Reactive Oxygen Species: Do They Play a Role in Adaptive Immunity?

The immune system protects the host from a plethora of microorganisms and toxins through its unique ability to distinguish self from non-self. To perform this delicate but essential task, the immune system relies on two lines of defense. The innate immune system, which is by nature fast acting, represents the first line of defense. It involves anatomical barriers, physiological factors as well as a subset of haematopoietically-derived cells generically call leukocytes. Activation of the innate immune response leads to a state of inflammation that serves to both warn about and combat the ongoing infection and delivers the antigenic information of the invading pathogens to initiate the slower but highly potent and specific second line of defense, the adaptive immune system. The adaptive immune response calls on T lymphocytes as well as the B lymphocytes essential for the elimination of pathogens and the establishment of the immunological memory. Reactive oxygen species (ROS) have been implicated in many aspects of the immune responses to pathogens, mostly in innate immune functions, such as the respiratory burst and inflammasome activation. Here in this mini review, we focus on the role of ROS in adaptive immunity. We examine how ROS contribute to T-cell biology and discuss whether this activity can be extrapolated to B cells.

Interestingly, although even at high concentrations O 2-. , NO and H 2 O 2 are not directly damaging to cells, they react with themselves or with metal ions to produce the extremely toxic secondary reactive species.OH, ONOOand HOCl. These secondary species are poorly controlled and rapidly and irreversibly react with virtually all classes of biomolecules causing oxidative damage. The accumulation of ROS can lead to a state of oxidative stress when the endogenous antioxidant machinery of the cell is overwhelmed (19)(20)(21)(22)(23)(24). Consequently, the cells accumulate oxidative damage within the DNA, lipids and proteins, causing cellular dysfunction and cell death (19)(20)(21)(22)(23). Excessive ROS production plays a major role in the initiation and amplification of cell death by modulating many signaling pathways. Consequently, ROS levels are contributing determinants for various forms of cell death, including apoptosis, necrosis/necroptosis, ferroptosis, pyroptosis and autophagic cell death (25)(26)(27)(28)(29)(30)(31)(32).
The immune system has the unique ability to distinguish self from non-self to protect the host organisms from a plethora of microorganisms and toxins (33)(34)(35). It eliminates foreign entities (pathogens and toxins) but tolerates the self (host's own tissues) and its associated microbiota (33,36,37). The innate immune system, the components of which are already present before any pathogenic intrusion, is fast acting. It relies on anatomical barriers (the skin and the mucosa lining the respiratory, gastrointestinal and urogenital tracts) to prevent foreign entities from entering the organism (33,34). These anatomical barriers are reinforced by soluble factors (complement system, pentraxins, collectins and the defensins antimicrobial peptides) as well as by leukocytes (macrophages, dendritic cells, mast cells, neutrophils, eosinophils, natural killer [NK] cells) that neutralize pathogens or kill the infected cells (33,34). The innate immune system is activated by the recognition of antigenic determinants common to a wide spectrum of microbes (the pathogen associated molecular patterns [PAMP]) and leads to a state of inflammation to alert and combat the ongoing infection (33,34,38). Importantly, the activated innate immune system delivers the antigenic information of the invading pathogens to activate the slower but highly potent and specific second line of defense known as the adaptive immune system. The adaptive immune response calls on T lymphocytes and B lymphocytes as, respectively, the effectors of the cellular adaptive immune response and as the antibody-producing cells with the essential functions of eliminating pathogens and establishing immunological memory (35,39,40).
ROS have been implicated in many aspects of the immune response to pathogens mainly related to innate immunity. Indeed, they have been proposed to be the common determinant of inflammasome activation, which is critical in the inflammatory process and thus necessary for an efficient immune response. ROS are also essential for pathogen killing by phagocytic cells, as illustrated in chronic granulomatous disease (CGD), an inherited disorder of NADPH oxidase characterized by recurrent and severe bacterial and fungal infections as phagocytes from these patients cannot do the respiratory burst. Here in this mini review, we focus on the role of ROS in adaptive immunity. We examine how ROS contribute to T-cell biology and briefly discuss whether these activities can be extrapolated to B cells.

ROS CONTRIBUTE TO TCR SIGNALING
Actually, within minutes of TCR stimulation, there is a production of both O 2-. and H 2 O 2 , which seems to originate from different TCR signaling pathways (48). Specifically, studies in Jurkat T cells showed that ROS increase the phosphorylation and activity of p56lck, ZAP-70, protein kinase C (PKC) and intracellular Ca 2+ levels ( Figure 1) (49)(50)(51). This results in a phosphoinositide-3 kinase (PI3K)/AKT/mTOR-, Myc-and ERRa-dependent augmentation of the global metabolism (52)(53)(54)(55)(56). It was further demonstrated that T cells from p47 phox -deficient mice do not undergo TCR-induced H 2 O 2 production, whereas TCR-induced O 2-. is unaffected following TCR stimulation in these cells, indicating that H 2 O 2 originates from a lymphocyte-encoded NADPH oxidase (NOX) (57). Additionally, using superoxide scavenger treatments or superoxide deficiency in OT-II.Ncf1 m1J mice having CD4 T cell-specific superoxide deficiency, it was shown that superoxide is necessary for Th1 responses as well as IL-12R and proinflammatory chemokine ligand expression in CD4 T cells (58,59). In fact, in T cells, ROS contribute not only to proximal but also to distal signaling pathways and modulate the activities of transcription factors NFAT, AP-1, and NF-kB to induce gene expression (60,61). Activated T cells take up large amounts of glucose and produce lactate, indicating that they are primarily glycolytic (61,62). Interestingly, during CD4 T-cell stimulation, mitochondrial oxygen consumption increases as an indication that mitochondrial function is also important for T-cell activation not only to support the glutamine requirement of these cells but also as a source of ROS ( Figure 1) (61,63). It was even shown that mitochondrial ROS specifically from respiratory chain complex III are required for CD4+ and CD8+ T cell expansion in vivo (61). Deletion of the Rieske iron sulfur protein (RISP), a subunit of mitochondrial complex III in T cells, resulted in a lack of oxidative phosphorylation and complex III-dependent ROS production and no expression of IL-2 upon CD3/CD28 stimulation (61). This phenotype was rescued by the addition of exogenous H 2 O 2 , clearly demonstrating the ROS requirement for full activation of the CD4 T cells (61). In this context, mitochondrial ROS were downstream of the TCR-mediated cytosolic and mitochondrial calcium increase, in agreement with the calcium dependency of mitochondrial TCA cycle dehydrogenases that fuel the electron transport chain (ETC) to increase mitochondrial membrane potential and ROS production (64,65). Interestingly, mitochondrial ROS are also necessary for CD8 T-cell activation, as inhibition of the respiratory chain complex I decreases the production of H 2 O 2 , calcium flux, and ERK1/2 phosphorylation and impairs CD8 T-cell activation and proliferation (66). Complex I inhibition not only decreases activation of naive cells but also decreases interferon (IFN)-g and TNF-a production as well as degranulation of effector and memory CD8+ T cells isolated from lymphocytic choriomeningitis virus-infected mice (66).
It is worth noting that some studies suggest mitochondria more than NADPH oxidase are the essential source of ROS involved in the activation process (61)(62)(63)(64)(65)(66). This apparent discrepancy could come from the activation status (naïve/primed) or developmental state (CD8/CD4, T helper 1/2/9/17 or regulatory CD4 T cells) considered, which might have different ROS requirements for full activation (67). Nevertheless, collectively these results showed that ROS play a role in the activation and maturation of both CD8 and CD4 T cells (57,61,66,68,69). Mechanistically, by inhibiting phosphatases, ROS might tilt the balance toward phosphorylation, ultimately potentiating the activation of kinase cascades and transcription factors such as NFAT, which is critical for IL-2 production (61).
Interesting, although they are required, ROS levels must be kept in check by the glutathione-dependent antioxidant machinery (70,71). Indeed, preventing glutathione (GSH) production impairs FIGURE 1 | Endogenous ROS contribute to T and B cell receptor signaling. The engagement of the B-cell receptor (BCR) or T-cell receptor (TCR) and their respective co-receptors CD28 and CD19/CD21 triggers intracellular phosphorylation signaling cascades resulting in the activation of transcription factors AP1, NF-kB, NFAT, OCA-B/OBF-1 IRF-4, which are critical for T and B lymphocyte activation. In T cell, dark grey, TCR/CD28 stimulation induces the phosphorylation and activation of the kinases p56lck (LCK) and ZAP-70, the phospholipase Cg (PLCg). CD28 recruits growth factor receptor bound protein 2 (Grb2), which docks the complex formed by SH2 domain containing leukocyte protein of 76kDa (SLP-76)/Linker for activation of T-cells (LAT)/signal transducer protein Vav1. The latter complex recruitment closer to the membrane facilitating its activation by ZAP70. PLCg generates inositol 3 phosphate (IP3) to mobilize intracellular Ca2 + stores resultingin the activation of the phosphatase calcineurin. PLCg also generates diacylglycerol (DAC) to activate protein kinase C (PKC), c-Jun N-terminal Kinase (JNK) and mitogen-activated protein kinase/extracellular signal-regulated kinases (MEK/ERK) cascade. Calcineurin dephosphorylates nuclear factor of activated T cells (NFAT), allowing its nuclear translocation. PKC allows NF-kB nuclear translocation by removing of the inhibitor (IkB), while JNK and ERK activate AP1. In B cells, light grey, BCR/CD19/CD21 stimulation initiates a phosphorylation cascade starting with the activation of the kinases Lyn and SYK, leading to the activation of B cell linker protein (BLNK), B Cell adaptor molecule for phosphoinositide 3-Kinase (PI3K) (BCAP) and PLCg, ultimately resulting in PKC, JNK and ERK activation. Both TCR and BCR signaling potentiate mitochondrial respiration and activate metabolic pathways through their action on the complex formed by PI3K, protein kinase B (PKB/AKT) and mammalian target of rapamycin (mTOR). These signaling cascades are potentiated by ROS (H2O2 and O2-.) from NADPH oxidases (NOX/DUAOX) and from the mitochondria (red arrows). In some instances, NOX triggers the oxidative modification of ZAP70 and LCK to precipitate their degradation and blunt activation (dashed bloc red line).
T-cell activation, as the energy and anabolic demands of these cells can no longer be met (72). GSH deficiency alters mammalian target of rapamycin (mTOR) and Myc activation, preventing the metabolic switch to glycolysis and glutaminolysis in an adenosine monophosphate-activated protein kinase (AMPK)-dependent manner (62,73). Paradoxically, it was reported that ROS can also downregulate T-cell activation by regulating the degradation of signaling molecules and the activation of cytoskeletal proteins (74,75). To prevent excessive ROS from triggering the mitochondrial permeability transition pore (PTP) opening and causing cell death, CD4 T cells upregulate microRNA (miR)-23a, which targets peptidylprolyl isomerase F (PPIF or Cyclophilin D), a key regulator of the PTP. The reduction in PPIF is expected to keep the mitochondrial PTP closed and reduce the escape of ROS, preserving CD4+ T-cell survival during the early hypermetabolic and inflammatory state of the activation process (76).

ROS CONTRIBUTE TO BCR SIGNALING
Unlike that of T cells, B-cell metabolism is less well characterized. However, it was recently demonstrated that energy demand is elevated during antigen (Ag)-driven proliferation and differentiation (77,78). B-cell stimulation with lipopolysaccharide (LPS) or anti-immunoglobulin M (IgM) antibodies drastically increases glucose import, although increased mitochondrial respiration still occurs, suggesting that here again mitochondrial function is important ( Figure 1) (79,80). Interestingly, ROS production in response to BCR stimulation occurs in two waves. An early NADPH oxidase 2-dependent ROS increase take place within minutes of BCR stimulation, and a second wave of increasing ROS levels from the mitochondria occurs at later time point (81). B cells deficient in early Nox2-dependent ROS production have no defects in proximal BCR signaling, cell activation or the ability to mount an antibody response following T cell-dependent Ag stimulation (81). However, preventing the later ROS increase attenuates BCR-dependent signaling, leading to defective activation, proliferation and response to BCR stimulation (81). These results indicate that the continuous production of mitochondrial ROS at later times during the activation process is critical for BCR signaling and optimal Ag-induced B-cell activation and proliferation, in agreement with findings from gene set enrichment analysis showing upregulation of OXPHOS and the TCA cycle in activated B cells (79,81).
Lymphocyte activation clearly requires a metabolic reprograming for a diversification of the source of energy and biosynthetic building blocks (62,82,83). Therefore, one could wonder whether the observed increase in ROS during cell activation could be a consequence of this metabolic reprograming and reciprocally any genetic manipulation of the ROS input could also alter the metabolism of these cells, questioning the real significance of the ROS in the activation process? This latter possibility is readily excluded by the fact that exogenous H 2 O 2 can rescue lymphocyte activation in the context of genetic ablation of complex III (61,66,79,81). Taken together, these results clearly demonstrate that cell intrinsic ROS signaling participates in the activation processes of both B and T lymphocytes.

ROS AND LYMPHOCYTE VIABILITY
We have seen that one of the proximal events following TCR signaling is an increase in ROS production both in the form of O − 2 and H 2 O 2 . One direct consequence of the increase in these ROS in the context of T cell blasts is the initiation of activationinduced cell death (AICD) following the induction of FasL expression (84,85). In fact, downstream of the TCR engagement, activated ZAP70 phosphorylates liker of activated T cells (LAT), which docks phospholipase Cg1 that generates inositol 3 phosphate (IP3) and diacylglycerol (DAG) (84). DAG activates protein kinase C q (PKC q ) and its translocation into the mitochondria to enhance the production of ROS in a mitochondrial complex I-dependent manner, which is necessary for the expression of the ligand of the death receptor Fas (FasL) (84). FasL engages Fas receptor and triggers apoptotic cell death, a process where mitochondrial ROS further play a role, as it was later shown that caspase 3 can induced a ROSdependent cell death by cleaving the respiratory chain complex I subunit NDUFS1 (30). We have also shown that ROS potentiate the apoptotic cascade by amplifying the release of apoptogenic factor from the mitochondria and increasing oligonucleosomal DNA fragmentation (86). Moreover, exposure to exogenous H 2 O 2 differentially affects T-cell viability, according to their subset and maturation status. Central memory and effector memory T cells are more sensitive to H 2 O 2 followed by naïve T cells, among which the CD8+ effector memory T-cell compartment is more sensitive to even low doses of H 2 O 2 ( Figure 2) (87,88). In this context, exogenous H 2 O 2 exposure triggers cell death in a mitochondrial pathway-dependent manner (87,89). T cells treated with H 2 O 2 experience the opening of the mitochondrial permeability transition pore (PTP), a rapid decrease in the mitochondrial transmembrane potential DYm, and the release of cytochrome C (89). Blocking the mitochondrial PTP opening or interference with the respiratory electron transport chain with rotenone or menadione abrogated H 2 O 2 cytotoxicity (89). Interestingly, antimycin A, a respiratory chain complex III inhibitor that increases the release of mitochondrial ROS, enhanced apoptosis, while overexpression of Bcl-2 and the viral antiapoptotic proteins BHRF-1 and E1B 19K counteracted H 2 O 2induced T-cell apoptosis (89). Furthermore, inhibition of the transcription factor NF-kB protected cells from H 2 O 2-induced cell death in a process that likely relies on the expression of a death effector gene such as p53 (89). Paradoxically, T regulatory cells, which have lower intracellular ROS levels, are particularly protected from H 2 O 2 -dependent inhibition of suppressive function and H 2 O 2 -induced death (90). Taken together, the higher sensitivity of effector memory CD8 T cells combined with the reduced susceptibility of T regulatory cells to H 2 O 2 -induced death suggest that the oxidized tumor microenvironment (TME) may be a particularly inhospitable site for CD8 T cells and detrimental to T cell-based adoptive cell transfer therapies. This is even more critical as effector memory T cells are the primary phenotype of cells administered during such therapeutic protocols. Thus, research is needed to determine the effect of the TME of chimeric antigen receptor (CAR)-T cell therapies. Beyond apoptosis, ROS critically regulate T-cell viability through the induction of ferroptosis. Both Agspecific CD8+ and CD4+ T cells deficient for glutathione peroxidase 4 (Gpx4) are unable to expand or protect against viral and parasitic infection (91). This phenotype can be rescued by dietary vitamin E supplementation, indicating that lipid peroxidation-dependent ferroptosis plays a critical role in the T-cell depletion during these antigenic challenges (91).
Exposure of B cells to relatively massive doses (100-250 µM) of H 2 O 2 has little effect on cell viability in the short term (30 minutes). However, 5 days later, half of the cells have died, even in the presence of CD40 stimulation (92). These doses of H 2 O 2 although massive are still in the physiological range, as it was estimated that in the proximity of activated neutrophils and macrophages the H 2 O 2 concentration can reach the 100s of µM (93)(94)(95). Moreover, this exogenous H 2 O 2 exposure completely suppresses the ability to CD40 stimulation to trigger antibody production ( Figure 2) (92). In fact, exogenous H 2 O 2 exposure dissociates TRAF2 from CD40, leading to inefficient IKK phosphorylation, IkBa degradation, and NF-kB activation, which altogether severely compromises B-cell activation (92).

ROS-MEDIATED LYMPHOCYTE DYSFUNCTION
As serial killers, cytotoxic T lymphocytes and NK cells must recognize, engage, kill, and detach from the first target cell before moving to the next (96)(97)(98). Detachment of the effector cell from its target requires its repolarization through the reorganization of its cytoskeleton for the disassembly of the first immunological synapse (99). Effector killer cells are quite sensitive to the redox status of their immediate environment (100)(101)(102). Oxidizing reagents curb killer cell degranulation, consequently inhibiting their cytotoxicity (102,103). Interestingly, it was reported that oxidized low-density lipoprotein (ox-LDL), used as oxidizing agent, inhibits killer cell degranulation (102). Pretreatment of NK cells or co-incubation of NK/target cell conjugates with non-cytotoxic doses of ox-LDL markedly and significantly reduces the NK cytotoxic activity against U937 tumor cells (102). This reduced NK cell cytotoxicity is not the consequence of their inability to engage the target cells, because the number of NK:target cell conjugates was not affected nor were the expression levels of CD11a, CD11b, CD18, CD2, and CD62L, key adhesion molecules involved in the effector-target cell interaction (102). Mechanistically, ox-LDL triggers a partial depolarization of the microtubule network that is critical for the polarization of the cytotoxic granules toward the immunological synapse formed between the effector and target cells. Similarly, exposure of mitogen-stimulated peripheral blood mononuclear cells (PBMCs) to ox-LDL reduces their production of TNF-a, IFNg and IL-12 (102). Likewise, exogenous and endogenous nitric oxide (NO) inhibits degranulation of lymphokine activated killer (LAK) cells (103). NO inhibits LAK cell exocytosis in part by decreasing the expression of RAS, a critical component of the exocytic signaling cascade, following destabilization of RAS mRNA (103). NO acts by interfering with the mRNA-stabilizing factor HuR, which binds and stabilizes AU-rich elements of the mRNA 3'-untranslated region (104). It was further demonstrated that ROS induced oxidation of the C-terminal portion of the TCRz chain, and the membrane proximal domain of p56(lck) and cofilin promote their degradation or inactivation, suggesting that ROS can also curb the TCR signaling cascade (74,75). Similarly, increasing evidence suggests that the dynamics of the immunological synapse can be regulated by ROS through their direct or indirect effects via plasma membrane polarization on calcium signaling and effector cell cytoskeletal reorganization (99,102,(105)(106)(107). As stated earlier, activation of the lymphocytes after engagement of their receptor initiates a phosphorylation cascade, resulting in, among other things, the mobilization of intracellular Ca 2+ stores, which is essential for the gene expression crucial for lymphocyte activation and the development of adaptive immunity (99,(108)(109)(110)(111). Depletion of Ca 2+ stored in the endoplasmic reticulum triggers store-operated Ca 2+ entry (SOCE). Compared to Orai1, the Ca 2+ channel involved in SOCE, Orai3 lacks the redox-sensitive cysteine 195 and therefore is redox-insensitive. Co-expression of Orail3 with Orai1 reduces SOCE sensitivity to ROS inhibition. Consequently, it is not surprising that T lymphocytes display upregulated Orai3 expression during their differentiation into effector T cells. This means that the modulation of the Orai1:Orai3 ratio could be a possible mechanism by which effector T lymphocytes preserve some responsiveness in oxidized environments, such as the hypoxic TME or inflamed tissues (106,112). ROS also regulate the effector function of B cells. Overexpression of a phosphorylation-defective mutant of succinate dehydrogenase A to model excessive mitochondrial ROS production suppresses Ig production, germinal center (GC) formation, and GC B-cell proliferation following an encounter with T cell-dependent Ag. Excessive mitochondrial ROS production also suppresses Ig production against T cell-independent Ag (113) as well as BCRdependent Lyn, Btk, and PLCg2 phosphorylation and CD19 expression. From these collective results, it was hypothesized that excessive mitochondrial ROS dampen B-cell activation most likely by reducing CD19 expression (113). Overall, it seems that mild to moderate exposure to exogenous ROS affects the lymphocyte effector functions, while acute exposure affects their viability ( Figure 2). The situation is complicated by the fact that the threshold between mild, moderate and acute exposure strongly depends upon the lymphocyte subset and maturation status.

ROS IN IMMUNE CELL DYSFUNCTION: THE CASE OF AUTOIMMUNITY
As we have discussed earlier, endogenous ROS contribute to lymphocyte activation; however, depending of the lymphocyte activation and/or differentiation status, exogenous ROS can affect their effector function and viability. Thus, we also would like to discuss whether ROS could play a role in the pathogenesis of immune-related disorders such as autoimmunity where the wrath of the immune response mistargets self-antigens (autoantigens). For instance, abnormal functions of T helper (Th)-17 cells, which in a normal setting are essential to fight against extracellular bacteria (114)(115)(116), are involved in multiple chronic inflammatory disorders such as psoriasis, multiple sclerosis (MS), inflammatory bowel disease (IBD), Sjögren's syndrome and rheumatoid arthritis (114). Interestingly, using sublethal doses of oligomycin A, an inhibitor of the respiratory chain ATP synthase/complex V, it was shown that mitochondrial oxidative phosphorylation (OXPHOS) plays a pivotal role in for Th17-mediated autoimmunity (117). Oligomycin treatment abolished Th17 pathogenicity by altering the expression of Th17 pathogenic signature genes, such as transforming growth factor beta 3 (TGFb3), interleukin 23 receptor (IL-23R), signal transducer and activator of transcription 4 (Stat4), and G protein-coupled receptor 65 (Gpr65), while genes inversely associated with Th17 pathogenicity such as suppressor of cytokine signaling 3 (Socs3) and IL-10R subunit alpha (IL-10Ra) were upregulated (117). Although the authors did not directly test this possibility, it is very likely that mitochondrial ROS could be involved in this process, as in their experimental condition the oligomycin treatment severely suppresses the basal mitochondrial oxygen consumption rate, which could result in a severe reduction in mitochondrial ROS, indicating that ROS could protect against the pathogenicity of Th17 cells (117). This agrees with previously work by Tse and coworkers showing that prevention of O − 2 production by macrophages and T cells skews T-cell polarization toward Th17 (118). Although they used a model of NOX-deficiency, collectively these results agree that the absence of ROS alters T-cell lineage commitment, pointing to a role for superoxide in the modulation Th17 versus Th1 T cell responses (118). From a more clinical stand point, it was shown that the hypomorphic allele of the Ncf1 gene encoding for p47 phox , a subunit of NOX2, is one of the strongest genetic predispositions for autoimmune arthritis, autoimmune encephalomyelitis and systemic lupus erythematosus (SLE), which are associated with increased numbers of autoreactive T cells (119)(120)(121)(122). Interestingly, results from two clinical trials, one using N acetyl cysteine the other Sirolimus to modulate respectively cellular GSH content and mTOR activity resulted in the improvement of SLE condition suggesting that modulation of the mitochondrial ROS output could also contribute to regulate pro inflammatory T cell development (123,124). Furthermore, by downmodulating the efficacy of antigen processing, ROS may further contribute to limiting the activation of autoreactive lymphocytes. In this regard, in the early stage of the processes, ROS should not simply be considered as effectors to eliminate invading pathogens, but also as modulators to fine-tune the inflammatory response depending on the timing, the site and the level of their production (125,126). By contrast, in the context of dysregulated and prolonged chronic inflammation, the local microenvironment is characterized by a low nutrient levels, increased lactate production, decreased pH, hypoxia and an increase level of ROS, which collectively lead to excessive tissue destruction. At this later stage, this excessive tissue destruction could promote the accessibility to cryptic neoantigens favoring the progression and exacerbation of autoimmunity (126).

ROS IN THE PROCESS OF CYTOTOXIC LYMPHOCYTE KILLING
Cytotoxic lymphocytes are particularly efficient at eliminating target cancer cells and virally infected cells. They mainly used the cytotoxic granule pathway relying on the degranulation of the pore forming protein perforin and a family of five serine proteases call granzymes in human (127)(128)(129). Although the granzymes trigger very distinct cell death pathways, we found that granzyme A and B (GA and GB) share the ability to induce ROS-dependent cell death. It was demonstrated that GA induces ROS-dependent death that is independent of the mitochondrial outer membrane permeabilization (MOMP) and insensitive to BCL2 but has all the morphological features of apoptosis (127,(130)(131)(132)(133). We also showed that ROS are necessary for the rapid cell death induction by GB. We found that K562 cells treated with a sublytic concentration of perforin (P) and GB undergo a rapid increase in ROS production and cell death that is inhibited in the presence of the well-characterized antioxidants N-acetyl cysteine (NAC), superoxide scavenger MnTBAP, or the mitochondrial targeted superoxide scavenger MitoQ (134). Moreover, GB and P-induced ROS and cell death are completely absent in pseudo rho cells deficient for mitochondrial DNA (mtDNA) and therefore lacking a functional respiratory chain (134). Both GA and GB induce ROS release from isolated intact mitochondria in the absence of cytoplasmic fraction S100 (130). Using organelle proteomics and bioinformatics, we found that GA and GB cleave NDUFS3, NDUFV1, NDUFS1 and NDUFS2 iron-sulfur (Fe-S) clustercontaining subunits of the respiratory chain complex I (86,130,132,135). Cleavage of complex I subunits exposes iron sulfur clusters and dramatically increases electron leak from the respiratory chain, leading to a rapid and sustained mitocentric ROS production, loss of complex I, II, and III activities, disorganization of the respiratory chain, mitochondrial respiration impairment, and loss of mitochondrial cristae junctions (86,130,132,135,136). It is worth noting that another study has also suggested the contribution of NOX as source of ROS during GBmediated cell death (137). However, we found that, GB-mediated killing of mouse embryonic fibroblasts (MEFs) from NOX-deficient animals proceeds as in wild-type MEFs (86). GB induction of mitocentric ROS promotes apoptogenic factor release and oligonucleosomal DNA fragmentation (138,139). Although granzymes do not express a mitochondrial targeting signal, they enter the mitochondria independently from the TOM40 complex, the organelle entry gate, and use instead the SAM50 channel (136,140). SAM50 is the core channel of the mitochondrial sorting and assembly machinery dedicated to the insertion of de novo b-barrel proteins into the mitochondrial outer membrane (141)(142)(143)(144). Preventing the entry of granzymes into the target cell mitochondria alters their cytotoxicity. Using a model of human glioma, a very aggressive primary brain tumor for which there is no cure, we showed that granzyme mitochondrial entry is also essential for the reduction of tumor burden in vivo (136,140). Collectively, these interesting results also indicated that respiratory chain complex I is at the crosstalk of GA, GB and caspase 3, three different cell death pathways. Complex I targeting is also conserved across phylum from bacteria to mammals. In collaboration with the Walch's group we showed GA-and GB-mediated disruption of bacterial complex I is also a necessary step for bacterial death (145). The central role of complex I alteration during cell death suggests that it is a very important step whose full range of function has yet to be unraveled. For more about the antimicrobial action of the granzymes, we refer readers to the review of the oxidative and non-oxidative antimicrobial activities of the granzymes by Marilyne Lavergne on this same research topic.

CANCER, OXIDATIVE STRESS, AND CYTOTOXIC LYMPHOCYTES
Uncontrolled proliferation and neoplastic transformation come with enormous demands for energy and macromolecule building blocks. These demands impose a severe metabolic stress, requiring a striking reprogramming of the cancer cell metabolism (146,147). The resulting altered metabolism combined with the hypoxic nature of the TME is accompanied by marked production of ROS (148)(149)(150)(151). This overproduction of ROS activates the cellular antioxidant response based on enzymatic and non-enzymatic antioxidant molecules, which is under the transcriptional control of the transcription factor nuclear factor (erythroid-derived 2)-like 2 (NRF2). Three isoforms of superoxide dismutase (SOD), cytosolic CuZn-SOD (SOD1), mitochondrial Mn-SOD (SOD2), and extracellular EC-SOD (SOD3), are involved in the rapid dismutation of O 2into H 2 O 2 (15,16). The homotetrameric catalase converts H 2 O 2 into water using NADPH as a cofactor (15,17). The glutathione peroxidases (GPx) use glutathione (GSH) and reduce H 2 O 2 and lipid hydroperoxides (15,18). H 2 O 2 removal also involves thioredoxin (TRX), thioredoxin reductase (TRR), thioredoxin peroxidase (PRX) and glutaredoxins (15). The most abundant non-enzymatic antioxidant molecule in the cell is GSH, which participates in the reduction of H 2 O 2 into H 2 O and O 2 , and is thereby oxidized to form GSSG. GSSG is then recycled into GSH by glutathione reductase still using as electron donor NAD(P)H. GSH also maintains aqueous and lipophilic levels of the antioxidant ascorbic acid (vitamin C) and a-tocopherol (vitamin E), respectively. Nevertheless, when this antioxidant system is overwhelmed, the pro-oxidant/anti-oxidant equilibrium is lost, and a state of oxidative stress is reached where the cells accumulate oxidative damage in all type of macromolecules, including DNA, RNA, lipids and proteins, which could lead to cell death (152)(153)(154). Oxidative DNA modifications generate 8hydroxy-2′-deoxyguanosine, which contributes to the accumulation of mutations that enhance aging and carcinogenesis (155). Consequently, transformed cells adapt and reach new redox balance, and paradoxically, ROS instead of killing, stimulate tumor development and progression by promoting cell proliferation through their mitogenic action as activator of extracellular-regulated kinase 1/2 (ERK1/2). This induces ligandindependent receptor tyrosine kinase (RTK) activation, activating Src kinase, NF-kB and phosphatidylinositol-3 kinase (PI3K)/Akt, to enable evasion of apoptosis and anoikis as well as to induce metalloproteinase (MMP) release in the extracellular matrix to favor invasion and promote angiogenesis (156)(157)(158)(159)(160)(161). ROS also contribute to epithelial to mesenchymal transition (EMT), an important process in the metastatic dissemination of cancer cells (2). Importantly, in the TME, cancer cells reprogram other cells, such as cancer-associated fibroblasts (CAFs), endothelial cells and cancer-associated macrophages (CAMs), in a ROS-dependent manner to favor tumor progression. CAFs contribute to tumor growth by promoting the tumor angiogenesis by secreting VEGF and angiopoietin, by generating anti-apoptotic factors and by the secretion of chemokines (CCL2 and CCL5) and MMPs to promote the dissemination while blocking the immune response through the secretion of immunosuppressive cytokines IL-6, IL-10 and TGF-b (162)(163)(164). This marked production of ROS also alters the phenotype of innate immune cells infiltrating the tumor parenchyma, contributing to the noxious nature of the TME (165)(166)(167)(168). Interestingly, it is worth noting that a direct link exists between the environmental ROS of the TME and inflammation (169). Intracellular ROS may regulate EMT in a NF-kB-and hypoxia-inducible factor 1 (HIF-1a)-dependent manner in a process requiring the activity of cyclooxygenase-2 (COX-2), the first enzyme in the synthesis of prostaglandins, prostacyclin and thromboxanes including prostaglandin E2 (PGE2). This suggests that this oxidized microenvironment favors a state of chronic inflammation in the TME. As stated earlier cytotoxic lymphocytes (NK cells and cytotoxic T lymphocytes) play an essential role in the immune response against cancer (97,(170)(171)(172)(173)(174)(175)(176)(177). It is therefore not surprising that harnessing the power of these innate and adaptive cytotoxic immune cells during immune check point blockage (ICB) or CAR-T/NK cell immunotherapies has produced very encouraging results (178)(179)(180). As we have discussed earlier, NK cells and CD8+ effector memory T cells are particularly sensitive to even low doses of H 2 O 2 while the T regulatory cells are protected from H 2 O 2 -dependent inhibition of their suppressive function and H 2 O 2 -induced death (87,88,90). Accordingly, tumor-infiltrating lymphocytes must adapt to this oxidized microenvironment among other things by modulating the ratio ORAI1:ORAI3 expression (106,112). Despite these adaptation mechanisms, as we have seen earlier, exposure to exogenous ROS can severely dampen lymphocytes' effector function, making the TME particularly hostile to infiltrating lymphocytes (181). Interestingly and counter intuitively, the inflamed nature of the TME further contributes to making the TME hostile for lymphocytes and NK cells. Indeed, it was recently reported that tumor-derived PGE2 achieves immune evasion by inhibiting NK cell-mediated remodeling of the TME and unleashing of cytotoxic T cells (173). Interestingly, F2-isoprostanes (F2-IsoPs) and isolevuglandins (IsoLGs), which are oxidized derivatives of PGE2, are extremely relevant disease biomarkers, as they are directly involved in the pathological processes (induction of inflammatory pathways, modulation of immune response, and induction of cell death) (182)(183)(184). Since inflammation is closely linked to ROS production, whether the oxidized form of PGE2 contributes to this immune evasion needs to be investigated. Collectively, the evidence supports that the oxidized nature of the TME is likely to affect the efficiency of infiltrating anti-tumor lymphocytes and the development of strategies to enable lymphocytes to withstand the oxidized nature of the TME could improve immunotherapies (Figure 3).

CONCLUDING REMARKS
Based on the available evidence, both NOX-and mitochondrialderived ROS play critical roles in lymphocyte activation, development, effector function, cytotoxicity, viability but also dysfunction. To this regard ROS directly contribute to both physiological and pathological adaptive immune responses. ROS from both sources contribute to the activation process of lymphocytes, however, based on strong genetic evidence relying on hypomorphic allele of the Ncf1 gene encoding for p47 phox , it is tempting to suggest that NOX-derived ROS would have a preponderant role as modulators to fine-tune the inflammatory response depending on the timing, the site and the level of their production. But more investigations are still required to seal this case. Moreover, since ROS also participate in innate cell function by potentiating the killing ability of phagocytes, an essential step in the antigen processing and presentation function of these phagocytes, ROS also indirectly contribute to adaptive immunity though the interplay between innate and adaptive immunity. Further characterization of the complex functions of ROS in lymphocyte biology will bring new insight for understanding the pathological conditions in which lymphocyte function is either detrimental or beneficial.

AUTHOR CONTRIBUTIONS
DM, EB, and MW wrote and made the illustrations. All authors contributed to the article and approved the submitted version.

FUNDING
This work was supported by grants from UNIPD SID 2018 and ERC starting grant ERC-2010-StG_20091118.