The role of p21 in regulating mammalian regeneration

The MRL (Murphy Roths Large) mouse has provided a unique model of adult mammalian regeneration as multiple tissues show this important phenotype. Furthermore, the healing employs a blastema-like structure similar to that seen in amphibian regenerating tissue. Cells from the MRL mouse display DNA damage, cell cycle G2/M arrest, and a reduced level of p21CIP1/WAF. A functional role for p21 was confirmed when tissue injury in an adult p21-/- mouse showed a healing phenotype that matched the MRL mouse, with the replacement of tissues, including cartilage, and with hair follicle formation and a lack of scarring. Since the major canonical function of p21 is part of the p53/p21 axis, we explored the consequences of p53 deletion. A regenerative response was not seen in a p53-/- mouse and the elimination of p53 from the MRL background had no negative effect on the regeneration of the MRL.p53-/- mouse. An exploration of other knockout mice to identify p21-dependent, p53-independent regulatory pathways involved in the regenerative response revealed another significant finding showing that elimination of transforming growth factor-β1 displayed a healing response as well. These results are discussed in terms of their effect on senescence and differentiation.


A mammalian model of regeneration, the MRL mouse
In 1998, the MRL (Murphy Roths Large) mouse, generated from cross-breeding AKR, C3H, C57BL/6(B6), and LG strains of mice [1], was shown to be able to close ear punches without showing residual signs of injury or scarring [2]. Multiple tissues were perfectly replaced, cartilage re-grew, and hair follicles reappeared. Furthermore, this type of perfect multi-tissue healing, known as epimorphic regeneration, occurred with the formation of a blastema-like structure that had been shown to be key to amphibian limb regeneration [3][4][5]. Th is phenomenon had earlier been seen in rabbit ear holes [6][7][8], and furthermore, a blastema-derived structure had also been described during antler re-growth [9]. Th e amphibian and mammalian ear hole regeneration processes have many features in common, including rapid reepithelialization of the wound [2], elimination of the basement membrane between the epidermal and dermal tissue layers [10,11], blastema formation, re-growth of cartilage and hair follicles, and scarless healing [2,12,13]. However, the existence of an inbred mouse model allowed this process to be genetically approachable. It was also determined that one of the strains used to generate the MRL mouse, the LG/J mouse, contributed the regeneration phenotype [14].
Ear hole closure has lent itself exceedingly well to genetic studies as this is a wound that is easy to access and measure and has proven to be a highly quantitative trait [15][16][17]. Recently, making use of an advanced intercross line (LG, SM F34 AIL) employing 1,200 mice and 3,600 single nucleotide polymorphisms [18], 18 quantitative trait loci were identifi ed for ear hole closure with small intervals from 0.661 to 7.141 Mb in length, which essentially reduced the healing intervals 10-to 50-fold from studies using F2 mice [15] (JM Cheverud et al., manuscript in preparation). Th is has allowed a more focused analysis of candidate genes. Further narrowing of

Abstract
The MRL (Murphy Roths Large) mouse has provided a unique model of adult mammalian regeneration as multiple tissues show this important phenotype. Furthermore, the healing employs a blastema-like structure similar to that seen in amphibian regenerating tissue. Cells from the MRL mouse display DNA damage, cell cycle G2/M arrest, and a reduced level of p21 CIP1/WAF . A functional role for p21 was confi rmed when tissue injury in an adult p21 -/mouse showed a healing phenotype that matched the MRL mouse, with the replacement of tissues, including cartilage, and with hair follicle formation and a lack of scarring. Since the major canonical function of p21 is part of the p53/p21 axis, we explored the consequences of p53 deletion. A regenerative response was not seen in a p53 -/mouse and the elimination of p53 from the MRL background had no negative eff ect on the regeneration of the MRL.p53 -/mouse. An exploration of other knockout mice to identify p21-dependent, p53-independent regulatory pathways involved in the regenerative response revealed another signifi cant fi nding showing that elimination of transforming growth factor-β1 displayed a healing response as well. These results are discussed in terms of their eff ect on senescence and diff erentiation. these loci and testing of candidates using gene knockouts should lead to the fi nal identifi cation of these genes.
Besides ear hole closure, multiple organ and injury systems have extended the MRL mouse's unusual healing properties. Th ey include regenerative studies in the heart [19][20][21], central nervous system stem cells and tissue [22][23][24], cartilage [25], cornea [26], digit [27,28] and myometrial healing [29]. Dorsal skin wound healing, which involves skin contracture, has been reported to be no diff erent or even worse in the MRL compared to controls [30,31]. However, a recent study shows that if the wound has a syngeneic or allogenic skin transplant, the MRL shows far better healing than the control [32]. One possible explanation for the healing diff erences in diff erent systems is that wound contracture, involving myofi broblasts or cells expressing Sma-1 (smooth muscle actin), known to be responsible for scarring, is diff erent in the MRL. Preliminary studies suggest this [33] (D Gourevitch, K Bedelbaeva, unpublished data). Th us, the wound site and type of wound need to be considered in the MRL's healing properties.

G2/M cell cycle accumulation of regenerating cells
Th e cells derived from the ear of regenerating and nonregenerating mice also show signifi cant diff erences from each other and represent what is seen in vivo. MRL fi broblast-like cells from uninjured ears display an uncommon metabolic profi le characteristic of an embryonic-type aerobic glycolysis, a feature of the adult MRL mouse itself, versus the more common metabolic state -oxidative phosphorylation -as seen in the B6 mouse [34]. Th ese cells express stem cell markers similar to adult MRL tissue that expresses these markers [34]. In a separate study, cells derived from the injured MRL ear blastema expressed stem cell markers as found in vivo [35] and displayed highly proliferative and migratory responses in vitro similar to human multipotential progenitor cells in this study [36].
Th e rapid growth rate of fi broblast-like cells from the uninjured MRL ear was noted early on and examination of cell cycle regulation comparing healer MRL to nonhealer B6 cells showed that the healer cells had an unusual accumulation of cells in G2/M [33]. A likely explanation of such G2/M accumulation or potential arrest was a DNA damage response and this was supported by an increased p53 response in the MRL [33] and confi rmed with data showing that foci of γH2AX and TopBP1, a phosphorylated histone and a protein recruited to sites of DNA damage, respectively, were highly increased in MRL cells and tissue [33]. DNA damage itself was tested using the comet assay and found in nearly 90% of healer cells compared to 5% of non-healer cells, showing both single-strand and double-strand breaks. Furthermore, the DNA repair protein RAD51 was increased in healer cells, suggesting that error-free homologous recombination was being used [33]. Th e cause of the DNA damage is still unclear, but the lack of the cell cycle protein p21 Cip1/Waf1 discussed below suggests a replicative stress mechanism.
Th ese results agree with many reports in the literature that G2/M accumulation is associated with regeneration in examples ranging from hydra [37] to amphibian [38] to mammalian liver [39,40]. Th e literature also shows that cells undergoing blastema formation synthesize DNA but have a low mitotic index, indicating an accumulation between S and M and implicating G2 [41][42][43][44][45][46][47]. Multiple in vitro studies have carefully explored cell cycle arrest and the factors involved in the re-entry of cells into S phase of the cell cycle and accumulation in G2, as seen in multinucleated muscle myotubes and myofi bers from regenerating amphibian limbs [48], in multinucleated mammalian myotubes generated from rat C2C12 cell line myoblasts, and in primary mouse myoblasts [49][50][51].
In MRL ear-derived cells, the fact that DNA damage was so widespread made one question why an accumulation of cells was seen in G2/M and not in G1/S. Th is led to an examination of G1 cell cycle regulatory proteins. Th e fi rst to be examined, the CDKN1A or p21 Cip1/Waf1 protein [52], was found to be repressed in these cultured cells. Examination of similar ear-derived cells from a CDKN1A-defi cient mouse [33] showed the same phenotype as MRL cells with increased DNA damage, γH2AX expression, and G2/M accumulation. But most striking was the fact that this mouse could fully close earhole injuries at least as well as the MRL mouse [33]. Th ere have been other mice that possess the ability to partially heal ear holes, including nude mice [53], mice expressing the transgene AGF (angiopoietin-related growth factor) in keratinocytes [54], and mice selected for infl ammatory potential [55]. However, what was surprising to us was that deletion of this single gene, as predicted from our in vitro ear dermal cell model, could actually result in the full MRL epimorphic regeneration phenotype.

The role of p21 CIP1/Waf1 , regeneration, and the retinoblastoma protein
Earlier studies have examined the role of p21 in regeneration of the mammalian liver. Gene expression of p21 plays a role in hepatic regeneration by both p53dependent and p53-independent control mechanisms [56]. Transgenic mice that over-express p21 produced large polyploid nuclei in a portion of the hepatocytes and the regenerative capacity of the livers was halted [57]. Over-expression of STAT-3 with resulting p21 upregulation impairs regeneration in fatty livers [58]. Consistent with this picture, repression of the p53/p21 pathway was shown to enhance liver regeneration [59]. Such studies parallel our recent fi ndings [33].
Th e overall understanding of the functions of p21 can be quite overwhelming considering the complexity of functions in which this protein has been implicated. p21 is involved in the response to cellular stresses, such as DNA damage, oxidative stress, cytokines, mitogens, tumor viruses, and anti-cancer agents, and can have tumor suppressive activities and oncogenic capabilities depending on the cell type and context [60,61]. For example, p21 is transcriptionally regulated by p53 for tumor suppressor activity and as an inhibitor of cell cycle progression through the inhibition of cyclin-dependent kinase (CDK)cyclin complexes and proliferating cell nuclear antigen, which can lead to diff erentiation, apop tosis, or senescence. Increasing this complexity is the fact that p21 can regulate gene expression and other cellular events, such as autophagy and a DNA damage repair response, through protein-protein interactions that depend on the cell type, subcellular localization, expres sion levels, protein stability, and post-translational modifi cations [62][63][64][65][66].
So which of these functions are involved in the regenera tion phenotype seen in the p21 -/mice? Some indication may come from in vitro studies in other regenerating systems. For example, adult urodele amphibians can regenerate limbs through a process that involves loss of diff erentiation markers, cell cycle reentry, proliferation, formation of a blastema, and diff eren tiation into adult tissue [12]. In an amphibian in vitro model of skeletal muscle regeneration, retinoblastoma (Rb) protein plays a predominant role in cell cycle reentry through phosphorylation by CDK4/6 [67]. Th is process requires serum to stimulate entry of the quiescent nuclei of multinuclear myotubes into S-phase with a serum-derived thrombin-activated factor being necessary for Rb hyperphosphorylation, resulting in its 'inactivation' [48,68]. Th ese cells enter S phase but arrest and do not separate into single cells, which would allow further progression of the cell cycle through mitosis. However, there are confl icting reports about mammalian cells. Myotubes from an Rb -/mouse are capable of cell cycle re-entry and show DNA synthesis upon serum stimulation but no mitosis in one study [50] but no cell cycle re-entry in another [51]. In a separate study using mammalian myotubes generated from the rat C2C12 myoblast line, newt regeneration blastema extract led to myotube cellularization to smaller myotubes and proliferating mononucleate cells, suggesting de-diff erentiation with reduced expression of mature muscle cell markers [49]. In addition, a recent report using primary myoblasts [69] suggests that another factor in addition to Rb, p19 arf , must be inactivated for cell cycle re-entry and de-diff erentiation in postmitotic mammalian muscle. Th e tumor suppressor protein p19 arf acts as a regeneration suppressor and is not found in regenerative vertebrates, suggesting that it has interesting potential as a key to mammalian regeneration. Th us, Rb inactivation has been shown to be important in both amphibian and mammalian regeneration in vitro.
Th e p21 protein, its major role being a CDK inhibitor found on chromosome 17 in the mouse, is known to block proliferation by preventing the phosphorylation of Rb and the transcription of cell cycle-regulated proproliferative proteins. Th e p21 protein binds to cyclin-CDK (2/4) complexes, not allowing them to function as kinases. Th ey in turn cannot phosphorylate Rb, which remains bound to E2F, a transcription factor responsible for proliferation, eff ectively blocking E2F function. Th us, p21 activity directly leads to suppression of cell cycle transit and the loss of p21 should promote E2F activity, lead to enhanced DNA synthesis and potentially to dediff erentiation. Rb function, then, in the studies above should be directly aff ected by p21 activity.
Not surprisingly, p53 and p21 have been shown to prevent the transition from fi broblasts to induced pluripotent stem cells [70][71][72]. Th e level of de-diff erentiation in the p21 -/mouse is being further explored, although we have previously reported that stem cell markers are over-expressed in MRL tissue [34].

The role of p53, senescence, and transforming growth factor-β in regeneration
As mentioned above, we found that p53 was up-regulated in MRL mouse ears, though p21 was absent. Is there a role for p53 in regeneration? Unlike the p21 -/mouse, which is a complete regenerator, p53 -/mice show no regenerative capacity [73]. Th is fi nding established a p53independent function of p21 that is important for regeneration. However, MRL.p53 -/crosses showed not only healing rates similar to or better than the MRL itself but also showed enhanced diff erentiation in the form of increased chondrogenesis and adipogenesis [73]. Th e major role played by p53 as the 'guardian' of the genome is due to its ability to respond to DNA damage and cellular stress by inhibiting cell cycle progression and then regulating DNA repair, cell cycle control, apoptosis, diff erentiation, autophagy induction, and senescence. It is not clear which of these functions or lack thereof could be responsible for the enhanced diff erentiation observed in MRL.p53 -/mice [64,71,[74][75][76][77][78][79]. One study suggests that removal of p53 allows for an accumulation of cells with elevated levels of DNA damage (on a repair-defi cient background mouse), which delays hair follicle renewal and regeneration [80,81]. However, we observed hair follicle formation in our MRL/p53 -/mice [73]. Further regeneration studies on diff erent tissue types need to be performed in order to determine the role of p53 in regeneration.
One potential area of interest are the roles of p21 and p53 in both diff erentiation and cellular senescence at wound sites. It has been shown that elimination of p21 in mouse stem cells with dysfunctional telomeres, a marker for senescence induction, increases stem cell function and the life span of these mice without an increase in cancer formation, providing a direct role for p21 in both stem cell diff erentiation and senescence [82]. One direct link for p21 in diff erentiation and senescence is sup pression by the Twist proteins, major regulators of embryogenesis [83]. Th e Twist proteins inhibit p21 in a p53independent manner and promote epithelial-mesenchymal transition and suppress cellular senescence [84].
Th e two major pathways for inducing senescence in cells of multiple tissues are p53/p21 [85][86][87][88][89][90][91] and p16 ink4a [75,[92][93][94][95]. In an earlier paper, we suggested that senescence was not a factor in MRL regeneration because of the lack of p53 requirement [73]. However, there is, in fact, evidence that p21 can induce senescence in the absence of p53 [87,[96][97][98] as well as p53-mediated p21independent activation of senescence [99][100][101]. It has been suggested that reactive oxygen species are necessary to maintain the senescence phenotype and that both p16 and p21 are involved [99,102,103]. Actually, we previously reported that reactive oxygen species levels are decreased in the MRL mouse [34], consistent with an aerobic glycolytic metabolism, which argues against senescence playing a functional role. In addition, the protein RhoD, which is required for transformation by the oncogenic protein Ras, is responsible for suppressing p21 induction and subsequent senescence [104,105]. Th e gene ID1 has been shown to repress HRAS-mediated senescence in the presence of increased amounts of p21 [106], arguing the other way. Recently, a publication showed that the matricellular protein CCN1, which is expressed at the sites of wounds, induces senescence through p53 and actually helps to prevent fi brosis during tissue repair [107]. In this case, however, the healing is tissue repair with scarring and not blastema-induced scarless regenera tion. Th us, the connection between senescence and regeneration, and its diff erence compared to oncogenesis, is yet to be determined.
Another major regulator of p21 is transforming growth factor (TGF)-β1, which is involved in anti-proliferation and diff erentiation [108]. TGF-β1 controls proliferation, diff erentiation, migration, and apoptosis in embryonic and adult tissue through the Smad3 pathway [109][110][111][112][113]. Multiple studies in mutant mice lacking the TGF-β1/ Smad3 pathway have implicated a regeneration phenotype in mice: mice lacking TGF-β1 show an increase in wound closure and epithelialization [114]; transgenic mice null for Smad3 show increased re-epithelialization and tissue renewal [115]; and Smad7 over-expression leads to Smad3 down-regulation and to enhanced liver regeneration through the TGF-β/Smad3/p21 pathway [116]. Smad3 has been implicated as a candidate gene in our genetic mapping studies of healer MRL and parental LG mice [15]. Contrary to these results, other transgenic studies on TGF-β1-null mice showed malfunctions in the repair of excisional back skin wounds due to altered infl ammatory responses [117][118][119]. Our studies have shown that a TGF-β1/Rag1 double knockout mouse is a partial healer [73]. An interesting fact is that TGF-β1 enhances Sma-1 production and myofi broblasts associated with scarring [120] and reduces regenerative healing, whereas the TGF-β isoform TGF-β3 enhances scar-free healing [121].

Conclusions
Th e MRL mouse is the fi rst genetically dissectible and molecularly tractable mammalian model of regeneration of multiple tissues in a single organism. It establishes the fact that regenerative capacity has not been lost to mammals through evolution but remains as a cryptic trait, which can be activated by the deletion of a single gene, p21. Th us, the p21-null mouse now should become a 'single gene' standard model for mammalian regenerative studies.
Th e lack of p21 may act to enhance the regenerative response in various ways. It could alter DNA damage and checkpoint responses, leading to enhanced proliferation. It could reduce TGF-β signaling, leading to reduced scar formation, and alter diff erentiation patterns. It could lead to lack of senescence and reduced cytokine responses. It could support progenitor cell stability as seen in induced pluripotent stem cell formation.
Besides determining exactly which function of p21 and its absence is responsible for enhanced ear hole closure, it will also be important to defi ne the critical pathways in the MRL mouse that actually lead to p21 down-regulation and regeneration.

Competing interests
The authors declare that they have no competing interests.