An Unexpected Friend – ROS in Apoptosis-induced Compensatory Proliferation: implications for regeneration and cancer

Abstract

Apoptosis-induced compensatory proliferation (AiP) is a form of compensatory proliferation that is triggered by apoptotic cell death to maintain tissue homeostasis. As such, AiP is essential for many tissue repair processes including regeneration. The apoptotic effectors, termed caspases, not only execute apoptosis, but are also directly involved in the generation of the signals required for AiP. Reactive oxygen species (ROS) play an important role for regenerative processes. Recently, it was shown in Drosophila that apoptotic caspases can mediate the generation of ROS for promoting AiP. This review summarizes and discusses these findings in the context of regenerative processes and cancer.

Introduction

Regeneration is a dynamic biological process that is aimed at restoring tissue integrity and maintaining homeostasis following injury. Injury is caused by wounding due to trauma, or pathologies like infection and cancer (1). Regeneration requires a coordinated series of various intracellular, extracellular and intercellular signaling events. Each of these events needs to be tightly regulated to achieve homeostasis and to prevent excessive healing responses. Defects in wound repair can cause non-healing ulcers, or excessive invasive proliferation leading to cancer. Indeed, a classical theory by Harold Dvorak postulates that “tumors are wounds that do not heal” (2). This theory was proposed based on observations that both wounds and tumors share common cellular properties. For example, the composition of the tumor stroma resembled the granulation tissue observed during wound healing. The major difference between wounds and tumors is that in contrast to wound healing, the process is not self-limiting in tumors, and continuous activation of the involved signaling events causes hyperproliferation, invasion, and finally metastasis.

The ability to repair tissue after an injury is a fundamental property of multicellular organisms, though the regenerative responses are quite diverse, and depend on the species, organ and the developmental stage (3). Most species have the capacity to regenerate missing body parts or organs, either completely or with formation of a scar, while a few species like primitive sponges, Hydra, and planarians can regenerate the entire organism from parts of their body (4). Some animals are capable of regenerating their tissue throughout life, while others show a developmental stage-specific restriction of the ability to regenerate. While most mature mammals maintain the ability to regenerate a few select organs such as the liver, they have largely lost the capacity to regenerate lost tissue.

Regeneration follows a succession of events including an immediate wound healing response, followed by formation of a regenerative structure called blastema, leading to proliferation and differentiation, and finally complete restoration of the lost tissue or appendage. Wound healing, the first stage of regeneration, is initiated immediately after an injury, and involves recruitment of immune cells like neutrophils and macrophages, inflammation, and formation of a clot. Actomyosin cables are extended across the wound edge, the extracellular matrix is remodeled, and the wound is closed. Formation of a blastema is a critical step in regeneration, and usually occurs prior to complete wound closure (5). The blastema is the site for regenerative proliferation, and consists of both differentiated cells and stem cells. The cells in the blastema give rise to new cells by multiple different mechanisms including stem cell proliferation, compensatory proliferation, cell cycle re-entry of differentiated cells, dedifferentiation, or transdifferentiation (3). Finally there is remodeling of the newly formed cells to establish epithelial integrity, which restores a fully patterned organ or limb.

Wound healing without blastema formation causes partial restoration of lost tissue, and gives rise to fibrotic scars. Healing by fibrotic scarring rather than by regeneration leads to tissue dysfunction, and can place a huge burden for the health of the animal (4). Therefore, understanding the molecular and cellular basis of regenerative growth has been an area of intense investigation during the past decades. Mammals show restrictive regeneration, and for this reason several different model organisms that show a high degree of regenerative capacity, like hydra, planaria, Drosophila, Xenopus, salamander or zebrafish, are being used with the goal to identify the mechanisms that underlie injury-induced cellular plasticity and regeneration (4). Studying regeneration in various model organisms can provide insight into why mammals have lost the capacity to regenerate, and whether it is possible to restore regeneration in mammals, forming the basis for regenerative medicine.

Apoptosis-induced Compensatory Proliferation (AiP)

Compensatory proliferation is one of the mechanisms by which a regenerative response is initiated post injury. In this mechanism, the uninjured cells increase proliferation to replace the damaged cells, and thus maintain tissue homeostasis (6) (Figure 1A). Studies in several model organisms have revealed that apoptotic cells are one of the driving forces for compensatory proliferation through secretion of mitogenic signals, hence this phenomenon was termed “apoptosis-induced compensatory proliferation” (AiP) (7, 8). Active caspases are the main inducers of AiP. Caspases are conserved cysteine proteases present in cells as inactive zymogens. They are divided into initiator caspases and effector caspases based on the length of their prodomains and their activation process. Initiator caspases like Caspase−2, −8, −9, −10, and Drosophila Dronc have long prodomains, and are activated by incorporation into large apoptotic protein complexes called the apoptosome (Figure 2A,B). Effector caspases like Caspase−3, −6, −7, and Drosophila DrICE and Dcp−1 are characterized by short prodomains, and are activated upon cleavage by upstream initiator caspases (9). After apoptosis induction, caspases are activated in a sequential cascade that culminates in the death of the cell (Figure 2A,B). Recent studies have shown that caspases also function independently of their role in apoptosis, and are involved in inflammation and immunity, cell migration, neurite pruning, cellular remodeling and differentiation, as well as AiP (reviewed in (10, 11)).

Figure 1. Genuine and Undead Models of Apoptosis-induced Proliferation (AiP).

(A) During genuine AiP, before they die, apoptotic cells (blue) emit signals (yellow arrows) to neighboring surviving cells (brown) to promote compensatory proliferation and maintain tissue homeostasis.

(B) Undead cells (red) emit proliferative signals (yellow arrows) to neighboring normal cells (brown). Because undead cells do not die, they continue to signal, triggering overgrowth. Undead cells themselves also contribute to the overgrowth.

(C) Examples of a normal (control) adult fly head (left) and an undead adult fly head (right). The head capsule is overgrown and displays patterning errors such as additional bristles and ocelli. This is indicated by the yellow double arrow. Genotypes: ey-Gal4 UAS-p35 (control) and ey-Gal4 UAS-hid UAS-p35 (ey>hid-p35; undead).

Figure 2. Apoptosis Pathways in Mammals and Drosophila and the undead AiP Model in Drosophila.

(A) The mammalian apoptosis pathway downstream of mitochondria. Mitochondria release IAP-antagonists and Cytochrome c (Cyt c) which trigger caspase activation and apoptosis. The apoptosome is composed of Apaf−1, Cyt c and Caspase−9. Activated Caspase−3 also activates calcium-independent Phospholipase A2 (iPLA2) which generates Prostaglandin E2 (PGE2) for AiP.

(B) The Drosophila apoptosis pathway. The IAP antagonists Reaper, Hid and Grim are localized at mitochondria and trigger apoptosis from there. The Drosophila apoptosome is composed of Dronc and Dark. Cyt c is not involved in Drosophila apoptosis. Homologous proteins to the mammalian apoptosis pathway have the same color as in (A).

(C) The undead AiP pathway in Drosophila. Expression of P35 inhibits the effector caspases DrICE and Dcp−1, thus blocking apoptosis. Dronc is not affected by P35 and can activate Jun N-terminal kinase (JNK) signaling which releases the mitogens Wingless (Wg), Decapentaplegic (Dpp) and Spitz (Spi) for AiP.

AiP initiated by caspases is seen in a variety of model organisms like Hydra, planarians, newts, Drosophila, Xenopus and mice (12). In Hydra, activation of effector caspases in apoptotic cells after mid-gastric amputation triggers release of the mitogen Wnt3, which stimulates proliferation and head regeneration (13). Similarly, in Xenopus tadpoles, tail amputation induces caspase activity at the site of injury, which is essential for regeneration (14). In mice, in an AiP-dependent process termed “Phoenix Rising”, caspases are important for epithelial wound healing, liver regeneration after partial hepatectomy, and tumor repopulation following chemotherapy or radiation. During Phoenix Rising, Caspase−3-induced activation of calcium-independent Phospholipase A2 (iPLA2) and prostaglandin E2 (PGE2) is required for compensatory proliferation (15, 16).

Pioneering work performed in Drosophila over the past decade have helped to identify the mechanisms and signaling events involved in AiP (reviewed in (12, 17, 18)). These genetic models of AiP take advantage of the high regenerative capacity of larval imaginal discs, developing epithelial primordia that give rise to adult structures such as eyes and wings. In these studies apoptotic wounds are induced in eye or wing imaginal discs, either genetically by over-expressing IAP antagonists (like hid or reaper) (Figure 2B,C) (19), or by irradiation, to study the role of active caspases for AiP. However, due to the transient nature of the apoptotic process, studying the non-apoptotic signaling events initiated by caspases is difficult. To overcome this limitation, the “undead” models of AiP were developed (20). In these models, apoptotic signaling is induced by expressing IAP antagonists, but cells are kept alive by co-expressing the baculovirus protein P35 that specifically inhibits the effector caspases DrICE and Dcp−1, but not the initiator caspase Dronc (Figure 2C) (21). Rendering cells in an undead state allows uncoupling of the apoptotic and non-apoptotic functions of Dronc, which can now persistently signal to induce compensatory proliferation leading to hyperplastic overgrowth (Figure 1B) (20, 22–25). For example, if hid and p35 are simultaneously expressed in the developing eye imaginal disc, hyperplastic tissue overgrowth is induced, and adult flies are characterized by enlarged head cuticles with several patterning defects (Figure 1C). The overgrowth phenotype of undead fly heads has been used for genetic screens aimed at identifying genes and mechanisms involved in AiP (26, 27).

Mechanistically, it was soon realized that in undead cells, Dronc promotes the activation of Jun-N-terminal kinase (JNK) signaling as the major inducer of AiP (25, 28) (Figure 2B,C). However, it has been unknown for a long time how active Dronc promotes JNK signaling. Recently it was reported that active Dronc promotes the generation of extracellular reactive oxygen species (eROS), which activate JNK signaling in the undead eye disc tissue (26). If this is the only mechanism by which Dronc activates JNK or if any other mechanisms exist remains to be seen. JNK functions in a positive feedback loop in AiP as it transcriptionally activates hid and reaper, thus amplifying the AiP process (26, 29, 30). Downstream of JNK signaling, undead cells produce and secrete several mitogens - including Wingless (Wg), a WNT-family member, Decapentaplegic (Dpp), a TGF-β family member, and Spitz (Spi), an EGF homolog (Figure 2C) (25, 28, 31). These mitogens then signal to the neighboring cells to initiate proliferation.

Although studies using undead models have provided a lot of insight into the mechanisms of AiP, the fact that cells are kept alive under constant apoptotic stress might alter their signaling properties, thus compromising the relevance for understanding physiological AiP. Consequently, p35-independent models of AiP have been developed, known as the “genuine” or regenerative AiP models. In the genuine models, tissue ablation is induced by a temporal pulse of apoptosis in a spatially-restricted manner. The affected imaginal discs are then allowed to recover and regenerate the lost tissue through AiP (Figure 1A) (32). Studies using the genuine AiP models have mostly corroborated the findings obtained in the undead models (32, 33), although as often seen with the use of different models, there is controversy about the involvement of individual components such as Wg and Dpp in AiP (31). Nevertheless, the studies using genuine models have confirmed the involvement of JNK signaling for inducing AiP (32, 34). Along with JNK signaling, p38 and JAK/STAT signaling pathways are also required for genuine AiP.

Furthermore, as described above for the undead model, ROS are also generated in the genuine models, and are required for activation of p38 and JNK signaling (33).

ROS and ROS signaling

Reactive oxygen species (ROS) are formed upon partial reduction of oxygen, and include superoxide anions (O₂⁻), hydroxyl radical (OH) and hydrogen peroxide (H₂O₂) (35). They are highly unstable with a relatively short half-life. O₂⁻ is generally considered to be the primary ROS and is abundantly generated by different endogenous and exogenous factors. O₂⁻ is rapidly dismutated to H₂O₂, and in the presence of Fe²⁺ or Cu²⁺ ions, H₂O₂ can further be converted to .OH via a process known as Fenton reaction (Figure 3).

Figure 3. Sources of ROS and Enzymes involved in Redox Signaling.

Abbreviations used: F – Flavin domain; EF – EF hand domain; P – peroxidase homology domain (PHD); SOD – superoxide dismutase; Cat – catalase; GPxs – glutathione peroxidases; PRxs -peroxiredoxins. See text for details.

Mitochondria are the primary source of intracellular O₂⁻ (36), followed by other organelles such as endoplasmic reticulum (ER), peroxisomes, and the phagosomes in specialized phagocytic cells that display localized generation of ROS (Figure 3) (37). In mitochondria, the electron transport chain (ETC) complexes transfer electrons from NADH and succinate to synthesize ATP during aerobic respiration. Leaks in the ETC, especially in complex I and III cause a one electron reduction of molecular O2 to form O₂⁻ in the mitochondrial matrix (36). In the ER, H₂O₂ is produced by post-translational oxidative modifications during protein folding (37, 38), while in the peroxisomes, ROS are generated as byproducts of catalytic functions of enzymes involved in various metabolic pathways like α- and β-oxidation of very long chain fatty acids, amino acid catabolism, and others (39, 40).

Another major source of ROS are the membrane-associated NADPH oxidases NOX and dual oxidase (DUOX) (Figure 3). They show widespread localization to different cell and organelle membranes, and account for the generation of extracellular ROS (eROS), in particular via DUOX. They catalyze the reduction of O2 to O₂⁻ by using NADPH as an electron donor (41). These enzymes are evolutionarily conserved in eukaryotes. There are total of 7 NOX family members - NOX1−5 and DUOX1−2 in mammals, while in Drosophila there is one homolog for NOX (dNOX) and one for DUOX (dDUOX), and in C. elegans there is no NOX homolog, but 2 DUOX enzymes present (42). NOX proteins are characterized by the presence of a carboxyterminal intracellular Flavin domain that contains binding sites for co-enzymes NADPH and FAD (Figure 3), an amino-terminal hydrophobic domain that forms 6 transmembranal α-helices with four highly conserved heme-binding histidine residues in the transmembrane domain that act as carrier to transport electrons across the membrane, and an additional amino-terminal intracellular Ca²⁺-binding EF hand domain observed in NOX5. DUOX proteins have a similar structure as NOX5 with the addition of a seventh transmembrane domain and an extracellular peroxidase-homology domain (PHD) (Figure 3). Thus, DUOX proteins can generate O₂⁻ through the catalytic core and potentially further process it through its own peroxidase domain (41).

There is a delicate balance between ROS generation and scavenging by the protective antioxidant defenses in the cell. Antioxidant systems present in the cells include enzymes like superoxide dismutase (SOD), catalases, glutathione peroxidases (GPxs) and peroxiredoxins (PRxs) (Figure 3) (38). SODs are responsible for catalyzing the rapid dismutation of highly reactive O₂⁻ anions to H₂O₂: SOD1 converts cytosolic O₂⁻, mitochondrial matrix-associated SOD2 converts O₂⁻ generated by mitochondrial ETC, while SOD3 is secreted extracellularly and is responsible for converting extracellular O₂⁻ produced by the NADPH oxidases (43). The antioxidant enzymes like catalases, GPxs and PRxs convert H₂O₂ to H2O and O2. The GPx family proteins reduce H₂O₂ by oxidizing glutathione, which is then reduced back by glutathione reductase using NADPH as an electron donor, thus normalizing the levels of reduced glutathione in the cells (44). PRx family enzymes contain a redox-sensitive cysteine residue in their active site, which is inactivated by H₂O₂-mediated oxidation. Thioredoxin acts as an electron donor to reduce and activate PRx, thus completing the catalytic cycle (45).

Historically, ROS were thought to be deleterious to the cell causing oxidative stress by indiscriminately damaging proteins, lipids, and nucleic acids. This holds true for the highly reactive O₂⁻ and OH radicals, as they can cause irreversible oxidative damage due to their strong oxidizing potential and lipid insolubility, thereby contributing to cellular dysfunction and various pathologies. However, it is becoming more appreciated now that ROS can also mediate important signaling functions, referred to as redox-signaling (46). For example, H₂O₂ is a perfect candidate to function as second messenger or as signaling molecule due to its relatively long half-life, high stability, and its ability to diffuse across membranes. Indeed oxidation of critical cysteine residues in redox-sensitive proteins is the most studied mechanism by which H₂O₂ functions as signaling molecule (47). The cellular targets of H₂O₂ that undergo this reversible cysteine oxidation encompass a vast range of biological processes. Examples include the phosphatases PTEN and PTP1B, kinases like MAPK and redox-reactive transcription factors like YAP1 in yeast and FOXO4 in mammals (37). H₂O₂ oxidizes the thiol side-chain of cysteine to form reactive sulfenic acid (-SOH) that can form intra- and inter-molecular disulphide (-S-S-) bonds or cyclic sulfenamide (-S-N-) structures, or can undergo hyperoxidation to form sulfinic (-SO2H) or sulfonic (-SO3H) acids (48). These reversible modifications may lead to changes in protein structure, function or activation state.

Along with cysteine thiols, H₂O₂ can oxidize several other amino acids like methionine, lysine, arginine, proline, histidine and tyrosine (49). Multiple mechanisms have been proposed for understanding how target proteins are selected for oxidation by H₂O₂. One mechanism proposes colocalization of ROS sources and targets, so that redox signaling events are triggered close to the source of ROS generation (37). For example, NOX proteins are often seen colocalized with putative targets like phosphatases or kinases at the plasma membrane, thereby influencing receptor tyrosine kinase signaling (50). Another mechanism termed as “redox relay” proposes that H₂O₂ oxidizes the scavenging enzymes like PRx or GPx, which subsequently transfer the oxidation to target proteins (51). This kind of relay process is seen for H₂O₂-mediated oxidation of apoptosis signaling kinase 1 (ASK1) and downstream phosphorylation of its MAPK substrate p38, which is dependent on formation of a ASK1-PRx1 disulphide intermediate (52). Yet another mechanism termed as “floodgate model” proposes that transient inactivation of scavenging enzymes by hyperoxidation or posttranslational modifications causes accumulation of H₂O₂ allowing oxidation of target proteins (53). Thus, by localized alterations in the redox buffering capacity, the cell can control ROS flux for selective signaling events (54).

Furthermore, depending on the location of a Cys residue within a protein, not all Cys residues are equally susceptible to H₂O₂-mediated oxidation (55, 56), further increasing specificity. Another form of physiological ROS regulation involves transport of H₂O₂ across cell membranes via aquaporins, which are integral membrane proteins involved in transport of water and small-molecule metabolites (57). Aquaporins enhance the membrane permeability of H₂O₂, and are useful for transporting the extracellular H₂O₂ produced by NADPH oxidases across the plasma membrane to mediate intracellular signaling cascades (58).

ROS in AiP and Regeneration

Tissue wounding and inflammation are associated with production of ROS, and recent studies show that generation of ROS, especially during the initial stages of wounding and regeneration are essential for an efficient wound healing response. ROS, in particular H₂O₂ function as an early damage signal to initiate and control different aspects of regenerative responses. In our recent work, we found that extracellular ROS (eROS) are produced by actively proliferating epithelial cells in response to activation of Dronc by pro-apoptotic signals. These eROS are required for the compensatory proliferation in both undead and genuine regenerative models of AiP in Drosophila eye and wing imaginal discs. Dronc triggers generation of eROS via activation of dDUOX specifically, and dNOX to a lesser extent (Figure 4A). Loss of eROS by silencing dDUOX or by expressing extracellular catalases caused an impaired proliferative response (26, 59). In the undead AiP model, eROS are necessary for recruitment and activation of hemocytes, Drosophila macrophage-like immune cells, to the eye imaginal discs. Hemocytes in turn secrete the TNF ortholog Eiger, which activates JNK signaling in the undead epithelial cells, and thus contributes to AiP (Figure 4A) (26). The presence of hemocytes on the overgrown undead epithelial tissue is reminiscent of tumor-associated macrophages (TAMs) seen in many solid tumors in human cancers. TAMs increase the inflammation linked with cancer, and function in tumor progression, immune suppression, angiogenesis and metastasis. High TAM-infiltration of tumors is associated with poor patient prognosis (60). Peripheral blood monocytes are recruited to the tumor, and get “alternatively activated” (M2 phenotype) in response to the tumor microenvironment to differentiate into TAMs. Several tumor-derived factors like chemokines, cytokines and microparticles, extracellular matrix components and immune cells in the tumor microenvironment help polarizing TAMs to adopt a M2-like phenotype (61). Along with these factors, it would be interesting to know if ROS from tumor cells contribute to activating TAMs, as is seen in the case of undead AiP model in Drosophila.

Figure 4. ROS Signaling in AiP and Regeneration in various model organisms.

(A) H₂O₂ attracts immune cells (hemocytes) to sites of wounding or damage in Drosophila. The left panel displays the signaling events in the undead AiP model. The right panel shows the signaling events after laser-induced tissue damage in Drosophila embryos.

(B) H₂O₂ signaling after wounding or amputation in several regeneration models indicated at the bottom of each panel.

Abbreviations used: TNF – Tumor Necrosis Factor; JNK – Jun N-terminal Kinase; Upd – Unpaired (Interleukin−6 homolog); Wnt – Wg and Int−1; FGF – Fibroblast Growth Factor; IL−11 – Interleukin 11; JAK/STAT – Janus Kinase/Signal Transducers and Activator of Transcription.

ROS-mediated recruitment of immune cells to wound sites is seen in several different model organisms. The role of H₂O₂ as a chemoattractant for recruiting leukocytes to the site of wounding was first demonstrated in zebrafish. Tail fin wounding of zebrafish larvae led to activation of DUOX at the site of injury, and generation of a tissue-scale gradient of H₂O₂ that promotes leukocyte chemotaxis to the wound. Loss of H₂O₂ gradient impaired leukocyte recruitment to the wound causing an imperfect wound healing response (62). Further evidence supporting the role of H₂O₂ as a chemoattractant was demonstrated by studies in Drosophila embryos. Laser wounding of the embryonic epidermis triggers a rapid calcium flash at the wound edge that activates dDUOX via its EF hand motifs, leading to generation of H₂O₂ (Figure 4A). H₂O₂ is responsible for recruitment of embryonic hemocytes to the wound, thus acting as early damage molecule in wound inflammatory response (63, 64).

However, these studies have not addressed the exact mechanism by which H₂O₂ promotes immune cell chemotaxis in vivo. It will be interesting to determine if H₂O₂ acts as a primary chemoattractant, or whether it stimulates production of downstream chemoattractants by the epithelial cells, or whether they modulate the extracellular matrix by interacting with the chemotactic plasma membrane receptors to attract immune cells. All these mechanisms are consistent with the paracrine signaling role of H₂O₂, and may act redundantly in different contexts. Direct chemotactic activity of H₂O₂ can be mediated via entering the cytoplasm and modulating intracellular signaling events in the immune cells. This was demonstrated in zebrafish larvae, where wound-derived H₂O₂ activated the Src family kinase Lyn in neutrophils through direct oxidation of a cysteine residue, thereby mediating leukocyte migration to the wound (65). How redox sensing by Lyn instructs directional migration of neutrophils remains to be determined. This was answered in part by a study in Drosophila embryos, where epidermal wounding generated H₂O₂ that activated the Lyn homolog Src42A in the hemocytes. Activated Src42A phosphorylated and activated Draper-I (a member of CED−1 family of phagocytic receptors), thereby leading to activation of the kinase Shark, and migration of hemocytes to the wounds. Hence, it was proposed that H₂O₂ acts as an activator signal rather than a chemoattractant to activate a signaling cascade in the macrophages to prime them to respond to wounds (66).

In addition to attracting immune cells to the site of injury, ROS also participate in proliferative responses during compensatory proliferation and regeneration. In Xenopus tadpoles, tail amputation induces sustained production of ROS over the span of regeneration, and decreasing ROS levels, especially early on after amputation, resulting in impaired tail regeneration. Amputation-induced ROS are also produced via NADPH oxidases, and activate Wnt/β-catenin and FGF signaling, thus initiating regeneration (Figure 4B) (67, 68). The role of caspases in ROS production was not investigated in this model, however, caspase−3 activity was previously shown to be important for proliferation during the first 24 hours after amputation (14). In adult zebrafish, caudal fin amputation causes sustained ROS production, unlike the transient ROS response observed after larval tail fin injury. ROS are immediately detected near the lesion after fin amputation, and are tightly regulated during the first 24 hours of regeneration, reaching a peak at 12 and 16 hours post amputation. This sustained ROS generation is specific for the regenerative response and is not observed during wound healing responses. In this regenerative response, ROS generated via enzymatic activity of NOX induce JNK activation and delayed apoptosis (Figure 4B). Both of these parallel processes are important for blastema formation and compensatory proliferation of epithelial cells (69).

In addition to epithelial regeneration, ROS production induced by amputation of the caudal fin also induces neuro-regeneration. In larvae, increased H₂O₂ produced by DUOX1 at wound sites is important for peripheral sensory axon regeneration and re-innervation of the skin (70). In adult zebrafish, sensory neurons, especially Schwann cells, induce H₂O₂ production via Hedgehog signaling post amputation, and this H₂O₂ stimulates the axonal growth and attracts peripheral axons to the regenerating blastema (71). In a genuine AiP model in Drosophila wing discs, ROS generation triggered in response to a transient pulse of apoptosis induced activation of JNK and p38 signaling in the surviving cells that resulted in expression of JAK/STAT pathway ligand Unpaired (Interleukin−6 (IL−6) homolog), leading to regeneration of lost tissue (Figure 4B) (33).

Similarly, another study showed that following acute liver injury in mice, dying hepatocytes produced ROS that induced production of IL−11 (a member of IL−6 family of cytokines). IL−11 triggers activation of JAK/STAT signaling in healthy hepatocytes, which results in compensatory proliferation (Figure 4B). However, it was not investigated whether involvement of caspases in hepatocytes was necessary for production of ROS or downstream expression of IL−11 in this model of liver injury (72). In Hydra, ROS are produced immediately at the wound edges following bisection, and are important for injury-induced cell death and MAPK activation in the head regenerating tips (73). In planaria, amputation of both the head and tail regions induces ROS production at the wound site, which is necessary for regeneration, patterning, polarization of proliferating cells, and early nervous system differentiation. This study provided the first evidence that ROS are involved in anterior body regeneration, and that production of ROS was independent of the orientation of the wound site (74). In mice, partial hepatectomy induced ROS production is tightly regulated by modulating activities of NADPH oxidases and scavenging enzymes. During the regenerative phase, NOX4 is activated while PRxs and catalases are downregulated to facilitate increased H₂O₂ production. On the other hand, during the termination phase of regeneration, the level of NOX4 is reduced, and that of PRxs and catalases are induced to decrease H₂O₂ production. In this model, H₂O₂ triggers two distinct signaling events in a dose-dependent manner. Elevated H₂O₂ levels induced cell proliferation by activation of ERK signaling, and in contrast, low H₂O₂ concentrations promoted cell quiescence by activation of p38 signaling. Thus, activity of specific enzymes and dose-dependent signaling by H₂O₂ triggers the proliferation-quiescence switch to govern liver regeneration after hepatectomy (75).

Implications for Cancer

ROS have a long history of being involved in the development and progression of cancer and increased ROS level is considered as a hallmark of many tumors (76, 77). Initially, it was thought that ROS would serve as chemical mutagens that would indiscriminately damage cellular macromolecules such as DNA by oxidative stress, and thus are tumorigenic by promoting genomic instability. In addition, the more recent findings that ROS, most notably H₂O₂, can also mediate several signaling processes in the cell with very specific oxidation targets has verified this view (recently reviewed in (78)). By regulating signaling events that control the cell proliferation rate, alter the metabolic states of the cell and promote angiogenesis, ROS are increasing the tumorigenic potential of the cancerous cells. Increased ROS production in cancer cells can be due to multiple different factors. Oncogenes such as Ras promote the generation of ROS (79–82). The tumor suppressor p53 is known to establish the redox balance by regulating the expression of antioxidants like GPx1 and SOD2, and loss in its activity can increase the oxidative burden in tumor cells (83–85). Increased metabolism of the tumor cells is another factor that increases the ROS levels in these cells (reviewed in (77)). Many human tumors and cultured tumor cell lines show an upregulation of the mRNA expression and/or protein levels of several different NOX enzymes, which are implicated in increasing the cellular ROS levels. Depending on the subcellular localization of these enzymes and the stage of tumorigenesis, NOX-derived ROS can mediate DNA damage causing genomic instability or signal to activate redox-sensitive pathways, to help in initiation and maintenance of tumorigenesis (86, 87). Along with this, a wide variety of human tumors harbor mutations in mitochondrial DNA-encoding ETC proteins, which are responsible for increasing the mitochondrial-derived ROS production (88).

NOX enzymes can regulate the MAPK/ERK and PI3K/Akt/mTOR signaling pathways through H₂O₂-mediated oxidation of phosphatases involved in these processes (89–91). PTEN is the primary target of H₂O₂, and oxidation of its active site cysteine (Cys124) inactivates the phosphatase, resulting in constitutive activation of the PI3K pathway (89). Inactivation of protein phosphatase 2A (PP2A) and PTP1B by H₂O₂-mediated oxidation causes an increase in Akt activation. This leads to an increased cell proliferation response, anchorage-independent growth and survival (90, 92, 93). Mitochondrial ROS are also responsible to activate the hypoxia-inducible factors (HIFs) in hypoxic tumor cells, thus allowing the tumor cells to adapt to the low oxygen microenvironment and help in its survival. Under hypoxic conditions, increased superoxide generated by mitochondrial ETC stabilizes HIF−1α and HIF−2α subunits (94, 95).

It has also been reported that AiP is involved in the etiology of human cancer ((96–106); recently reviewed in (18)). Chemo- and radiotherapy of human patients often aims to induce the death of the tumor cells. However, apoptotic tumor cells in return may generate signals for AiP and despite initial tumor regression, the tumor cells repopulate and the tumor grows back (97–99, 102). Although a direct role of ROS for this type of tumor AiP has not been demonstrated so far, based on the work in several model organisms, it is possible that this is the case. Future work will need to address this question.

ROS have not only a tumor-promoting role, but they can also have a tumor-suppressive function. For example, genetically engineered mice carrying oncogenic K-Ras and B-Raf mutations significantly increased tumor development and mortality upon dietary supplementation of the anti-oxidants N-acetylcysteine (NAC) or Vitamin E, suggesting that ROS prevented tumor growth in these animals (107). Thus, for the application of therapies which potentially target ROS, we need to have a very detailed context-specific understanding of the role of ROS in these tumors. That will be a challenge for the future.

Conclusions

The initial view that ROS are only produced as a by-product of cellular stress, and cause deleterious oxidative damages in the cells has been replaced after the discovery that cells can temporally and spatially control ROS production, and that ROS-mediated signaling events play an important role in different cellular processes. As discussed here, ROS are critical mediators of AiP for regenerative processes and are also involved in tumorigenesis. Future work is needed to reveal the precise role of ROS and the targets of ROS modification during these processes.