice, can regenerate ear hole punches due in part to sustained high Erk signaling. (Right) Progenitors located in the mouse dermal fascia generate fibroblasts with pro-scarring, pro-inflammatory profiles that recruit immune cells, while also producing myofibroblasts that promote wound contraction.

C. Antler velvet contains fibroblasts with a pro-regenerative profile that do not recruit immune cells. On the contrary, back skin has pro-scarring fibroblasts with a pro-inflammatory profile.

Immune cells.

Immune cells play a major role in tissue regeneration and exhibit extensive heterogeneity. Prevailing immune cell subpopulations and their kinetics define whether regeneration is favored or inhibited after injury. For example, in zebrafish skeletal muscle, a specific macrophage subset populates injury border zones and secretes the cytokine nicotinamide phosphoribosyltransferase to promote the activity of muscle stem cells¹²¹. Rapid installment of anti-inflammatory macrophages and clearance of neutrophils, collagen, and tempering of the fibrotic response have all been associated with sustained regeneration across tissue types and species^{122,123 124–126}. Tissue-resident macrophages, not recruited from bone marrow monocytes, have been shown to be abundant in the context of a pro-regenerative environment, as for instance the injured neonatal mouse heart. In contrast, an influx of non-tissue resident, circulating macrophages predominates in the poorly regenerating adult mouse heart¹²⁷. Yet, other studies that investigated lung and skin repair, indicate that monocytes/macrophages recruited from the circulation can be pro-regenerative^128,129. It is currently unknown whether a unifying principle can be extracted regarding the role of macrophages in diverse tissue regeneration contexts or whether each tissue context shows distinctive interactions with macrophage subsets.

What attracts pro-regenerative macrophages and in which ways do macrophages encourage tissue regeneration? Macrophage involvement in regeneration is one of the more popular areas of study in the field, and only a fraction of available information can be covered here. Genetic evidence from appendage regeneration in zebrafish and axolotl have implicated Wnt and Midkine factors in directing macrophages to sites of regeneration^126,130. In mouse skin, macrophage recruitment is mediated by lipids released by wound-adjacent adipocytes¹³¹. Evidence in zebrafish indicates that TNF-α secreted by macrophages promotes neurogenesis in radial glial cells through the Tnfrsf1a receptor during spinal cord regeneration in larval zebrafish. In contrast, later-stage macrophages act to decrease levels of IL1B, a cytokine that promotes regeneration^132,133. Other innate immune cell types also modulate regeneration, including innate lymphoid cells and neutrophils. For instance, thymic natural killer T cells (NKTs) restrain regeneration of amputated mouse digit tips, whereas splenic NKTs have a pro-regenerative influence¹³⁴.

Adaptive immunity is increasingly recognized as a contributor to regeneration outcomes. Diverse roles for T_reg cells in the regeneration of multiple tissues have emerged after their initial discovery in regenerating skeletal muscle¹³⁵. These cells have mitogenic roles in regeneration of multiple other tissues including heart, skin, and supporting the intestinal stem cell niche^136–138. A comparison of regenerating organs in zebrafish, including retina, spinal cord, and heart, revealed that T_reg cells show common and organ-specific expression of regeneration-promoting factors, including Neurotrophin3 in the spinal cord, Neuregulin1 in the heart, and Igf1 in the retina¹³⁹. Whether T_reg cells acquire tissue-specific programs upon injury and regeneration, or instead pre-exist in these tissues is not fully understood. Beyond T_reg cells, other types of T cells have also been implicated in local responses to tissue trauma. Wounded skin relies on gamma delta T cells to provide Fgf9 that elicits hair follicle neogenesis in large skin wounds and IL1 to promote hair follicle stem cell proliferation^140,141. Furthermore IL-17A production from Th17 cells, a response that is normally associated with protection against microbiota after skin disruption, was recently implicated in the regeneration of sensory nerves¹⁴².

Nerves.

Salamander limb regeneration depends on innervation. This observation was initially made in the early 19^th century and foreshadowed contemporary findings indicating a contribution of the nervous system in many regeneration contexts¹⁴³. Early studies of limb regeneration found that the quantity of innervation rather than the nerve type (motor or sensory) was important, and trophic factors were proposed to play a dominant role¹⁴⁴. Since then, many proteinaceous “growth factors” that may substitute nerves for blastema formation have been identified, including Neuregulin, FGF2 GDF5, and BMP7, as well as Anterior gradient and transferrin^{145 146,147 148,149}. Many of these factors are thought to act directly on mesenchymal blastema cells to promote their proliferation, although Anterior gradient is proposed to act by relaying nerve signaling to the epidermis¹⁵⁰. Denervation of mouse digit tips by sciatic nerve resection causes defects in regeneration, with growth factors PDGF-AA or OSM able to partially rescue these defects. Innervation has also been associated with maintaining high FGF2 levels that promote cell proliferation^151,152. Experiments employing digit nerve resection, by contrast, indicate no requirement for regeneration after tip amputation¹⁵³. In the mouse skin, nerve-derived Shh ligand encourages hair follicle stem cells in the bulge at telogen stage to contribute to anagen stage hair follicles; surprisingly, Shh also shapes a niche that allows upper bulge cells to migrate into the regenerating epidermis¹⁵⁴. Such signals also derive from supporting Schwann cells that in the regenerating mouse digit tip can be a source of TGF-B, which initiates skin healing, or OSM and PDGF-AA, which promote blastemal cell proliferation^152,155.

Beyond classical growth factors, signaling of sympathetic nerves by catecholamine activation of adrenergic receptors has emerged in several contexts of stem cell activity during regeneration. In mice, ablation of sensory and sympathetic nerves depletes the bone marrow of HSCs, in which B2 and B3 adrenergic signaling from nerves is required to sustain the niche¹⁵⁶. These Leptin receptor-positive (Lepr⁺) stromal niche cells express Nerve growth factor that promotes innervation of the bone marrow, and the nerves in turn activate B adrenergic signaling in the Lepr⁺ cells to sustain the HSC niche¹⁵⁷. Stressful conditions affect hair follicles through increased B adrenergic signaling, driving hyperproliferation, depletion of melanocyte stem cells and hair graying¹⁵⁸. Furthermore, channel activity is implicated in the process of innervation during early stages of zebrafish fin regeneration, as revealed by conditional genetic mutants in the major neuronal voltage-gated sodium channel (VGSC) gene scn8ab¹⁵⁹.

Cells communicate not only as pairs but within complex circuits of cell-cell communication, choreographing the cell behaviors required from different actors. Recently, a pharmacological screen led to a model in which heart regeneration in zebrafish depends on nerve-derived adrenergic signaling, acting through a subset of macrophages that ultimately influence ECM deposition and cardiomyocyte proliferation¹⁶⁰. In another example of cellular coordination, the early interactions of nerves and macrophages with the wound epidermis was proposed to create a niche for blastema cell proliferation and expansion in regenerating axolotl limbs¹³⁰. These types of multicellular relays are now the focus of combined theoretical modeling and experimental approaches that indicate that the sum of multi-cellular interactions may define regeneration outcomes^161,162.

Fibroblasts.

Long considered to be a homogeneous cell type that maintains structural integrity and responds to injury by secreting matrix components, such as collagen, fibroblasts are receiving an increasing amount of attention for their diversity of programs and functions. The classic fibroblast injury response provides a quick fix for major injuries to tissues, such as the skin or the heart. There are obvious benefits to quick-scarring mechanisms that close surface wounds, preserve barrier function, and fill in the gaps of lost myocardium after an ischemic injury. Yet, tissue biomechanics and function are often downgraded after these quick repair mechanisms and the area retains susceptibilities. Animal responses to trauma have thus often been framed as “Regeneration versus Scarring”; and there is truth to this. For instance, zebrafish hearts will heal by scarring if cardiomyogenesis is experimentally blocked, and murine hearts show more effective repair when cardiac fibroblasts are depleted by a CAR-T strategy or a small percentage of them is reprogrammed into myocyte-like cells^163–165. Similarly, pericyte-derived fibroblast-like cells contribute to spinal cord injury scars, and genetically disabling division of pericyte progeny in mice improves repair and connectivity after a cord hemisection injury¹⁶⁶.

Sequencing experiments have made it clear that fibroblasts constitute a heterogeneous population. Like immune cell types, pro-regenerative and anti-regenerative (pro-scarring) resident fibroblasts can help dictate injury outcomes. In mice, two skin fibroblast types with differential fibrotic injury responses exist that once expressed the transcription factor Engrailed 1 (En1) during embryonic development (EN1 lineage-positive fibroblasts (EPFs) and EN1-naiïve fibroblasts (ENFs))¹⁶⁷. EPFs arise at a point in embryonic development in which scarring of skin wounds replaces a more regenerative, scar-free response¹⁶⁸, and they are dominant contributors to scarring in adult skin¹⁶⁷. ENFs normally adjudicate skin development and fetal regenerative responses to injury¹⁶⁸, although a subpopulation of ENFs can induce En1 expression and acquire EPF properties in adult skin wounds¹⁶⁹. Perhaps analogously, Erk signaling has been proposed to toggle regenerative and scarring properties of fibroblasts in mice. Spiny mice (Acomys) have an enhanced capacity to regenerate ear hole punches and large back skin wounds as compared to house mice (Mus)¹⁷⁰. Their healing responses are associated with sustained high Erk signaling¹⁷¹ (Fig 4B). Erk inhibition promotes scarring in Acomys, whereas sustained Erk activation promotes regeneration in Mus, and differentially Erk-responsive fibroblast subsets are suspected to contribute to species differences¹⁷¹.

Pro-regenerative and pro-scarring fibroblast subtypes display distinct immunomodulatory profiles and properties. Progenitors located in the fascia of mouse dermis undergo a sequence of steps to generate fibroblasts with pro-inflammatory profiles that recruit immune cells, such as macrophages and monocytes. These fibroblasts subsequently convert into myofibroblasts that promote wound contraction¹⁷² (Fig 4B). Analogous transitional events happen in injured lungs to generate pro-fibrotic alveolar fibroblasts¹⁷³. Velvet covering the antlers of reindeer – massive bony appendages cast and regenerated each year – regenerate large hole punches without scarring. Antler velvet contains fibroblasts retaining a fetal profile in contrast to immune-activating profiles of fibroblasts from poorly regenerating back skin. Inflammatory cell recruitment by velvet fibroblasts is muted compared to back skin fibroblasts, even when velvet fibroblasts are grafted ectopically onto back skin followed by injury, indicating that fibroblast-immune cell crosstalk measurably determines regenerative capacity¹⁷⁴ (Fig 4C).

4. Size control and tissue integration

Positional, or geometric, memory is a defining feature of regeneration, which brings back what it has lost and only that, weeks to months to years after the initial formation of that lost structure during embryogenesis. These events require mechanisms to retain or instill memory of the correct programs for cell identity in the injured tissue stump, as well as mechanisms to create scale to the correct adult size. As the myriad cellular constituents of tissues return at appropriate proportions, they must be physically and, in some cases, electrically, integrated to regenerate function of that lost organ or appendage.

Epigenetic memory of position.

Trauma unleashes widespread transcriptional changes that ultimately yield to regeneration of order -- cell types return in correct proportion and pattern. This occurs at least in part by recalling epigenetic memories of cell and positional identities that template the pattern and determine the spectrum of cell identities that are produced during regeneration. It is likely that processes to establish and follow epigenetic settings for cell lineage and positional states help regenerating tissues steer clear of cancer programs.

Salamander limb regeneration relies on stably sustained cellular position codes that exist in each limb segment (upper limb, lower limb, hand/foot) as well as anterior-posterior (thumb-pinky) and dorsal-ventral (palm-back of hand)¹⁷⁵. Here, fibroblasts maintain stable identities according to proximodistal limb region and radial location around the limb, in this way guiding regeneration of the correct structures. Fibroblasts within different limb segments launch distinct homeodomain expression programs that constrain them to regenerate their own segment and those more distal but not more proximal seqments. This positional constraint arises at least in part from the repressive histone mark H3K27me3 that differentially occupies homeodomain protein genes in fibroblasts from the different limb segments. In the upper arm, the hand-associated Hox gene, such as HoxA13 is occupied by H3K27me3 marks and repressed, so that an upper arm gene like Meis fires early during upper arm regeneration. In contrast, in the hand, HoxA13 has an underoccupied H3K27me3 signature and has corresponding elevated gene expression during hand regeneration ^176,177. In essence, limb amputation prompts the resultant blastema to re-initiate the limb development program at the segment level of the amputation.

Radially arranged cell identities are exploited by the limb to help ensure that regeneration occurs only upon amputation and not after simple wounding, by establishing a co-dependent signaling relationship between cells derived from the anterior and posterior limb¹⁷⁸. Posterior fibroblasts, as they transform into blastema cells, recreate a key signaling center -- localized Shh expression -- that signals to anterior limb progenitor cells to promote growth and patterning. The epigenetic basis of this memory of posterior limb cells to induce Shh relies on a sustained transcriptional circuit in adult tissue (Fig 5A). Expression of the posterior transcription factor Hand2 persists in the mature limb tissue in axolotl, as well as in mature pectoral fins of zebrafish, which also maintains patterned domains of the anterior transcription factor alx4a and other genes^179,180. Hand2 is necessary and sufficient to produce posterior blastema cells that initiate Shh gene expression (Fig 5A). A similar system exists long the D/V axis involving expression of the transcription factor Lmx1b¹⁸¹.

Fig 5. Size control and tissue integration.

A. Epigenetic memory of position is represented at least in part from the repressive histone mark H3K27me3 in fibroblasts of different limb segments. Posterior fibroblasts of the salamander blastema express the transcription factor Hand2 that initiates SHH signaling to progenitor cells located in the anterior limb.

B. (Left) Wnt and Erk signaling across the body axis of planaria. After animal bissection, the Wnt signaling gradient rapidly re-scales, re-directing neoblasts to form proportionally scaled organs. (Right) Regenerated cardiomyocytes integrate with existing heart muscle in zebrafish, involving maturation and upregulation of Lrrc10 after proliferation.

Positional memory also exists in mice and humans, and transcriptional profiling of fibroblasts across the body has identified positional codes of gene expression that appear to define regions, even after culturing in vitro. Similarly, HOXA11 is specifically expressed in lower arm but not upper arm fibroblasts of adult mice, and lineage and functional studies uncovered a role for this spatially restricted transcription factor in fracture repair ^182–184. A current unknown in cross species comparisons is whether mammalian stem cells can reactivate the sequential expression of Hox genes, as observed in regenerating salamander limbs¹⁷⁶.

Functional integration of regenerated tissue.

Regeneration outmatches current bioengineered devices in its return of function. For instance, recreating the sensory ability of an arm and its digits is a major challenge for prosthetic device. Tissue matchmaking, to form a seamless interface between old and new tissue, helps regeneration reach its functional goals. For example, if the salamander limb is amputated in the humerus, how does the remaining humerus part connect to its new skeletal elements? Riquelme-Guzmán and colleagues found that continuity between the old bone with regenerated bone relies on osteoclasts to trim the bone at the amputation plane¹⁸⁵. During limb regeneration, an intermediate mini-limb is formed and expands to match the size of the animal. Thus, embryonic-like fields of cells have been specified early to make tissues representing limb regions, whereas cells within each tissue compartment continue to proliferate and promote tissue expansion. By denervating limbs at late stages of limb regeneration, a study indicated that nerve input is required for the late, expansion phase of growth¹⁸⁶. The authors proposed that the amount of nervous tissue might be a yardstick by which proliferative cells within limb regions hit their target sizes¹⁸⁶.

For the heart to effectively regenerate lost muscle, electrical and mechanical coupling of newly formed cardiomyocytes with existing heart tissue is critical. Relevant to this, induced pluripotent stem cell (iPSC)-derived cardiomyocytes that possess fetal morphology and physiology survive (Fig 5B), integrate, and function when introduced into hosts as cell suspensions or patches, yet they may stay immature and can cause arrhythmogenesis^187,188. When innate regeneration through dedifferentiation and proliferation of endogenous cardiomyocytes occurs, as in zebrafish, cardiomyocytes at the injury border zone first show embryonic-type physiology as in the nature of calcium ion fluxes¹⁸⁹ (Fig 5B). Upregulation of the leucine rich repeat protein (Lrrc10) is required to help cardiomyocytes mature away from the embryonic phenotype (Fig 5B). This maturation occurs by ensuring the structural integrity of the nascent cardiac dyad and modulation of calcium ion dynamics. Lrrc10, first identified in zebrafish heart regeneration¹⁹⁰, also induces structural maturation in human iPSC-derived cardiomyocytes¹⁸⁹. Thus, cardiomyocyte regeneration requires appropriate redifferentation as well as dedifferentiation^189,191, and this knowledge might guide therapeutic applications.

Scaling.

Achieving size during regeneration can occur in multiple ways. Salamander limbs form a blastema that gradually differentiates into the appropriate limb structures, whereas zebrafish fins maintain a blastemal growth zone at their distal tips and add bone segment upon segment until length is achieved. Paradoxically, the crowned king of mammalian regeneration, the liver, reconstitutes an organ-to-body mass ratio from intact lobules after partial hepatectomy, but does not reconstruct the missing lobule from a primordium. Regeneration of liver mass relies at least in part on systemic FGF15 secreted from the gut, that acts through the Klotho receptor, along with liver-derived RSPO that coordinates Wnt and other signaling pathways^192,193. At this tissue level, the liver is zonated with hepatocytes expressing different enzymes depending on their distance along the portal-to-central vein axis, the unit structure of the liver. Transcriptome sequencing and genetic fate-mapping has revealed differences in contributions to regeneration from the different zones of hepatocytes, based on these locations with the lobes^194–196. This organization is faithfully recapitulated after injury, and depends on LGR4 and LGR5 signaling, whereby the expression of these receptors is stratified along the zones. This signaling helps to control the expansion or contraction of different hepatocyte populations to attain an appropriate proportionality^{196 193}. Interestingly, somatic mutations in growth-promoting genes acquired during development can influence and even increase the regenerative capacity of hepatocyte clones in liver disease¹⁹⁷.

Proportional scaling with faithful spatial patterning is a hallmark of whole-body regeneration in planarians. Flatworms grow and shrink according to nutrient availability, and, after bissection, rescale their anatomy although the regenerating bodies are a fraction of their previous size. Sustained positional transcriptional states are important, in which body wall muscle cells display a positional code of signaling molecules, including regulators of Wnt and BMP signaling pathways, that guide neoblasts in building position-appropriate body parts^198,199. In planarians, Wnt and ERK signaling are graded across the body axis to control neoblast differentiation and determine the head-to-tail axis and its proportionality^200–203. After animal bissection, the Wnt signaling gradient rapidly re-scales, re-directing neoblasts in the severed body to form organs appropriate to the new scaling. Thus, a phoenix-like transformation of body part into whole organism occurs (Fig 5B). How Wnt signaling gradients span millimeter distances is incompletely understood. In particular, whether true gradients of single morphogens exist across such distances or whether these are relay mechanisms between cells is not clear. Dynamic signaling patterns are also represented in regenerating zebrafish scales, in which live imaging of ERK kinase activity revealed a timed series of signaling waves in osteoblasts, traveling millimeter distances in a ringed pattern from the center out to the periphery of the scale²⁰⁴. Each time a wave hits a region of osteoblasts, they see a transient burst of growth. How spatially regulated upstream and downstream players are controlled and sculpt tissue is central to regeneration and relevant to engineering approaches in adult-scale regenerating organs.

Knowing when to stop is important not only when writing literature reviews, but also during regeneration. Several zebrafish mutants that flaunt lengthened fins have shed light on this important issue. Two of the underlying mutations affect the activity of components of potassium leak channels, Kcnk5b and Slc12a7a/Kcc4a, and they cause fins to grow longer with lengthier bony segments, both ontogenetically and during regeneration^205,206. The activities of these proteins are linked to signaling through regulation of Kcnk5b phosphorylation by the phosphatase calcineurin, affecting channel conductance^207–209. Within this signaling network, another potassium channel, Kcnh2a, acts upstream of calcineurin to control cessation of fin growth²¹⁰. Further examination of developmental gene expression in channel mutants revealed upregulation of key signaling factors and a tissue phenotype of a larger blastema and faster skeletal growth²⁰⁸. This phenotype is reminiscent of the effects of agents that influence bioelectric signaling and elicit ectopic regenerates in amputated Xenopus limbs²¹¹. Together, these observations suggest that cellular ionic setpoints interface with signaling pathways and developmental gene regulatory networks to manage regeneration of entire structures.

Perspectives

Each hallmark mentioned here has knowledge gaps to be filled by discovery research and technological advances. A pipeline of more advanced animal and plant models of regeneration will help identify answers as well as new questions. Although single-cell RNA sequencing has increased the cellular resolution of regenerative events, it does not obviate the need for direct visualization of cellular and subcellular events in live, regenerating tissue. There is an infinite space for the generation of new tools and the combination of these tools with advanced imaging platforms. Importantly, a growing number of studies of developing tissue integrate mechanical cues with molecular signals and cellular behaviors. Mechanical signals are also implicated in regeneration, for example by physically enabling or inhibiting cell division, or by controlling factor release or localization^212–214. Advancing methods to spatiotemporally control mechanical influences in injured and regenerating tissues will enable a full appreciation of their impact.

The most common question regenerative biologists receive is “Why?” -- why do certain species and tissues have great regenerative capacity when others are so limited? It is challenging to directly address this topic with rigorously tested hypotheses; one is more likely to accidentally bump into illuminating evidence by exploring how regeneration works. An important arm of trying to address the evolution of regenerative capacity is to understand how enhancers, silencers, insulators or other non-coding DNA elements became associated with protein coding components used earlier in initial embryonic development.

A bright future lies ahead in harvesting the growing body of conceptual and molecular information to engineer regeneration-competent cell states and tissue environments (Fig 6). There are many markers on the road to that destination. Although it is recognized that influential aspects of regenerative biology, such as inflammation and signal presence, can be modulated, the ability to alter intrinsic epigenetic states to access memories of developmental competence is a roadblock that must be overcome²¹⁵ (Fig. 5). A better understanding of the factors that allow controlled transdifferentiation in certain cell types and contexts, but not in others, will aid in this pursuit. Modulation itself needs precision for safety and efficacy, and this precision can be enforced by incorporating regeneration enhancer elements to dose and target pro-regenerative factors, for instance within gene therapy vectors²¹⁶. Building synthetic circuits for regeneration based on engineered ligand systems can also infuse elements of control and amplitude. Finally, highly regenerative organisms remind us that stem cell location within a regenerating primordium can impact the competency to express key morphogens and other signaling molecules. Directed spatial targeting of competent stem cells may be most helpful in future applications of regenerative medicine.

Fig 6. Representation of hallmarks of regeneration in proficient and inefficient models, and potential interventions at bottlenecks.

Left: Key cellular and signaling events occur during limb regeneration in the salamander. Middle: Human limb injury leads to fibrosis without restoration of tissue patterns. Right: Manipulation of gene expression states with modern tools such as CRISPR-mediated transcriptional activation (CRISPRa), immunomodulation, and restitution of developmental genetic circuits can help unleash a regenerative response.