Therapy-induced senescence through the redox lens

Abstract

Therapy-induced senescent tumor cells have emerged as significant drivers of tumor recurrence and disease relapse. Interestingly, reactive oxygen species (ROS) production and its associated redox signaling networks are intertwined with initiation and establishment of therapy-induced senescence. Therapy-induced senescent cells influence neighboring cells and the tumor microenvironment via their bioactive secretome known as the senescence-associated secretory phenotype (SASP). The intracellular effects of ROS are dose and context-dependent. Under normal physiological conditions, ROS is involved in various signalling pathways and cellular processes important for maintenance of cellular homeostasis, such as redox balance, stress response, inflammatory signalling, cell proliferation and cell death among others. However excess ROS accompanied by a pro-oxidant microenvironment can engender oxidative DNA damage, triggering cellular senescence. In this review, we discuss the role of ROS and the redox state dynamics in fine-tuning homeostatic processes that drive therapy-induced cell fate towards senescence establishment, as well as their influence in stimulating inflammatory signalling and SASP production. We also offer insights into interventional strategies, specifically senotherapeutics, that could potentially leverage on modulation of redox and antioxidant pathways. Lastly, we evaluate possible implications of redox rewiring during escape from therapy-induced senescence, an emerging area of research. We envision that examining therapy-induced senescence through the redox lens, integrated with time-resolved single-cell RNA sequencing combined with spatiotemporal multi-omics, could further enhance our understanding of its functional heterogeneity. This could aid identification of targetable signalling nodes to reduce disease relapse, as well as inform strategies for development of broad-spectrum senotherapeutics. Overall, our review aims to delineate redox-driven mechanisms which contribute to the biology of therapy-induced senescence and beyond, while highlighting implications for tumor initiation and recurrence.

1. Introduction

1.1. Redox homeostasis and signaling

Redox signaling comprises a network of pathways interconnecting chains of redox reactions, oxidative post-translational modifications, and regulation of reactive oxygen species (ROS) scavenging systems, all orchestrated carefully to achieve a state of redox homeostasis in the cell. ROS itself is a broad term for unstable derivatives of oxygen molecules, generated as an inevitable consequence of ATP production and cellular metabolism. Notably, two species of ROS - hydrogen peroxide (H₂O₂) and superoxide anion radical (O₂^·−) - have been regarded as key redox signaling agents in cellular processes. Aside from H₂O₂ and O₂^·−, there has also been growing interest in other species of ROS, such as the hydroxyl radical (•OH), hypochlorous acid (HOCl) and organoid hydroperoxides (ROOH) [1]. Importantly, ROS interfaces with ROS scavenging systems consisting of antioxidant enzymes such as superoxide dismutases (SOD), glutathione peroxidases (GPx), peroxidoxins (Prxs), and catalase (CAT) as well as glutathione (GSH) and thioredoxin (Trx) that serve as reducing agents [2]. These scavenging systems perform critical functions in controlling oxidative stress, characterized by elevated levels of unstable oxidizing agents such as oxygen, hydroxyl free radicals and so on.

A notable direct effect of oxidative stress is oxidative DNA damage. While oxidative DNA adducts are constantly repaired at basal level, both mild or severe oxidative stress can result in oxidative damage that may invoke senescence, apoptosis, or even necrosis [3,4]. Despite this, recent observations seem to indicate that ROS should not be associated exclusively with a particular pathological state. Indeed, the state of redox homeostasis itself demonstrates that the redox state is not static but rather dynamic, and dependent upon various internal and external factors exerted upon the cell. As the generation of ROS is an inevitable consequence of energy production, ROS is intricately connected with essential physiological processes. For instance, ROS has been shown to regulate cell polarity during epithelial development [5,6], and implicated in growth factor signaling, proliferation, and adaptation to hypoxia [[7], [8], [9], [10], [11], [12]]. Hence ROS as such is not a toxic substance per se as widely-believed, but should be thought of as an important regulator and signaling component of cellular pathways.

1.2. Role of redox signaling in carcinogenesis

1.2.1. ROS-related redox signaling as an early event that supports ‘pro-tumor’ behavior

Along with a growing appreciation of its role in specific molecular and regulatory functions, the role of ROS has also been implicated in a broader range of cellular processes. For instance, according to the current paradigm on cancer development, the ‘pro-oxidant’ redox state present in neoplastic lesions is characterized as having excess ROS and chronic oxidative stress [13].

Excess ROS levels and oxidative stress have been associated with malignant behaviors, in particular, invasion and metastasis via the epithelial-mesenchymal transition (EMT). As cancer cells metastasize, cells are likely to be exposed to common oxidants present within the blood, thereby increasing intrinsic oxidative stress [14]. The small number of cancer cells that survive this metastatic process must undergo metabolic changes to allow them to cope with this increased oxidative stress. Indeed, metastasizing cancer cells with altered redox signaling pathways have been shown to display higher levels of oxidative stress resistance [15]. In line with this, metastasis-competent circulating tumor cells from breast, prostate, and lung cancers have been documented to upregulate the β-globin gene (HBB) via Kruppel-like factor 4 (KLF4), which confers protection from ROS-induced apoptosis [16]. Additionally, a study from Ralph DeBerardinis' and Sean Morrison's labs that infused ¹³C-labelled nutrients into subcutaneous tumors from patient-derived melanomas that were either efficient or inefficient metastasizers, found that the efficient metastasizers upregulated the lactate transporter monocarboxylate transporter 1 (MCT1) [17]. MCT1 is a bidirectional transporter of lactate and other monocarboxylates such as pyruvate, but the main physiological role of MCT1 is lactate import. As lactate is co-transported with a proton, the maintenance of intracellular pH by MCT1-dependent lactate import was shown to be important. Indeed, treatment with the MCT1 inhibitor AZD3965 increased intracellular pH, thereby activating phosphofructokinase (PFK) activity while suppressing glucose-6-phosphate dehydrogenase (G6PD) activity. This in turn led to reduced pentose phosphate pathway (PPP) flux relative to glycolysis [17]. This reduced PPP flux is consistent with increased ROS levels observed in AZD3965-treated melanomas. Importantly, the oxidative PPP produces NADPH, which is then used by cells to counteract ROS and decrease oxidative stress. In this way, MCT1-dependent lactate uptake contributes to oxidative stress resistance and allows metastatic melanoma cells to survive in the blood during dissemination.

Oxidative stress caused by excess ROS may also lead to DNA mutations via oxidative DNA damage. One of the most studied oxidative damage products is 8-oxo-dG, an oxidized derivative of deoxyguanosine. It acts as a mutagenic lesion as its base-pairing specificity is lost, potentially introducing G:C to T:A base transversions [18]. Other oxidative products may include 2-OH-dA and 8-oxo-dA, derivatives of deoxyadenosine, which are also known to be mutagenic [19,20]. ROS-mediated DNA mutations may lead to inactivation of tumor suppressors or activation of oncogenes [21,22]. These not only engender cancer initiation, but also fuel clonal evolution.

1.2.2. ROS-related redox signaling as a barrier to carcinogenesis

On the flip side to what we have just discussed is the fact that excessive ROS levels and oxidative stress have also been described as barriers to carcinogenesis as they invoke multiple stress pathways that culminate in cell death such as apoptosis, necroptosis or ferroptosis.

During chronic oxidative stress, mitochondrial damage may accumulate excessively due to a ‘vicious cycle’ where impaired oxidative phosphorylation due to oxidative damage further produces mitochondrial ROS, thus exacerbating oxidative stress [23,24]. A key event in apoptosis is the disruption of cytochrome c interaction with cardiolipin in the mitochondrial inner membrane [25]. Importantly, cardiolipin can be oxidized by ROS, disrupting its association with cytochrome c and mediating its release [25]. Excessive ROS can also activate the p53 pathway by phosphorylating p53, leading to the inhibition of Mdm2-mediated degradation and consequently p53 protein stabilization [26,27]. p53 can then promote induction of intrinsic apoptotic pathways via upregulation of pro-apoptotic proteins such as BIM, PUMA and NOXA, leading to MOMP stabilization which then mediates cytochrome c release [28].

Necroptosis on the other hand, is a regulated form of necrosis, which is triggered via cell death receptors, tumor necrosis factor receptor 1 (TNFR1), Fas, tumor necrosis factor-related apoptosis-inducing ligand (TRAIL) receptor 1 (TRAILR1) and TRAIL receptor 2 (TRAILR2) [29]. Upon activation of cell death receptors, a key event in the signaling pathway is the formation of complex IIb, known as the necrosome, consisting of inactive form of caspase-8, RIP1, RIP3, and FADD. RIP1 phosphorylates RIP3, which in turn phosphorylates and activates mixed lineage kinase domain-like (MLKL), that then translocates to the plasma membrane. This increases membrane permeability and triggers cell death. Here, ROS plays an important role in necroptosis initiation via recruitment of the necrosome complex [30,31]. ROS has been shown to facilitate autophosphorylation of RIP1 via modification of its cysteine residues C257, C268 and C586, resulting in the recruitment of RIP3 to form the necrosome [32]. Additionally, tumor necrosis factor (TNF) stimulation has been shown to activate Nox2 NADPH, via complex formation with TRADD, RIP1, and Rac1, leading to superoxide production that is key for TNF-induced necroptosis commitment [30]. Importantly, antioxidant butylated hydroxyanisole (BHA) was found to inhibit superoxide generation, and consequently TNF-induced necroptosis in the mouse fibrosarcoma L929 cell line [33]. Given that cancer cells can evade apoptosis, inducing necroptosis via ROS modulation seems to be a promising strategy to overcome apoptotic resistance.

In addition to necroptosis, ferroptosis is another key non-apoptotic cell death pathway that can be modulated by ROS. Ferroptosis is induced by lipid peroxide accumulation due to the intracellular presence of excess unstable ferrous iron Fe²⁺. Excess Fe²⁺ produces large amounts of ROS through the Fenton reaction which mediates peroxidation of polyunsaturated fatty acids (PUFA), ultimately resulting in the destruction of the phospholipid bilayer and cell death [34]. The gluthathione antioxidant system is critical in regulating ferroptosis via the GPx4 enzyme, which allows oxidation of GSH to GSSG for conversion of toxic phospholipid hydroperoxides to non-toxic phospholipid alcohols [35]. An altered redox state compromises this GSH-dependent system and has been shown to result in the accumulation of lipid peroxidation products, leading to cell death [36,37]. Interestingly, certain cancers that inactivate the tumor suppressor NRF2, a master regulator of antioxidant pathways, show increased sensitivity to ferroptosis [38]. Thus, inhibition of antioxidant enzymes can be viewed as an ‘Achilles heel’ that can be targeted to induce ferroptotic cell death.

1.2.3. ROS-related redox signaling regulates both rapid and long-term adaptation

The dichotomous roles of ROS and its influence on the redox state in carcinogenesis illustrate an effective way for fine-tuned adjustments of cellular function (Fig. 1). Under normal physiological conditions, low levels of ROS have been demonstrated to be involved in regulating self-renewal, and serve as specific second messengers in innate and adaptive immune responses [39]. In cells with oncogenic lesions and a pro-malignant microenvironment, the gradual accumulation of ROS and chronic oxidative stress is favored to support malignant transformation and behavior, as well as selection for high oxidative stress-resistant phenotypes. In cells with benign lesions and non-neoplastic environments, high levels of oxidative stress may tip the balance and render acute and excessive damage to DNA and organelles. Various stress pathways that induce apoptosis are more likely to be invoked in this situation. Hence, redox signaling can be viewed as a ‘molecular toolkit’ that regularly fine-tunes cellular and physiological processes that enables rapid and long-term cellular adaptative responses based on internal and external stress perturbations. This is critical for maintenance of cellular homeostasis, essential for human health.

Fig. 1.

ROS as a “molecular toolkit” that fine-tunes molecular and physiological processes for adaptation and malignancy. Basal (low) ROS levels: In normal physiological environments, ROS/redox signalling acts as a physiological rheostat to regulate critical cellular functions for development and adaptation to the surrounding millieu. This maintenance of redox homeostasis is essential to normal cellular activity. High ROS levels: During cancer initiation, intrinsic factors such as oncogenic lesions, and external factors such as the chronic inflammatory environment, shift redox homeostasis into a ‘pro-oxidant’ state. This ‘pro-oxidant’ state in turn facilitates pro-tumorigenic behaviors and selects for higher tolerance against oxidative stress. In this way, alterations in ROS and redox signalling allow fine-tuned adjustments that underlie both physiological and pathological conditions.

1.3. Therapy-induced senescence (TIS)

1.3.1. Hallmarks of TIS and key markers of senescence

Leonard Hayflick first demonstrated that human diploid fibroblasts failed to proliferate after about fifty passages in culture, and termed the phenomenon ‘cellular senescence’ [40]. Today, cellular senescence is regarded as a fundamental tumor cell response to therapy, as a growing body of evidence points to the accumulation of senescent cells in patients receiving radiation therapy and chemotherapy [[41], [42], [43]]. The term “therapy-induced senescence” (TIS) refers to the stress response exerted by cancer therapies that culminates in a stable cell cycle arrest.

As with other types of senescence such as replicative senescence and oncogene-induced senescence, TIS cells exhibit the main hallmarks of senescence: an enlarged, flattened and irregularly-shaped morphology, increased lysosomal content, as well as Lamin B1 depletion [44]. Fortunately, these characteristics are detectable and easily visualized, and therefore measurable, via simple laboratory procedures such as microscopy and immunoblotting. These methodologies include the histochemical detection of senescence-associated beta-galactosidase (SA-β-gal) enzymatic activity at pH 6.0 (in normal cells, SA-β-gal shows optimal activity at pH 4.0) [45], as well as the loss of LaminB1, usually measured by immunoblotting. These are also often used as senescence-associated markers in various tissues, such as skin senescence [46]. In addition, the presence of DNA damage is also an important feature of TIS cells, especially in the early phases of senescence. The severity of DNA damage may manifest from localized double-stranded breaks (DSBs) to genome-wide instability, such as aneuploidy, chromosomal rearrangements, micronuclei formation, or even extrachromosomal DNA [[47], [48], [49]]. As a result, TIS cells often exhibit a persistent DNA damage response (DDR), which involves the recruitment and activation of DNA damage-sensing kinases such as ataxia telangiectasia and Rad3-related (ATR) or ataxia-telangiectasia mutated (ATM) and, possibly, DNA-dependent protein kinase (DNA-PK). Activation of ATM via DSBs for example, leads to phosphorylation of signalling proteins such as nibrin (NBS1) and the histone variant H2AX (γH2AX), ultimately leading to activation of p53 and senescence via transactivation of cyclin-dependent kinase (CDK) inhibitors p16, p21 and p27 [50]. Hence, increased expression of γH2AX, phosphorylated p53, and upregulation of p16, p21, p27 are commonly used to demarcate entry into senescence.

1.3.2. Heterogeneity of the therapy-induced senescence secretome

Despite the existence of defined markers, the senescence state displays much heterogeneity, and is accompanied by a bioactive secretome known as the senescence-associated secretory phenotype or SASP. The SASP is highly variable both in terms of its components and its downstream phenotypic effects [51,52]. A case in point is a study by Marco Demaria and colleagues that highlighted this in its comparison of CDK4/6 inhibitor (CDK4/6i)-induced senescence and doxorubicin (Doxo)-induced senescence [53]. A key finding here was that while CDK4/6i- and Doxo-induced senescent cells both displayed senescence hallmarks, their secretory profiles and consequent downstream phenotypes differed from each other. The Doxo-induced senescent cells displayed a canonical NF‐κB-driven SASP that was pro-inflammatory in nature. In contrast, CDK4/6i-induced senescent cells displayed paracrine senescence activity, did not acquire tumorigenic properties, and its p53-driven SASP was found to support immunosurveillance. Indeed, CDK4/6i were found to trigger anti-tumor immunity by suppressing the proliferation of regulatory T cells (Tregs) and promoting cytotoxic T cell-mediated clearance of tumor cells [54].

Due to the heterogeneity of SASP components, determining the effects of SASP in the context of cancer progression can be challenging. While TIS cancer cells are generally thought to be drivers of tumor recurrence and mediators of chronic inflammation [[55], [56], [57]], recent studies focused on immune checkpoint inhibitors have revealed that senescent cancer cells are exceptionally immunogenic. For instance, a recent study from Manuel Serrano's group [58] demonstrated that cancer immunosurveillance could be induced by immunizing mice with senescent cancer cells. Remarkably, these anticancer immune responses were reported to be even superior to that of cells undergoing immunogenic cell death (ICD) during immunization, considered the gold standard for eliciting cell-based antitumor immune responses [59]. While the secretion levels of damage-associated molecular patterns (DAMPs) such as adenosine triphosphate (ATP) and calreticulin (CALR) were observed to be similar in senescent cancer cells and ICD-cells, a distinguishing feature was that senescent cancer cells were retained for a longer period of time in the skin, resulting in the persistent production of inflammatory SASP and DAMPs compared to the ICD-cells [58]. In line with the theme of immunogenicity, another study described the use of senescent cancer cell-derived nanovesicles (SCCNV), which induced superior antitumor immune responses and suppression of metastasis in mice compared to non-senescent CCNV [60]. Interestingly, both studies [58,60] reported similar patterns of immune activation by senescent cancer cells and SCCNV, specifically a higher dendritic cell (DC) maturation and activation of CD8⁺ T cells, in contrast to normal physiological contexts where senescent cells are mainly cleared by natural killer (NK) cells and macrophages [61].

Importantly, the intricacies of SASP and its dynamics present an opportunity for intervention since it is highly malleable. Reprogramming of SASP factors can be achieved by inhibition of key pathways that sustain a pro-tumorigenic SASP [62]. Inhibition of NF-κB by metformin, for instance, has been shown to reduce the expression of various SASP factors [63]. The mTOR inhibitor rapamycin has also been shown to blunt pro-tumorigenic SASP production via NF-κB inhibition [64]. Aside from NF-κB, the p38MAPK pathway is also key to driving and sustaining pro-tumorigenic SASP [[65], [66], [67]]. Indeed, several studies have indicated that p38MAPK inhibition is able to alleviate bone loss and metastasis in a murine breast cancer TIS model via suppression of the SASP [68,69]. Alternatively, targeting the epigenetic regulators of SASP could prove to be beneficial. For instance, targeting of KDM4, a histone H3-specific demethylase, has been shown to dampen the SASP of senescent stromal cells, leading to increased cancer apoptosis and prolonged survival in prostate cancer animal models [70]. In addition, the inhibition of EZH2 has recently been shown to relieve repression of pro-inflammatory SASP genes, leading to increased cancer immunosurveillance activity in PDAC tumor models [71].

While the modulation of SASP to achieve anti-tumorigenicity seems to be a viable strategy, this should be taken into careful consideration as SASP effects are highly context-dependent. As a case in point, silencing of the NF-κB p65 subunit in cyclophosphamide-induced senescent lymphomas has been shown to not only abrogate SASP, but also results in earlier relapse and shorter overall survival [72]. A second example involves CDK4/6i-induced senescence, where modulation of SASP might not provide a readily-identifiable benefit as its p53-driven SASP response seems to already favor clearance and antitumor immunity [53]. Hence, understanding how various stimuli and activation of different senescence programs influence the SASP phenotype is key to inform strategies for development of novel and improved interventions for SASP reprogramming.

2. Redox dyshomeostasis and acquisition of therapy-induced senescence

2.1. DNA damage and DNA damage repair signaling as fundamental drivers of cellular senescence

An important factor to consider with regards to whether a cell commits to TIS or activates cell death pathways is that the determinants of this cell fate decision are highly context-dependent. The current paradigm implicates both DNA damage and DDR signaling as the crux of senescence induction [73]. A key takeaway from studies that have contributed to this paradigm is that presence of DNA damage alone is insufficient, and that it has to also be accompanied by persistent DDR signaling to induce senescence.

Indeed, an elegant study from Rene Medema's group demonstrated this by using non-transformed telomerase-immortalized (hTERT) retinal pigment epithelial RPE-1 cells already arrested at the G2 cell cycle phase by the DNA damage checkpoint. Here, delaying homologous recombination (HR)-mediated repair of double strand breaks in these cells enhanced ATR signaling which triggered p21 accumulation [74]. The elevated levels of p21 were observed to drive cell cycle exit from the G2-checkpoint arrest and induce senescence. Interestingly, the number of DNA breaks induced did not seem to determine cell cycle exit. Rather, the presence of difficult-to-repair breaks, along with the simultaneous presence of easier-to-resolve breaks, were suggested to be more accurate predictors of senescence.

Related to DNA breaks driving senescence is another important study from Corinne Abbadie's group that took advantage of normal human dermal fibroblasts (NHDFs) and normal human epidermal keratinocytes (NHEKs) that differed in their ability to spontaneously evade senescence and generate neoplastic cells [75]. Here, accumulation of persistent single-stranded breaks (SSBs), characterized by XRCC1 foci, induced a transient senescent plateau via activation of the p38MAPK/p16/Rb pathway. This transient senescence was followed by accumulation of clonal populations of transformed and mutated cells, indicating a role in engendering tumor initiation.

A critical sub-region of the chromosome that is overtly distinct in terms of DNA repair compared to other regions is the telomere. Indeed, DSB repair of telomeres tend to specifically suppress the non-homologous end joining (NHEJ) pathway to avoid end-to-end fusion of chromosomes [76]. A study by Joao Passos and colleagues using AcGFP-53BP1 and telomere probes, identified an accumulation of telomere-associated foci (TAFs) both in vitro and in vivo in the liver and small intestines of aged mice, independent of damage foci induced via telomere shortening and uncapping [77]. Importantly, time-resolved fluorescence microscopy experiments further showed that these TAFs took longer to resolve than non-TAFs, suggesting reduced repair efficiency.

2.2. ROS and the pro-oxidant state support acquisition of cellular senescence

Given our discussion thus far on senescence drivers, one might ask, how does ROS play into these concepts and influence senescence acquisition?

In some way, ROS could be considered a key initiator of senescence via the exacerbation of oxidative stress (Fig. 2A). Telomeres, in particular, have been shown to be prone to oxidative damage. An elegant study by Patricia Opresko's group previously highlighted this by leveraging on a FAP-mCerulean-TRF1 fusion protein to specifically introduce 8-oxo-guanine (8oxoG) lesions at the telomeres [78]. Remarkably, acute telomeric 8oxoG was able to initiate rapid premature senescence, characterized by ATM-driven DDR signaling that led to p53 activation. Telomeric 8oxoG-induced DDR, which arose from replication-associated telomeric fragility, rendered replicating cells more sensitive to telomeric 8oxoG-induced senescence than quiescent cells. In addition, the telomeric state has also been shown to affect mitochondrial function and subsequently ROS production, via control of mitochondrial biogenesis factors [79]. This places the telomeres and mitochondria in a positive regulatory loop, of which the dysfunction of one may exacerbate dysfunction of the other. Despite this, acute telomeric 8oxoG did not seem to affect cancer cells, even those deficient in 8oxoG glycosylase (OGG1) activity [80]. Considering the non-discriminatory cytotoxic activity of chemotherapies that specifically target rapidly-proliferating cells, one could speculate that such mechanisms could be involved in driving chemotherapy-induced side-effects and accelerate aging phenotypes in cancer patients and survivors post-treatment.

Fig. 2.

Therapy-induced senescence through the redox lens. (A) Cancer therapy induces high ROS levels and high oxidative stress in cells, causing the formation of telomeric 8-oxoguanine and telomeric oxidative damage. The telomeric lesion is sensed by DNA damage-sensing proteins, causing activation of p53/p21 pathways, resulting in senescence. (B) Genotoxic chemotherapy causes inhibition of DNA replication and the cell cycle, which induces global DNA damage. Double-stranded and single-stranded breaks are sensed by ATM and ATR respectively, inducing a p53 response that results in activation of the senescence program. The p53/p21 response may in turn induce production of ROS via mitochondrial dysfunction, causing increased oxidative DNA damage and persistent DNA damage response signaling. (C) SASP induces paracrine DNA damage as well as increased ROS production via upregulation of NADPH oxidase (NOX). In the absence of genotoxic drugs, ROS-induced DNA damage may play a more prominent role in driving p53 activation, resulting in paracrine senescence. (D) Senescent cells may accumulate at the invasive front of cancers and create high ROS levels, high oxidative stress, and a pro-inflammatory environment via the SASP. Pro-Inflammatory SASP factors then facilitate aggressive and invasive behavior, while the high ROS microenvironment creates an immunosuppresive, tumorigenic-permissive millieu. (E) A small subpopulation of therapy-induced senescent cells may re-enter the cell cycle via upregulation of Cdk1. These “senescence-escaped” cells are phenotypically different to their non-senescent counterparts prior to escape, and are characterized by high self-renewal capacity via upregulation of stemness-related genes. Upon escape from senescence, these cells possess a low ROS, high antioxidant phenotype, in contrast to the senescence-associated pro-oxidant environment. This suggests a redox-rewiring event associated with senescence escape with implications for tumor recurrence.

On the other hand, initiation of DNA damage and DDR are more likely to originate from cytotoxic drugs such as alkylating agents, anthracyclines, and topoisomerase inhibitors, or radiation therapy, which directly act on DNA, and essential cellular processes such as DNA synthesis or replication [81]. Within this context, the generation of ROS and subsequently the pro-oxidant state, play important secondary roles in reinforcing DNA damage and induction of persistent DDR necessary for commitment to senescence (Fig. 2B). As an example, it has previously been reported that the long-term activation of p21 could induce mitochondrial dysfunction and elevate intracellular ROS via the GADD45A-p38MAPK-GRB2-TGFBRII-TGFβ pathway [82]. This led to replenishment of short-lived DNA damage foci and maintenance of the DDR. Another study also described a positive feedback loop of the accumulation of p53 and elevated intracellular ROS, specific in stress-induced senescent cells, via negative regulation of the ubiquitin E3 ligase CUL4B [83]. Here, ectopic expression of CUL4B successfully attenuated senescence by blunting p53 activation and ROS production. Taken together, this suggests that disrupting the p53/p21-ROS positive feedback loop could offer an effective strategy to modulate cell fate by use of antioxidants or ROS-inducing agents. Indeed, the ROS inhibitor N-acetylcysteine (NAC) was able to prevent cisplatin-induced senescence in wild-type p53 hepatocellular carcinoma cells [84], while an exogenous source of ROS combined with physiological levels of p53 was able to drive cells towards the apoptotic pathway instead of cellular senescence [83]. Collectively, the p53/p21 and ROS feedback loop not only represent crucial elements of senescence establishment but also a targetable axis for therapeutic intervention, possibly in conjunction with chemotherapy, to influence cellular fates.

3. The redox state and its non-cell autonomous impact

3.1. ROS and altered redox states mediate senescent cell bystander effects

Aside from the initiation and establishment of senescence, ROS and altered redox states may also affect the functional phenotype of senescent cells, which can be dictated via their secretome or SASP. One of the notable features of the SASP is its ability to induce paracrine senescence in neighboring cells [85], a phenomenon often referred to as the “senescent cell bystander” effect [86]. In principle, molecular machineries that establish secondary senescence are similar to that of primary senescence, characterized by the presence of DNA damage, DDR, p53/p21 activation, and avoidance of programmed cell death pathways, suggesting that there are certain SASP entities, amongst a complex composition, that are able to induce DNA damage in neighboring cells. In this context, there has been evidence of at least a partial ROS control of the “bystander effect” via modulation of pathways that alter SASP composition. In stress-induced senescent human lung fibroblast MRC5 and IMR90 cells for instance, hyperproduction of ROS due to mitochondrial dysfunction has been shown to drive NF-κB-driven SASP that contributed to increased 53BP1 foci formation and activation of DDR in young bystander cells [87]. Interestingly, elevated oxidative stress has been found to be a consistent theme in bystander cells exposed to the senescent secretome [[87], [88], [89]]. Indeed, within the cytokine-rich inflammatory SASP, several cytokine members, such as TNF-α, IL-1β and IFN-γ have been found to increase ROS in recipient cells [90]. In drug-induced paracrine senescent non-transformed BJ fibroblast cells, IL-1β, in conjunction with TGF-β, has been reported to drive DNA damage via upregulation of the NADPH oxidase Nox4 [91]. This could mediate excessive H₂O₂ production along with its binding partner p22^phox [92]. Studies on bystander senescent cells arising from low-dose alpha-particles and microbeams-induced irradiation on human cell cultures, further supported the role of oxidative stress as an inducer of DNA damage that led to upregulation of NADPH oxidase, along with a concomitant increase of superoxide anions and H₂O₂ [[93], [94], [95]]. These also contributed to activation of p53 and p21^Waf1. Taken together, these studies suggest that there might be heavy reliance on ROS for sustainable maintenance of DNA damage that can eventually culminate in senescence, even in the absence of potent DNA damaging agents (Fig. 2C).

3.2. ROS and oxidative stress support a pro-tumorigenic SASP program

Constitutive ROS production associated with induction and establishment of the senescent phenotype is also frequently connected to global changes in the redox proteomics landscape or a rewiring of redox signaling networks. Examples of these include bulk changes of cysteine oxidation observed during stressed conditions [96,97], and significant rewiring of redox signaling in aging models in vivo [98,99]. Such alteration of redox signaling networks may not just affect the redox homeostatic balance and its associated antioxidant networks, but also various cellular functions and phenotypes. A notable example of this was reported in a study that used the SILAC-iodoTMT labelling approach to track redox-sensitive cysteine residues in ionizing radiation-induced senescent RPE-1 cells [100]. This revealed a shift in redox signaling that favored the preservation of antioxidants and proteins from the peroxiredoxin (Prx) family such as CAT, GPx1, GPx4, PDXN, PRDX1, and PRDX2, which led to higher tolerance of oxidative stress. Remarkably, knockdown of peroxiredoxin 6 (PRDX6) was not only able to sensitize senescent RPE-1 cells to exogenous H₂O₂, but also altered the cytokine landscape, characterized by the suppression of IL-6, TGFB-1, and SERPINE-1 [100]. Similar observations have also been obtained in stress-induced senescent dermal fibroblast cells, where treatment with NAC managed to attenuate production of IL-6, IL-8, and TGFB3, likely via suppression of the TAK1/IKK-B/IRF5 axis [101].

At the invasive front of cancers, inflammatory responses act as powerful drivers of cancer invasion and metastasis [102]. Indeed, the invasive front of breast cancer tissues was previously found to display higher incidence of senescence markers and SASP factors [47]. Consistent with this, a recent study that examined spatial heterogeneity of colorectal cancers identified major populations of senescent tumor cells residing at the invasive front. This was characterized by p16^INK4A staining, secretion of inflammatory SASP, and upregulation of matrix metalloproteinase-7 (MMP7) that can drive metastatic phenotypes [103]. Senescent colorectal tumor cells were found to originate from proliferating cancer cells located at the tumor centre, gradually transitioning to a partial EMT phenotype, concomitant with the acquisition of p16^INK4A, as they migrated to the invasive front. Of note, upregulation of p16^INK4A and MMP7 expression was highly associated with low expression of DNA methyltransferase 1 (DNMT1), which was attributed to high ROS production at the invasive front [103]. In addition, CD44v9, a transcriptional variant of CD44 known to generate glutathione, was also found to be suppressed in p16^INK4 positive senescent tumor cells, suggesting CD44v9-ROS-DNMT1 axis acts as a key player of tumor senescence-induced metastatic phenotype at the invasive front [103]. One might conclude therefore that while the specific pathways that drive SASP may vary, oxidative stress seems to be a key contributor to creating a conducive environment for inflammatory and pro-tumorigenic SASP production (Fig. 2D).

3.3. ROS mediates senescence-associated alterations of the immune landscape

With the advent of immune checkpoint therapy and advancement in single-cell technologies for identification of specific immune cells in bulk samples, the immune microenvironment has been regarded as an important aspect of cancer pathology. While senescent cells are mainly eliminated via immune-mediated clearance, impaired immunosurveillance may engender a chronic senescent microenvironment [58]. ROS are not only mediators of oxidative stress, but also serve as pivotal players in immune cell regulation during tumor development. ROS mostly act as immunosuppressive participants in tumor progression. High levels of ROS may mediate immunosuppressive effects; for instance, exposure of T cells to high ROS levels can downregulate T cell activity [104]. On the other hand, low levels of ROS can induce the immunoregulatory enzyme, indoleamine 2,3-dioxygenase, and enhance the function of regulatory T cells (Tregs) that suppress the immune response [105]. Interestingly, analyses of publicly-available cancer databases suggested an association between an increased oxidative state (high redox score) and an ‘immune-desert’ tumor microenvironment in pancreatic cancers with poor sensitivity to chemotherapy and dismal patient outcome [106]. Notably, a study investigating ROS-induced senescent SW480 colorectal cancer cells, which possess SASP features similar to that of p16^INK4A positive colorectal cancer cells in vivo, showed inhibition of CD8⁺ T-cell migration [107]. Subsequent investigations found that CXCL12 (C-X-C motif chemokine 12) secreted by the senescent cells, induced loss of CXCR4 CD8⁺ T cells and consequently impaired migration. Interestingly, physiological levels of CXCL12 could still mediate CD8 T-cell chemotaxis, suggesting a mechanism of bi-directional cues exerted by CXCL12 based on concentration levels [108]. This further highlights the fact that overproduction of senescence-associated chemokines might not necessarily translate to immune infiltration, contributing further to the complexity of SASP.

On the other hand, especially in the context of pathological inflammatory environments associated with chronic disease, immune cells may pose collateral damage as part of an evolutionary trade‐off between the efficacy of the immune system and age‐related pathology, as observed during inflammaging [109]. Supporting this hypothesis in relation to inflammatory pathogenesis, interesting work from Joao Passos’ group in the context of acute liver injury revealed that neutrophils could cause ROS-dependent telomere dysfunction in non-immune cells, culminating in senescence [110]. Importantly, these neutrophils were found to be recruited by senescent p16^INK4A positive cells, suggesting a potential ROS-dependent mechanism by which paracrine senescence can be mediated. This could constitute a positive feedback loop where chronic senescent environment is reinforced [111]. Taken together, current evidence demonstrates that excessive ROS and oxidative stress environments created by senescent cells engender an immunosuppressive milieu to permit pro-tumorigenic tendencies (Fig. 2D). It is worth noting that this is an area of emerging research, and we propose that heterogeneity of SASP be considered while evaluating mechanisms underlying alterations in ROS-associated immune landscape.

4. Interventions that impinge on the redox-TIS axis

It is clear that the ability of ROS and associated redox signaling to modulate cell fates and senescent cell function makes it an attractive target for intervention. There are two broad interventional categories that define the current field of senotherapeutics: The first class of compounds, the senolytics, aim to selectively induce cell death of senescent cells. The second, the senomorphics (or senostatics), is focused on modulation of the proinflammatory senescent secretome or SASP. In the following sections, we discuss potential to leverage on ROS and redox signaling as a means of achieving senolytic or senomorphic function, as well as describe some recent examples of compounds that have exploited the senescent redox state as part of interventional strategies.

4.1. Leveraging on ROS and redox signaling for broad-spectrum senolytic function

It is now well-established that senescent cells rely on constitutive upregulation of the anti-apoptotic BCL family of proteins for maintenance of cell viability. Indeed, BCL2 antagonists, such as the well-known ABT-263 (navitoclax), have been widely used as senolytics in vitro and in vivo to exploit this senescence-associated vulnerability [112]. This approach, however, has been met with several challenges and highlights a caveat. For example, cells lacking BAK and BAX are known to be resistant to ABT-263 [113]. In addition, paracrine senescent cells have been shown to express reduced BCL-2 compared to primary senescent cells [114]. This accounts for higher incidence of resistance against BCL-2 inhibitor in paracrine senescent cells compared to primary senescent cells, essentially highlighting that diversity of the senescent phenotype may impair therapeutic efficacy.

Leveraging on the senescent cell redox signaling network as a way of activating cell death pathways may prove advantageous in tackling the heterogeneity of senescent cells. Indeed, several compounds have been investigated for their ability to target specific players in the senescent redox signaling network to achieve selective activation of programmed cell death.

The natural product piperlongumine (PL) for instance, was identified from a library screen using irradiation-induced senescent WI-38 fibroblast cells to have the ability to selectively kill senescent cells [115]. Compared to other senolytic agents, PL was found to confer low toxicity, possess an excellent pharmacokinetic-pharmacodynamic (PK/PD) profile and oral bioavailability [116]. Further investigation using probe-based pulldown and mass spectrometry proteomic analysis identified the protein oxidation resistance 1 (OXR1) to be one of the molecular targets bound by PL [117]. Expression of OXR1 was found to be significantly higher in senescent WI-38 cells, compared to non-senescent cells. Subsequent experiments further demonstrated that PL induced the proteasomal degradation of OXR1. Knockdown of OXR1 in senescent cells was found to mimic the senolytic function of PL, causing cells to undergo apoptosis via ROS accumulation and downregulation of antioxidant enzymes such as GPx2 and CAT. Importantly, knockdown of OXR1 selectively sensitized senescent cells, but not non-senescent cells to H₂O₂-induced apoptosis [117], suggesting a rewired redox signaling network that relied heavily on OXR1 for tolerance of oxidative stress. In short, leveraging on OXR1 as a target can be beneficial for the development of novel senolytics. This could then be tested in TIS cancer cells to possibly improve the outcome of cancer chemotherapies, which brings us to the next example.

The synthetic flavonoid GL-V9 (5-hydroxy-8-methoxy-2-phenyl-7-(4-(pyrrolidin-1-yl) butoxy)-4-H-chromen-4-one), derived from the natural product wogonin, demonstrated ROS-dependent senolytic activity in chemotherapy-treated TIS MDA-MB-231 breast cancer cells and replicatively-senescent mouse embryonic fibroblasts (MEFs) [118]. Interestingly, GL-V9 was found to mimic the lysosomal alkalinization activity of chloroquine in senescent cells, preventing acidification of the autolysosome, a key step required in autophagy to clear dysfunctional mitochondria [119]. Indeed, GL-V9-treated senescent MDA-MB-231 cells was shown to possess increased mitochondrial abundance as observed with Mito-Tracker Green. Further findings demonstrated that the elevated levels of ROS driven by dysfunctional mitochondria clearance eventually contributed to a ROS-dependent apoptotic pathway, underlying the basis for the senolytic activity of GL-V9 [118]. In this way, GL-V9 was postulated to be a promising senolytic drug beneficial in the elimination of TIS cells.

Aside from ROS-dependent apoptotic induction, the induction of ferroptosis leveraging on impaired iron homeostasis represents an alternative approach to selectively induce cell death of senescent cells. For instance, a recent study found that the ferroptosis inducer FIN56 could selectively ablate both primary and paracrine senescent cells [114]. Subsequent investigation revealed reliance on glutathione peroxidase GPx4 in senescent cells in causing the ferroptotic cell death. Importantly, FIN56 was also shown to exhibit similar efficacy in both paracrine and primary senescent cells, supporting its potential use as a broad-spectrum senolytic agent.

4.2. Leveraging on ROS and redox signaling to achieve synergistic effect of senomorphic function

Senescent cells are known to play beneficial roles in normal physiological function, such as in wound healing and development [120,121]. Elimination of these physiologically-relevant senescent cells have been found to be detrimental. An early observation of this was reported by Judith Campisi's group, where elimination of p16⁺ cells by ganciclovir retarded cutaneous wound repair [120]. In a more recent study, the elimination of sentine1 p16⁺ cells in the basement membrane of the lung by dasatinib and quercetin was found to reduce regeneration of airway stem cells [122].

However, there are caveats to the use of senolytic agents. A well-known example is ABT-263 which can induce severe side-effects such as nausea, anemia, neutropenia and thrombocytopenia. Thrombocytopenia for instance, can arise due to the unintended off-target lytic effects on non-senescent cells such as circulating platelets and T lymphocytes since these also rely on Bcl-2 family proteins for survival [123]. Concurrent dosing with chemotherapy or higher doses of ABT-263 may also increase the likelihood and intensity of these adverse reactions. In addition, it is crucial to recognize that senescent cells are highly immunogenic, a property that can be exploited for activation of anti-tumor immune response [56]. Use of a senomorphic approach to dampen or suppress the effect of SASP without eliminating senescent cells may address some of the concerns associated with administration of senolytics.

Interestingly, while most senomorphics directly target drivers of SASP such as NF-kB, p38MAPK, and mTOR [51], some senomorphic drugs may also simultaneously alleviate other senescence-associated processes by modulating ROS and the redox state. Metformin, for instance, has been shown to not only be a potent SASP modulator, but also a chemosensitizer. In NSCLC cells, metformin treatment was shown to be able to reverse oxidative stress induced chemoresistance via downregulation of Nrf2, as well as its subsequent downstream antioxidant proteins [124]. Mechanistically, metformin was found to induce dephosphorylation of Nrf2 and restore its polyubiquitination, thereby accelerating its proteasomal degradation. Treatment in conjunction with the ROS scavenger NAC appeared to blunt this chemosensitizing activity by restoring antioxidant activity. Maintenance of Nrf2 expression in patients after neoadjuvant chemotherapy was also found to correlate with poor survival and chemoresistant phenotype [124]. Consistent with the role of Nrf2 as a key redox signaling transcription factor, rapamycin has also been shown to increase tolerance against oxidative stress-induced senescence that was Nrf2-dependent [125]. This senescence prevention capability was also found to be concomitant with activation of autophagy. In another study, rapamycin loaded in ZIF-8 (Zeolitic imidazolate framework-8) nanoparticles was found to be beneficial for direct tumor killing and alleviating doxorubicin chemoresistance via ZIF8-stimulated ROS production and mTOR inhibition [126]. Further to this, while rapamycin is not considered a cytotoxic agent per se, there is evidence that synergistic use of rapamycin with other compounds could elicit therapeutic apoptotic effect. As an example, a study using a gastric cancer mouse model explored use of synergistic treatment with rapamycin and EF24, a curcumin analog [127]. Surprisingly, EF24 was found to enhance rapamycin-induced ROS production by increasing ER stress and mitochondrial dysfunction. Interestingly and importantly, ROS production by rapamycin was found to be independent of its mTOR inhibition, suggesting that in this synergistic context, the senomorphic activity of rapamycin dependent on mTOR inhibition would still be likely intact. Taken together, one could extrapolate that exploiting ROS and redox state modulation could potentially allow flexible and synergistic usage of senomorphics, as chemosensitizers or apoptosis inducers for instance, while still retaining their SASP-modulating activity via targeting of SASP drivers.

5. Therapy-induced senescence escape – a new frontier in redox biology

Traditionally, therapy-induced senescence (TIS) has been defined as a stable form of growth arrest following therapy such as radiation or chemotherapy. An emerging area of research that has challenged this notion and gained traction in recent years is “senescence escape”. Here, a small subpopulation of TIS cells acquire the ability to escape senescence [128,129]. It is important to note that these senescence-escaped cells which re-enter the cell cycle, differ phenotypically to non-senescent cells prior to escape (Fig. 2E). Initial observations of this was first reported by Daniel Wu's group in 2005, where a subpopulation of TIS p53-null, p16-deficient NSCLC H1299 cells escaped replicative arrest and re-entered the cell cycle [130]. These escaped cells were found to express abnormally high levels of the cyclin-dependent kinase Cdc2/Cdk1. In a separate study, TIS HCT116 colon cancer cells treated with autophagy inhibitor Bafilomycin A1 (Baf1A) was shown to reactivate its proliferative potential [131]. These reactivated cells were found to possess increased stemness, as characterized by increased NANOG expression, CD24⁺ population, and colony formation. Consistently, analyses of senescent and non-senescent B-cell lymphomas from Eμ-Myc transgenic displayed upregulation of a stem-cell signature, concomitant with activation of WNT signaling. These signatures were also observed in cells which re-enter the cell cycle after senescence [132], indicating that certain aspects, such as increased stemness, were not lost after release from senescence arrest. Interestingly, this was also observed in oncogene-induced senescence-escaped cells, where senescence-associated chromatin marks persisted as epigenetic memory in post-senescent cells [133]. This supported the idea that senescence-associated epigenetic reprogramming could be inherited by post-senescent cells, forming a “senescence scar”.

Importantly, from the redox perspective, the redox state in “senescence-escapees” completely differs from cells in the senescence state. While the senescence state is characterized by high ROS levels and a pro-oxidant microenvironment, a study using TIS-escaped breast cancer cells demonstrated a redox environment with low ROS, concomitant with overexpression of superoxide dismutase 1/2 (SOD1/2), glutathione peroxidase 1/2 (GPx1/2) and stabilization of Nrf2 [134]. This low ROS environment appears to mimic that of cancer stem cells [135]. Indeed, the stabilization of Nrf2 in TIS-escaped cells was found to be p21-mediated and necessary for tumor stem cell enrichment [134].

It is unclear why the senescence-associated pro-oxidant redox state is not preserved in senescence-escaped cells. One could speculate that transition to a low ROS environment is intertwined with CDK-driven cell cycle-related processes and checkpoints that differ between senescent and senescence-escaped cells. It is also possible that such redox reprogramming may be important in triggering the initiation of senescence escape and mediating early steps of cell cycle re-entry. A time-resolved study tracking the dynamics of the redox state, specifically during the window of senescence escape transition, would be key in addressing this. It may be beneficial to approach this via a multi-omics approach to dissect any accompanying alterations in epigenetics marks and metabolic nodes that may collaboratively work with redox changes to mediate cell cycle re-entry post-senescence escape.

6. Conclusion

Examining therapy-induced senescence through a redox lens opens up an exciting and promising route for understanding senescence heterogeneity to combat cancer progression and therapy resistance. We envision this will not only identify suitable targetable nodes, but will also inform strategies for the development of novel broad-spectrum senotherapeutics. Integrating this with the use of time-resolved single-cell RNA-sequencing combined with spatiotemporal multi-omics, should also provide insights into the distribution of senescent cells and its redox signalling dynamics. In addition, tumor stem cell-like cells with low ROS levels offers vulnerabilities to exploit for the emerging field of senescence escape (Fig. 2E). As TIS-escape represents a driver of tumor recurrence [136], it would be paramount to elucidate the role of ROS and redox state dynamics during this process. We look forward to novel and exciting discoveries in this regard.

Funding

This work was supported by the National University of Singapore (NUHSRO/2020/114).

CRediT authorship contribution statement

Matius Robert: Writing – review & editing, Writing – original draft. Brian K. Kennedy: Writing – review & editing, Supervision, Funding acquisition. Karen C. Crasta: Writing – review & editing, Writing – original draft, Conceptualization.

Declaration of competing interest

The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.

Data availability

No data was used for the research described in the article.