The DNA Damage Response and Inflammation in Cancer

According to widely accepted models, oncogenesis as the result of malignant transformation (i.e., the conversion of a healthy cell into a malignant cell precursor) followed by tumor progression (i.e., the expansion of malignant cells to generate clinically relevant neoplasms) involves two major components (1–3). On the one hand, genetic and/or epigenetic defects provide healthy cells with the ability to evade cell-intrinsic oncosuppressive mechanisms that would otherwise prevent transformation, such as regulated cell death (RCD; ref. 1). On the other hand, newly formed malignant cell precursors acquire additional genetic and/or epigenetic alterations that allow for the evasion of cell-extrinsic oncosuppression, notably anticancer immunosurveillance (2, 3). In line with this notion, numerous oncoproteins (e.g., KRAS, MYCN) not only provide cancer cell precursors with aggressive features linked to malignancy (e.g., accrued proliferative potential) but also favor immunoevasion and/or the establishment of an immunosuppressive microenvironment (4). Conversely, various oncosuppressor proteins (e.g., TP53, PTEN) not only support the preservation of homeostasis in healthy cells and/or favor the demise of cells damaged beyond repair (which are prone to undergo malignant transformation) but also mediate immunostimulatory effects that ultimately engage anticancer immunosurveillance (4). Cell-intrinsic and cell-extrinsic oncosuppression mechanisms are so efficient that only a few (if any) clinically manifest neoplasms develop in the lifespan of an individual despite abundant somatic mutagenesis in virtually every cell of the body (5, 6).

In healthy cells, the DNA damage response (DDR; Fig. 1) mediates robust oncosuppressive functions that involve both such cell-intrinsic and cell-extrinsic components. First, efficient DNA repair de facto prevents the accumulation of genetic defects that may initiate malignant transformation. Second, if DNA damage cannot be satisfactorily repaired, the DDR actively engages RCD programs that ensure the demise of cells at high risk for malignant transformation. Finally, some components of the DDR machinery directly regulate the emission of immunostimulatory signals that recruit or activate immune cells, hence supporting the elimination of potentially oncogenic cells (7). For instance, oncogene activation in healthy hepatocytes has been shown to stimulate tumor protein p53 (TP53, best known as p53)-dependent cellular senescence, coupled to the secretion of chemokines and cytokines that ultimately drive an innate and adaptive immune response against senescent cells (8, 9). Supporting the critical oncosuppressive role of the DDR, individuals with xeroderma pigmentosum (XP), an autosomal recessive disease driven by germline mutations affecting nucleotide excision repair (NER), exhibit a severe predisposition to ultraviolet (UV) light-driven tumors (10). Moreover, multiple bona fide tumor suppressors are intimately involved in DNA repair and/or the activation of RCD or cellular senescence downstream of failed DNA repair. These proteins, which are lost or mutated in a high percentage of human tumors, include (but are not limited to): p53 (11), ATM serine/threonine kinase (ATM; ref. 12), BRCA1 DNA repair associated (BRCA1; ref. 13), BRCA2 (13), and phosphatase and tensin homolog (PTEN; ref. 14). In line with this notion, virtually all established tumors exhibit at least some DDR defects (Table 1; ref. 15), which de facto provides the scientific rationale for using DNA-damaging agents including conventional chemotherapeutics and radiotherapy (RT) for clinical cancer management (16).

That said, DDR signaling in transformed cells can also favor tumor progression and resistance to therapy. For instance, some tumors that arise as a consequence of genetic DDR defects develop non-oncogene addiction to alternative/complementary DDR pathways as a means to limit genetic instability to degrees that are compatible with cell survival (17–19). This is the case of malignancies bearing BRCA1 or BRCA2 mutations, which require poly(ADP-ribose) polymerase 1 (PARP1) for survival (and hence are exquisitely sensitive to PARP inhibitors in a setting of synthetic lethality; ref. 20). Moreover, although cancer cells exhibit DDR alterations that enable the use of DNA-damaging agents in the clinic (16), numerous DDR inhibitors including (but not limited to) agents specific for ATM, ATR serine/threonine kinase (ATR), checkpoint kinase 1 (CHEK1), and WEE1 G2 checkpoint kinase (WEE1) are being investigated as chemo- or radiosensitizers, at least in part based on their ability to compromise DNA repair and thus promote therapy-driven RCD or cellular senescence (Fig. 1; refs. 20, 21). Finally, the hyperproliferative phenotype of malignant cells as well as the relatively adverse conditions of the tumor microenvironment (TME) is per se associated with high degrees of DNA damage, requiring at least some DDR activity not only to limit potentially lethal genetic drifts but also to contain inflammatory and immune reactions that are elicited by the unwarranted accumulation of nucleic acids in ectopic subcellular compartments and may result in the activation of anticancer immunosurveillance (22–24). In line with this notion, genetic signatures of poor tumor infiltration by CD8⁺ T cells (a class of lymphocytes that are critical for anticancer immunity) have recently been linked with an increased expression of DNA repair genes in various cohorts of patients with pediatric and adult tumors (25). Finally, XP-related tumors (which are driven by UV light in the context of genetic NER defects, see above) are susceptible to nonspecific immunostimulants including the Toll-like receptor 7 (TLR7) agonist imiquimod (26), suggesting the presence of an immunologically active TME at baseline.

That said, although a proficient DDR is strictly beneficial for healthy cells as it ensures oncosuppression, progressing neoplasms may obtain superior benefits from incomplete DNA repair and subcytotoxic DDR activation. Indeed, suboptimal DNA repair and DDR signaling enable the rapid accumulation of mutations that (when compatible with cell survival and unable to drive RCD) foster tumor heterogeneity and clonal diversification, generally endowing the tumor ecosystem as a whole with accrued capacities for adaptation to environmental and therapeutic challenges that operate as selective pressures according to Darwinian principles (27). This is particularly true for defects in DNA mismatch repair (MMR) that culminate with the establishment of microsatellite instability (MSI), a condition of genetic hypermutability that provides progressing tumors with a superior potential for adaptation (27).

Importantly, inflammatory and immune reactions (inclu­ding those elicited or inhibited by the DDR) may have opposed effects on tumor progression and response to therapy (22). On the one hand, efficient anticancer immunosurveillance (be it natural or elicited by therapy) relies on the initiation of acute and robust but ultimately resolving inflammatory reactions that involve activated immune effector cells (2, 3). Such effector cells include innate natural killer (NK) cells (which may be involved in early anticancer immunosurveillance as well as in the control of metastatic cancer dissemination; ref. 28), as well as adaptive, tumor-specific CD4⁺ and CD8⁺ T cells (which are critical for the optimal efficacy of numerous treatment regimens including conventional chemotherapy, RT and targeted anticancer agents; refs. 29–31). On the other hand, chronic, indolent and nonresolving inflammatory reactions have been shown to promote tumor progression and resistance to treatment (32). Indeed, nonresolving inflammation (as in nonhealing wounds) is generally associated with the establishment of a microenvironment that is dominated by immunosuppressive cytokines such as transforming growth factor beta 1 (TGFβ1) and interleukin 10 (IL10; ref. 33), as well as by regulatory cell populations, including so-called “M2-like” tumor-associated macrophages (TAM; ref. 34) and CD4⁺CD25⁺FOXP3⁺ regulatory T cells (Treg; ref. 35), coupled to the functional exhaustion of effector immune cells (36).

Here, we critically discuss the intricate links between the DDR and inflammation in the context of malignant transformation, tumor progression, and response to treatment, with emphasis on strategies harnessing DDR modulators to alter the immunological tumor configuration for therapeutic purposes. Mechanistic aspects of DNA repair and the DDR unrelated to inflammation have been recently reviewed elsewhere (23, 37) and hence will not be covered by this review.

Base excision repair (BER) is a DDR pathway activated by DNA lesions that do not considerably distort the tertiary structure of the double helix, notably bases that are oxidized by reactive oxygen species (ROS) or reactive nitrogen species (RNS), two major sources of DNA damage (38, 39). BER initiates with the removal of damaged/oxidized or misincorporated bases by DNA glycosidases, such as 8-oxoguanine DNA glycosylase (OGG1) and mutY DNA glycosylase (MUTYH), resulting in the generation of abasic sites that are detected and excised by apurinic/apyrimidinic endodeoxyribonuclease 1 (APEX1, best known as APE1). This culminates with the formation of single-strand breaks (SSB) that are sensed by PARP1 (see below), leading to the recruitment of additional BER and SSB repair factors to damaged sites (40). Finally, DNA synthesis by DNA polymerase beta (POLB) or delta (POLD), coupled to DNA ligation by DNA ligase 1 (LIG1) or LIG3 in complex with X-ray repair cross complementing 1 (XRCC1), can restore the original DNA sequence (Fig. 1). Although defects in BER have only sporadically been associated with human oncogenesis (40), multiple BER components have been shown to regulate inflammatory reactions in healthy or malignant cells.

For instance, OGG1 appears to mediate inflammatory functions in a variety of disease models, prompting the development of pharmacologic OGG1 inhibitors for use as nonsteroidal anti-inflammatory drugs (41). In line with this notion, the lack of Ogg1 limits inflammatory lesions induced in mice orally inoculated with the carcinogenic bacteria Helicobacter pylori, correlating with reduced serological signs of inflammation and blunted mutagenesis (42). However, Ogg1^−/− mice are more susceptible to colorectal carcinogenesis as driven by azoxymethane (AOM) plus dextran sodium sulfate (DSS), which—similar to H. pylori-driven gastric carcinogenesis—involves a considerable inflammatory component, than their wild-type (WT) counterparts (43, 44). Similar findings have been obtained in mice lacking N-methylpurine DNA glycosylase (MPG), another glycosidase involved in BER (45). Conversely, Mutyh^−/− mice appear to be protected from DSS/AOM-driven carcinogenesis (46). That said, the whole-body codeletion of Ogg1 and Mutyh considerably shortens the mouse lifespan, correlating with a considerable increase in the propensity for spontaneous oncogenesis, two effects that are further aggravated by the additional deletion of Msh2 (coding for an MMR component; ref. 47). Taken together, these findings exemplify the critical role of BER as a suppressor of oncogenesis driven by ROS and RNS. Accordingly, both OGG1 and MUTYH are mutated in a wide panel of human malignancies (48, 49).

APE1 is active in gastric epithelial cells infected by H. pylori, resulting in the activation of proinflammatory transcription factors such as AP-1 and NF-κB, interleukin 8 (IL8) production, and the establishment of an inflammatory response that supports not only gastric ulcer but also gastric transformation (50). Such a proinflammatory activity of APE1 also emerges in the context of tumor necrosis factor (TNF) signaling, which is key for organismal responses to chronic H. pylori infection and gastric carcinogenesis (51). APE1 also appears to be required for ROS generation in gastric epithelial cells exposed to H. pylori (52), knowing that (i) ROS and RNS are prominent sources of DNA damage and mutagens (39), (ii) ROS are potent activator of the NLR family pyrin domain containing 3 (NLRP3) inflammasome (53), and (iii) NLRP3 signaling coupled to IL1β and IL18 secretion has been linked to H. pylori pathogenicity (54). Thus, APE1-driven inflammatory reactions appear to contribute to gastric carcinogenesis in the context of chronic H. pylori infection.

Interestingly, APE1 drives tumor-promoting inflammation (TPI) largely because of its role in redox responses, rather than through its functions in BER. Moreover, it seems that the ability of APE1 to drive TPI may also involve its ability to regulate redox homeostasis in macrophages (55) and helper T cells (56). In line with this notion, POLB (which operates downstream of APE1 in BER) reportedly mediates robust anti-inflammatory, rather than proinflammatory, functions. Specifically, expression of a POLB variant that is associated with human gastric carcinoma (L22P) in mouse gastric epithelial cells not only favors chromosomal instability but also establishes TPI and promotes gastric carcinogenesis both as a standalone intervention (57) and in the context of H. pylori infection (58). Finally, depletion of POLB from human mammary carcinoma MCF7 cells drives potent type I interferon (IFN) synthesis downstream of faulty DNA repair, accrued cytosolic accumulation of double-strand DNA (dsDNA), and consequent activation of cyclic GMP-AMP synthase (CGAS), resulting in superior cancer cell sensitivity to elimination by NK cells (59). These findings suggest that proficient BER is key for preserving genetic integrity and inhibiting unwarranted inflammatory reactions. In line with this notion, low levels of XRCC1 correlated with increased tumor infiltration by CD8⁺ T cells, albeit in the context of high tumor grade and other features of aggressiveness, in two large cohorts of patients with breast cancer (60).

These observations exemplify the critical impact of BER components (via BER-dependent or independent mechanisms) on inflammatory responses in healthy and malignant tissues (Supplementary Table S1).

MMR resolves base mismatches and short insertions or deletions (61). Such alterations are detected by heterodimers consisting of mutS homolog 2 (MSH2) and MSH6 (known as MutSα), or MSH2 and MSH3 (known as MutSβ). In turn, MutSα (or MutSβ) provides a platform for mutL homolog 1 (MLH1) and PMS1 homolog 2, mismatch-repair system component (PMS2) heterodimers, an endonuclease complex stimulated by proliferating cell nuclear antigen (PCNA). This results in the incision of the nascent DNA strand, mismatch removal by exonuclease 1 (EXO1), and DNA repair synthesis (Fig. 1; ref. 61).

Heterozygous germline mutations in MMR components cause the so-called Lynch syndrome (LS), an inheritable predisposition for colorectal cancer (62). Moreover, somatic MMR defects provoke genetic hypermutability in the context of MSI, which has recently been approved as a tumor-agnostic marker for the use of immune-checkpoint inhibitors (ICI) targeting programmed cell death 1 (PDCD1, best known as PD-1) in patients with cancer (63). In line with this notion, a defective MMR correlates not only with an increased tumor mutational burden (TMB, which has been linked to ICI sensitivity in multiple cancer patient cohorts; ref. 64) but also with exacerbated inflammatory reactions and an inflamed TME at baseline (which has likewise been consistently associated with clinical ICI effectiveness; refs. 2, 65).

As compared with their WT counterparts, Msh2^−/− mice are more sensitive to DSS/AOM-driven colorectal carcinogenesis (66), an effect that can be attenuated, at least in the distal colon, by the nonsteroidal anti-inflammatory drug sulindac (67). Along similar lines, the loss of Msh2 exacerbates the formation of colorectal lesions in tumor-prone APC^Min/+ mice by a mechanism that does not originate from compromised repair of nitric oxide-induced DNA lesions (68) but rather involves the gut microbiome, both as a source of oncogenic metabolites (69) and as a driver of inflammation (70). Intriguingly, MSH2 is downregulated in response to TNF signaling in the liver (71), suggesting that, at least in some settings, TPI and DDR defects may be interconnected via a vicious feedforward loop (Fig. 2). Finally, although the combined loss of MSH2 and its binding partner MSH6 has been associated with a high TMB in a heterogeneous collection of MSI-high cancers (72), in a data set including 5 distinct cohorts of patients with lung adenocarcinoma, high MSH2 expression levels correlated positively not only with signatures of MMR proficiency but also with a high TMB, expression of CD274 (best known as PD-L1), and tumor infiltration by CD8⁺ T cells, pointing to an ICI-sensitive TME (73). Although the precise reasons underlying such an apparent discrepancy remain to be elucidated, tissue of origin may considerably influence the impact of MSH2 defects on the TME, as previously documented in prostate versus colorectal or endometrial tumors (74). On the other hand, several case reports confirm that LS-associated germline MSH6 mutations are linked to extraordinary sensitivity to ICIs not only in patients with colorectal cancer (75) but also in patients with uterine serous carcinoma (76) or endometrial tumors (77).

Loss of MLH1 and its binding partner PMS2 has been linked to increased TMB in MSI-high solid tumors, although to lower levels than the combined loss of the MutSα components MSH2 and MSH6 (72). In endometrial carcinoma, loss of MLH1 and PMS2 has also been independently associated with PD-L1 positivity (78), again associating MSI with therapeutically actionable features of the TME. Accordingly, gastrointestinal tumors evolving in Mlh1^−/− mice could be efficiently controlled by combining immunogenic chemotherapeutics such as gemcitabine or cyclophosphamide (29), a tumor-targeting vaccine (79, 80), and/or a PD-L1 blocker (81, 82).

Importantly, such a response pattern does not only reflect the abundant tumor neoantigen (TNA) load caused by the loss of MLH1 (which per se may induce some degree of anticancer immunity; ref. 83), but also the active secretion of proinflammatory factors, including type I IFN, C–C motif chemokine ligand 5 (CCL5) and C–X–C motif chemokine ligand 10 (CXCL10) via a CGAS- and stimulator of interferon response cGAMP interactor 1 (STING1)-dependent mechanism, by cells experiencing MMR defects (84–86). At least in some settings, such a response is initiated by EXO1-driven DNA hyperexcision and consequent accumulation of CGAS-activating DNA in the cytosol (87). Accordingly, Mlh1^−/− mice are exquisitely sensitive to inflammation-driven colorectal carcinogenesis (88), a phenotype that can be further aggravated by mild inflammatory cues including low-dose radiation (89). Of note, MLH1 also resembles MSH2 in being (epigenetically) repressed by inflammatory cues (90) in the context of a vicious oncogenic cycle involving defective DDR and TPI (Fig. 2).

Taken together, these observations exemplify the major genoprotective and anti-inflammatory roles of a proficient MMR in both healthy and malignant cells (Supplementary Table S2).

Nonhomologous end-joining (NHEJ) is a potentially error-prone mechanism for the repair of DNA double-strand breaks (DSB) via direct ligation of DNA ends (91, 92). NHEJ is initiated by the binding of a heterodimer consisting of XRCC6 (best known as Ku70) and XRCC5 (best known as Ku80) to DSBs. The Ku70/Ku80 heterodimer promotes the recruitment and activation of protein kinase, DNA-activated, catalytic polypeptide (PRKDC, best known as DNA-PKcs). Processing of incompatible DSB ends by nucleases, like DNA cross-link repair 1C (DCLRE1C, best known as Artemis), and DNA polymerases, such as DNA polymerase mu (POLM) and lambda (POLL), facilitates DNA end-ligation by LIG4, XRCC4, and nonhomologous end-joining factor 1 (NHEJ1, best known XLF; refs. 91, 92; Fig. 1). Defects in NHEJ are fairly uncommon in human neoplasms, possibly with the exception of single-nucleotide polymorphisms in XRCC5, XRCC6, or their promoter regions, which have been associated with an increased risk for multiple solid tumors (93).

At odds with other DDR components (see above), both XRCC6 and XRCC5 appear to be upregulated in the context of proinflammatory signaling via TLR4 or prostaglandin-endoperoxide synthase 2 (PTGS2, best known as COX2) in healthy tissues including the liver (94), as well as in malignant cells (95). Diethylnitrosamine (DEN)-driven hepatic carcinogenesis was accelerated in male mice bearing a missense Tlr4 mutation, correlating with increased oxidative DNA damage, failed Ku70 and Ku80 upregulation early upon DEN-mediated hepatic injury, limited inflammatory signaling, and reduced numbers of intrahepatic macrophages (94). Importantly, most of these hepatic phenotypes could be reverted by adenovirus-driven Ku80 overexpression in hepatocytes, culminating with decelerated DEN-driven carcinogenesis (94). Of note, Ku80 levels negatively correlate with PD-L1 expression in post-RT biopsies from patients with a variety of tumors (96), suggesting that a proficient NHEJ or other NHEJ-unrelated Ku70 or Ku80 functions limit tumor-targeting immune responses culminating with interferon gamma (IFNG)-dependent PD-L1 upregulation in the TME.

Supporting the latter possibility, deletion of Xrcc6 (alone or in the context of mutant p53) causes intestinal alterations associated with a pronounced inflammatory response and increased propensity for carcinogenesis, a phenotype that the deletion of Lig4 cannot recapitulate (97). Of note, extranuclear Ku70 has been proposed to drive type III IFN secretion in response to the cytosolic accumulation of DNA, via a mechanism that involves the Ku70 DNA-binding domain and STING1 (98, 99). However, it seems unlikely that the loss of cytosolic DNA sensing by Ku70 may explain inflammation-associated colorectal carcinogenesis in Xrcc6^−/− mice.

DNA-PKcs has also been proposed to operate as cytosolic DNA sensor in the context of viral infection, ultimately driving type I IFN production via a STING1- and interferon regulatory factor 3 (IRF3)-dependent mechanism (100). Such an extranuclear DNA-PKcs activity, which is antagonized by a variety of viral factors (101, 102), does not appear to involve physical interactions with Ku70 (103) and may not be operational in all cells. Indeed, pharmacologic inhibition (rather than activation) of DNA-PKcs with AZD7648 synergizes with RT to promote therapeutically relevant type I IFN responses in various syngeneic mouse tumor models (104). More­over, not only missense mutations of PRKDC are associated with autoimmune disorders (105), but DNA-PKcs has also been shown to catalyze the inactivating phosphorylation of CGAS, de facto suppressing antiviral immunity (106). Finally, a nuclear pool of STING1 has been proposed to support genomic stability by binding to DNA-PKcs and its interactors (107). Taken together, these latter observations suggest that the ability of DNA-PKcs to operate as a cytosolic DNA sensor should be revisited. Nonetheless, it appears that both DNA-PKcs and less so the DDR signal transducers ATR and mitogen-activated protein kinase 14 (MAPK14, best known as p38^MAPK) are important for cancer cells experiencing DNA damage to emit adjuvant signals that actively engage tumor-specific immune responses irrespective of–or prior to–immunogenic cell death (ICD; refs. 108, 109). The precise molecular mechanisms underlying these latter observations, however, remain elusive.

Yet another NHEJ component, XRCC4, has been shown to affect inflammatory reactions (110). Specifically, XRCC4 physically interacts with RNA sensor RIG-I (RIGI, best known as RIG-I), a cytosolic pattern recognition receptor (PRR) that binds ectopic RNA molecules (111). Besides inhibiting NHEJ (and hence interfering with viral integration in the host genome), such an interaction promotes the oligomerization and ubiquitination of RIG-I in support of improved antiviral activity (110). However, whether the interaction between XRCC4 and RIG-I also modulates inflammatory responses in tumor models or patients with cancer has not yet been investigated.

Thus, NHEJ components operate as part of the DDR to repair DSBs in the nucleus and as regulators of inflammatory reactions in the cytosol (Supplementary Table S3).

Homologous recombination (HR) is a generally error-free pathway that operates in the S and G₂ phases of the cell cycle to promote the repair of DSBs using homologous DNA templates (91, 92). HR is initiated by the 5′ to 3′ resection of DNA ends by the MRN complex, a heterotrimer consisting of MRE11 homolog, double-strand break repair nuclease (MRE11), RAD50 double-strand break repair protein (RAD50) and nibrin (NBN, best known as NBS1); and RB binding protein 8, endonuclease (RBBP8, best known as CTIP). This generates short stretches of single-strand DNA (ssDNA) that are elongated through the activities of nucleases, like EXO1 and DNA2, and helicases, such as BLM RecQ like helicase (BLM), resulting in long ssDNA overhangs that are coated and protected by replication protein A1 (RPA1, best known as RPA). BRCA1, BRCA2, and partner and localizer of BRCA2 (PALB2) favor the exchange of RPA with RAD51 recombinase (RAD51), which forms a nucleofilament that can invade the undamaged sister chromatid in search of homologous regions to utilize as template for repair. The invading strand initiates DNA synthesis, which can be followed by synthesis-dependent strand annealing, involving the release of the newly synthesized strand and its consequent reannealing to the second DSB end. Alternatively, double Holliday junctions can form, which can either undergo resolution by Holliday junction resolvases or dissolution by a complex containing BLM and DNA topoisomerase III (TOPOIII), resulting in the generation of crossover or noncrossover products (Fig. 1; refs. 91, 92). HR has been extensively investigated in the context of cancer, especially upon the identification of germline BRCA1 and BRCA2 mutations as risk factors for mammary and ovarian carcinogenesis (112), and the discovery that neoplasms bearing BRCA1 or BRCA2 mutations are exquisitely sensitive to PARP inhibition (20).

Each of the components of the MRN complex has been involved in the regulation of inflammatory reactions. For instance, an extranuclear pool of MRE11 has been shown to serve as a cytosolic DNA sensor via a mechanism that involves its partner RAD50 and culminates with STING1 activation followed by IRF3-dependent type I IFN synthesis (113). Moreover, in patients with Fanconi anemia (FA, see below), mitochondrial MRE11 actively degrades nascent mitochondrial DNA (mtDNA), resulting in mtDNA cytosolic spillage and CGAS activation (114), a process that is generally under negative regulation by autophagy (which degrades dysfunctional mitochondria; ref. 115) and caspases (which proteolytically inactivate CGAS and accelerate cell death; refs. 116, 117). Finally, the exonuclease activity of MRE11 can generate CGAS-activating DNA fragments and hence elicits innate immune signaling, at least in cells lacking RAD51, which would otherwise protect the newly synthesized DNA from MRE11 degradation (118). Extranuclear RAD50 also appears to interact with the PRR caspase recruitment domain family member 9 (CARD9), thus enabling IL1β production in response to cytosolic DNA accumulation in infected dendritic cells (DC) via an NF-κB–dependent mechanism (119). Whether this latter process is operational in noninfected or epithelial cells remains to be formally demonstrated.

Dissociation of the MRN complex as caused by adenoviral infections has also been suggested to promote NBS1 segregation in nuclear foci, coupled with ATM hyperactivation and consequent NF-κB signaling (120). Although this may result in genomic instability coupled with polyploidization (120), its impact on immunogenicity has not been clarified yet. Interestingly (at least in some circumstances), intracellular PD-L1 can act as an mRNA-binding protein to stabilize selected transcripts including NBS1, BRCA1, and other mRNAs encoding DDR components (121). This results in superior DDR proficiency, at least hypothetically coupled to the suppression of potentially oncosuppressive inflammatory reactions driven by suboptimal HR (121). Both NBS1 (122) and BRCA1 defects (see below) have indeed been associated with accrued inflammatory responses at baseline or in response to carcinogens.

EXO1 promotes HR by catalyzing long-range end-resection (123). EXO1 is also involved in other DDR pathways, including NER and MMR, largely reflecting its ability to remove DNA stretches containing lesions prior to the restoration of original DNA sequences (61, 124). EXO1 nuclease activity can generate ssDNA fragments that—unless degraded by the cytosolic exonuclease three prime repair exonuclease 1 (TREX1)—mediate interferogenic effects (125). A similar activity has also been documented for the end-resection factor BLM (125). Moreover, EXO1-dependent DNA degradation appears to underlie the ability of DSBs accumulating in response to RT to elicit PD-L1 exposure in cancer cells, a mechanism that is finely regulated by multiple HR components (126).

BRCA1 and BRCA2 promote HR by favoring DSB end-resection (BRCA1) and loading the recombinase RAD51 on resected DNA ends (BRCA1 and BRCA2; ref. 13). Besides generating non-oncogene addiction to PARP, loss of BRCA1 has been shown to promote oncogenesis and tumor progression in the context of immunosuppressive TME alterations that (at least in some settings) may nonetheless be actionable for therapeutic purposes. For instance, BRCA1 mutations have been associated with the recruitment of myeloid-derived suppressor cells (MDSC) to the mammary TME via CXCL12, resulting in therapeutic resistance to PD-1 blockers that could be abrogated with CXCL12-targeting strategies (127). Similarly, in preclinical models of ovarian cancer, loss of Brca1 results in an inflamed (but immunoresistant) TME at baseline as a consequence of chronic CGAS/STING1 activation and consequent secretion of vascular endothelial growth factor A (VEGFA; ref. 128). In this setting, sensitivity to dual PD-1 and cytotoxic T-lymphocyte associated protein 4 (CTLA4) blockage can be restored with VEGFA-neutralizing agents (or Sting1 deletion) optionally combined with PARP inhibition (128). Signatures of CGAS/STING1 and IFNG signaling have also been detected by comparing isogenic BRCA1-deficient versus BRCA1-proficient breast and ovarian cell lines, as well as by comparing patients with mutant vs. WT BRCA1 from The Cancer Genome Atlas (TCGA; ref. 129).

In triple-negative breast cancer (TNBC) models, Brca1 defects correlate with increased TMB and superior tumor infiltration by lymphocytes expressing PD-1 and CTLA4, offering a therapeutically actionable profile for dual immune-checkpoint blockage plus cisplatin-based immunotherapy (130). Disrupting the binding between BRCA1 and its physical interactor PALB2 facilitates age-driven hepatic oncogenesis, but results in the formation of tumors that respond to PD-1 blockage as a consequence of robust type I IFN secretion via CGAS (131). These findings suggest that the net effect of BRCA1 mutations on the TME and sensitivity to ICIs exhibits at least some degree of context dependency. In line with this notion, BRCA1 mutations have been correlated with increased tumor infiltration by a variety of immune cells including CD4⁺ and CD8⁺ T cells, in patients with hepatocellular carcinoma (HCC; ref. 132), breast cancer (133), or high-grade serous ovarian carcinoma (HGSOC; ref. 134). However, although in HCC this was linked to MSI and poor prognosis (132), a robust immune infiltrate and/or enhanced spatial interactions between malignant cells and T cells had positive prognostic value in patients with BRCA1-mutated breast cancer (133) and HGSOC (134).

Recent data argue against the assumption that BRCA1 and BRCA2 mutations would have similar immunological effects on developing tumors. Specifically, gene-expression programs related to both adaptive and innate immunity are enriched in BRCA2- over BRCA1-deficient tumors (135). Moreover, truncating mutations in BRCA2 are associated with superior response to ICIs as compared with their BRCA1 counterparts, not only as modeled in various mouse cancer cell lines but also as assessed in a patient cohort encompassing 95 patients with diverse malignancies (135). Along these lines, although the acute transcriptional response to experimental BRCA2 loss is largely dominated by genes involved in cell-cycle progression, DNA replication, and repair, late transcriptional alterations are mostly linked to IFN signaling, a process that is further magnified by PARP inhibition (136). Specifically, BRCA2 inactivation appears to promote the formation of micronuclei that recruit CGAS to initiate a potent interferogenic response that is accompanied by TNF secretion, and restores cell killing by exogenous TNF administration in a variety of otherwise TNF-resistant human cancer cell lines (137). In line with this notion, human cancer cells with homozygous or heterozygous deletions of BRCA2 have been attributed superior sensitivity to RCD driven by TNF-like death receptor ligands (138). These observations reveal unsuspected differences between the immune correlates of BRCA1 and BRCA2 defects that require additional investigation.

Similar to BRCA1 defects, RAD51 expression has been consistently associated with increased TMB and robust tumor infiltration by immune cells including CD4⁺ and CD8⁺ T cells in patients with HCC (139) and breast cancer (140, 141). However, in all these settings, elevated RAD51 levels correlated with advanced tumor stage, high grade, and dismal prognosis (139–141). These observations may reflect the establishment of chronic, indolent TPI with a detrimental impact on oncogenesis, and tumor progression. Supporting this notion, transcriptional signatures of type I and IFNγ signaling have consistently been associated with poor prognosis in patients with HCC and breast cancer (116, 142). Apparently at odds with these observations, however, depletion of RAD51 or its interacting partner RPA has been shown to cause robust accumulation of cytosolic ssDNA coupled to potent CGAS activation, especially in the context of defects in TREX1 (143). Conversely, CGAS appears to potently inhibit HR by compacting template DNA into a hypercondensed state that is less amenable to invasion by RAD51-coated ssDNA filaments (144), and by binding to and hence interfering with the activation of PARP1 (145). Such an effect, however, is independent from canonical CGAS/STING1 signaling.

In conclusion, HR signaling heavily impinges on the regulation of inflammatory reactions in cancer (Supplementary Table S4).

PARP1 and other members of the PARP protein family detect a variety of DNA lesions including SSBs resulting from the APE1-mediated processing of abasic sites during BER, ssDNA gaps formed during DNA replication, as well as DSBs in the context of HR and NHEJ, ultimately participating in the initiation of multiple DDR pathways (146). PARP inhibitors have been intensively investigated (and are currently approved for use) in patients with breast or ovarian neoplasms bearing BRCA1 or BRCA2 mutations (20). An expanding literature suggests that at least part of the therapeutic benefits mediated by PARP inhibitors in these patients emerges from immunologic mechanisms.

PARP inhibition reportedly mediates strong, STING1-dependent inflammatory responses culminating with type I IFN and/or CCL5 secretion in models of BRCA1-deficient and proficient ovarian cancer (147, 148), BRCA1-deficient (and less so BRCA1-proficient) TNBC (149, 150), BRCA1-competent non–small lung cell carcinoma (NSCLC) with defects in NER (149), BRCA2-deficient NSCLC and TNBC cells (136), as well as in renal cell carcinoma models lacking polybromo 1 (PBRM1), a chromatin remodeling enzyme (151), and models of small-cell lung carcinoma (SCLC; ref. 152).

Intriguingly, such an immunostimulatory activity has been linked to the ability of some PARP inhibitors to induce PARP1 trapping on DNA (153), resulting in accrued replication stress (see below), aggravated genomic instability, and reduced availability of cytosolic PARP1, which directly inhibits CGAS via PARylation (154). Conversely, CGAS accumulating in the nucleus as a consequence of DSBs interferes with PARP1 activation during HR (145). Such an effect, however, does not seem to involve canonical CGAS signaling but instead the dephosphorylation of CGAS on Y215, which is a target of BLK proto-oncogene, Src family tyrosine kinase (BLK; ref. 145).

Other members of the PARP family, including PARP7 (and most likely PARP2), share with PARP1 the ability to strongly suppress type I IFN signaling (155). In some models, genetic or pharmacologic PARP inhibition promoted the establishment of an inflamed TME in vivo (152, 156–158), correlating with the IFNG-driven upregulation of PD-L1 (149, 159) and synergism with ICIs (148, 152, 157–160) as well as immunotherapeutic agents stimulating STING1 signaling (161), targeting immunosuppressive macrophages (162), or promoting antibody-dependent cellular cytotoxicity by NK cells (163). Intriguingly, the ability of PARP inhibitors to drive PD-L1 expression in the TME may originate not only from the activation of cellular immunity coupled with IFNγ production but also from the ability of PARP1 to suppress the transcription of CD274 by PARylating signal transducer and activator of transcription 3 (STAT3; ref. 164). Along, similar lines, PARP1 has been shown to actively suppress the expression of various NK cell–activating ligands (NKAL) by acute myeloid leukemia cells, hence supporting their ability to evade NK cell–dependent immunosurveillance (which is particularly prominent in leukemia; refs. 165, 166).

Of note, Parp1^−/− mice appear to be less (rather than more) sensitive to DSS/AOM-induced colorectal carcinogenesis, correlating with moderately increased DNA damage but decreased signs of inflammation in the intestinal epithelium (167). Intriguingly, such a chemopreventive effect appears to be maximal in heterozygous Parp1^+/− hosts or in the context of incomplete pharmacologic inhibition of PARP, potentially reflecting the inhibition of TPI, notably the tumor-promoting activity of MDSCs, coupled to the reestablishment of robust anticancer immunosurveillance (156, 168).

Altogether, these data suggest that interfering with the kinetic whereby PARP normally interacts with damaged DNA, notably PARP trapping, results in robust inflammatory responses often linked to cytosolic STING1 activation (Supplementary Table S5).

The main DDR kinases are ATM, which is activated upon DSB formation (12), and ATR, which is activated in response to ssDNA stretches generated following DSB end-processing or replication fork stalling (169). The catalytic activity of ATM and ATR regulates (either directly or indirectly) various effector molecules that promote DNA repair in the context of a transient cell-cycle arrest, or actively drive cellular senescence or RCD when DNA is damaged beyond recovery. These effectors include CHEK2 (the prototypical ATM substrate) and CHEK1 (the prototypical ATR substrate) as well as p53, p38^MAPK, and WEE1 (Fig. 1; refs. 12, 169). Somatic mutations or epigenetic mechanisms suppress (at least to some degree) DDR signaling in most human tumors (12, 170). This reflects the robust oncosuppressive function of proficient DDR signaling, both as a genoprotective mechanism in the context of mild DNA damage and as a driver of cellular senescence or RCD in cells bearing irreparable DNA lesions (171). That said, most cancer cells tend to become highly reliant on alternative DDR pathways for survival and proliferation, which has raised considerable interest in the development of DDR inhibitors (171). Accumulating evidence suggests that these agents may also evoke therapeutically relevant anticancer immune responses.

Atm^−/− mice resemble their Msh2^−/− counterparts as they exhibit increased sensitivity to colitis-driven colorectal carcinogenesis compared with WT mice (172). Along similar lines, the loss of ATM (which is frequent in human tumors) may generate non-oncogene addiction to PARP1 (at least in models of colorectal cancer and mantle cell lymphoma), which also occurs in the context of BRCA1 defects (173, 174). Moreover, ATM inhibition has been shown to derepress tumor-targeting immune responses by a variety of mechanisms. First, pharmacologic inhibition of ATM drives robust type I IFN secretion in preclinical tumor models, via both CGAS/STING1-independent (175) and -dependent processes (176, 177), representing an efficient therapeutic partner for ICIs (175, 176). Of note, at least in some of these settings, similar results could be obtained with inhibitors of CHEK2 (177). Second, ATM depletion by RNA interferences suppresses the expression of integrin αvβ3 by cancer cells experiencing DNA damage (178), knowing that integrin αvβ3 facilitates the uptake of viable (as compared with dying) cancer cells by antigen-presenting cells (APCs) and hence suppresses CD8⁺ T-cell cross-priming (178). Finally, pharmacologic ATM inhibition prevents PD-L1 expression as driven by chemotherapy in preclinical models of prostate cancer (179). In line with this notion, the loss of ATM as well as ATM mutations have been linked with a beneficial immune configuration of the TME and/or with increased sensitivity to ICIs in various cohorts of patients with cancer (177, 180, 181).

Although ATM signaling in neoplastic cells dampens inflammation, ATM activation in immune cells apparently supports (rather than inhibits) inflammatory responses. For instance, an extranuclear pool of ATM has been shown to underlie the ability of macrophages to acquire proinflammatory, antitumoral activities in response to the activation of NADPH oxidase 2 by radiotherapy (182). Similarly, ATM signaling in APCs exposed to low-dose chemotherapy drives an NF-κB–dependent signal transduction pathway that ultimately enhances antigen presentation by upregulating MHC class II surface levels (183). Whether the latter is associated with improved immune tumor control has not been explored. Of note, inflammatory signaling as driven by CGAS also regulates ATM functions. Specifically, the CGAS-dependent synthesis of cyclic GMP-AMP (cGAMP) in response to cytosolic dsDNA has been shown to promote ATM activation in STING1- and TANK binding kinase 1 (TBK1)-dependent (but type I IFN-independent) manner, resulting in cell-cycle arrest downstream of CHEK2 signaling but inhibited HR as a consequence of NAD⁺ depletion (184). These latter observations exemplify yet another scenario in which inflammatory cues and DDR inhibition may interact in a feedforward, self-amplifying mechanism (Fig. 2).

ATR activation as driven by Epstein–Barr virus (EBV) infection (which is responsible for most nasopharyngeal carcinoma cases) promotes the establishment of a strongly immunosuppressive microenvironment dominated by M2-like macrophages that is permissive for oncogenesis and tumor progression (185). Moreover, inhibited ATR signaling due to baculoviral IAP repeat containing 6 (Birc6) deletion in hepatocytes drives tumor-promoting inflammatory reactions that aggravate DEN-driven hepatic carcinogenesis (186). ATR underlies multiple immunosuppressive mechanisms including the upregulation of PD-L1 and CD47 (an antiphagocytic signal) also in cancer cells (187), de facto impairing the ability of RT to elicit therapeutically relevant anticancer immune responses as a consequence of defects in both immune priming (187) and immune execution (188). Thus, proficient ATR signaling mediates anti-inflammatory effects in both normal and malignant tissues. In stark contrast, however, Atr haploinsufficiency accelerates the formation of melanomas driven by the loss of Pten and activation of B-Raf proto-oncogene, serine/threonine kinase (BRAF) in mice, a phenotype that is accompanied by the establishment of an immunosuppressive TME with M2-like macrophages and PD-L1 expression (189). Whether such an apparently paradoxical effect originates from gene dosage changes remains to be clarified.

Irrespective of this unknown, pharmacologic ATR inhibitors reportedly mediate robust immunostimulatory effects, often dependent on cytosolic dsDNA accumulation and CGAS signaling, in preclinical models of prostate cancer (190), acute lymphocytic leukemia (ALL; ref. 191), head and neck squamous cell carcinoma (HNSCC; ref. 192), PBRM1-deficient renal cell carcinoma (151), and BRCA2-defective cervical carcinoma (193). Moreover, ATR inhibition synergizes with RT at driving CGAS-dependent immunostimulatory effects of therapeutic relevance that (at least in some settings) can be exacerbated by ICIs in models of hepatocellular carcinoma (194), human papillomavirus (HPV)⁺ lung carcinoma (195), and KRAS-mutant tumors (196). Intriguingly, a similar interferogenic response has been documented in nontransformed mammary epithelial MCF10A cells exposed to RT and ATR inhibition, although, in this context, it was mechanistically linked to sensing of cytosolic RNA, not DNA (197). ATR inhibitors have also been shown to synergize at the activation of therapeutically relevant (and sometimes ICI-actionable) tumor-targeting immune responses upon combination with WEE1 inhibitors (in models of ovarian cancer; ref. 198), oxaliplatin (in models of oxaliplatin-resistant colorectal cancer; ref. 199), and aurora kinase A (AURKA) blockers (in models of MYCN-amplified neuroblastoma; ref. 200). These latter observations reinforce the notion that inhibiting multiple components of the DDR may drive superior immunostimulatory effects, perhaps by abrogating (or at least interfering with the emergence of) compensatory non-oncogene addiction.

Immunostimulatory effects have also been observed upon inhibition of CHEK1. For instance, pharmacologic CHEK1 inhibition has been shown to synergize with the immunogenic chemotherapeutic gemcitabine and ICIs in preclinical models of SCLC, correlating with accrued tumor infiltration by CD8⁺ T cells, DCs, and proinflammatory M1-like macrophages at the expenses of anti-inflammatory M2-like macrophages and MDSCs (201). Similar to ATR blockers, CHEK1 inhibitors can also promote TNF secretion by ALL cells (191), although the immunologic consequences of this process remain unclear. Finally, pharmacologic CHEK1 inhibition has been shown to mediate potent, CGAS-dependent immunostimulatory effects and hence synergize with PD-L1 blockers in mouse models of SCLC (152).

That said, pharmacologic CHEK1 inhibition reportedly limits (rather than aggravates) colorectal carcinogenesis in the DSS/AOM model, apparently as a consequence of inhibited C–C motif chemokine ligand 2 (CCL2) expression and hence scarce recruitment of CCR2⁺ macrophages to the colon (202). Finally, it seems that multiple inflammatory cues converging on IRF1 activation (including, but not limited to, IFNG signaling) promote CHEK1 downregulation (203, 204). Such a circuitry, however, appears more plausible to facilitate the apoptotic demise of cancer cells experiencing DNA damage or replication stress than to inhibit DNA repair.

Activation of p53 in hepatocytes undergoing oncogene-driven senescence mediates robust cell-extrinsic oncosuppressive functions via CD4⁺ T cells and NK cells (8, 9). Intriguingly, such a homeostatic effect appears to require physiologic p53 signaling, as constitutive p53 activation resulting from the deletion of transformed mouse 3T3 cell double minute 2 (Mdm2, which encodes the main endogenous p53 inhibitor) instead paradoxically promotes KRAS-driven hepatic carcinogenesis upon the establishment of TPI (205). That said, p53 loss and cancer-associated TP53 mutations have been consistently linked to the establishment of TPI in support of tumor progression in the context of immunoevasion. For instance, a common p53 mutant (p53^R249S) binds TBK1 and prevents it from interacting with STING1 and IRF3, de facto blocking type I IFN secretion downstream of CGAS signaling and hence limiting the infiltration of mouse TNBCs by NK cells in favor of increased amounts of immunosuppressive. M2-like macrophages, and accelerating tumor growth (206). Along similar lines, a gain-of-function mutation of p53 that is common in human neoplasms (R172H) alters the immune infiltrate of experimental KRAS-driven pancreatic tumors to promote the accumulation of neutrophils at the expenses of CD8⁺ and CD4⁺ T cells, resulting in accrued resistance to ICIs that can be ameliorated by neutrophil depletion (207). Furthermore, the loss of Trp53 (the mouse homolog of human TP53) specifically in KRAS-driven pancreatic cancer cells promotes the recruitment of immunosuppressive myeloid cells to the TME, culminating with accelerated tumor progression in the context of suppressed anticancer T cell–dependent immunity (208). In breast cancer, loss of p53 is also accompanied by the secretion of WNT ligands that stimulate local macrophages to produce IL1β, favoring a systemic inflammatory response that supports metastatic dissemination (209).

In line with these observations, p53 reactivation has been associated with restored cancer immunosurveillance and responsiveness to immunotherapy in numerous tumor models. For instance, pharmacologic MDM2 inhibition by nutlin-3a stimulates systemic antitumor immunity and tumor regression in highly infiltrated mouse lymphoma models (210) as well as in preclinical models of melanoma, especially when combined with an AURKA inhibitor (211). Moreover, pharmacologic MDM2 inhibition promotes interferogenic effects downstream of the derepression of endogenous retroviruses that are sensed as cytosolic double-strand RNA (dsRNA; ref. 212), and reportedly upregulates the costimulatory molecule CD80 on epithelial cancer cells (213). Along these lines, restoring p53 expression with nanoparticles reconfigures the immunologic TME in mouse models of HCC and oral squamous cell carcinoma as it synergizes with PD-1–targeting ICIs (214, 215). Similar immunostimulatory effects have been reported for nutlin-3a and a peptide-based MDM2-targeting vaccine in syngeneic models of HNSCC (216), for HDM201 (a potent and selective second-generation MDM2 inhibitor) combined with PD-1 or PD-L1 blockers in models of colorectal cancer and TNBC (217), as well as the for MDM2 antagonist APG-115 plus PD-1 inhibitors in models of colorectal cancer and HCC (218). Pharmacologic MDM2 antagonism has also been shown to increase the susceptibility of malignant cells to lysis by immune effector cells (219). Whether this effect mechanistically depends on p53, however, remains unclear. Further supporting the critical role of anticancer immunosurveillance in the efficacy of p53-restoring strategies, the orally active MDM2 antagonist DS-527 mediates optimal antileukemic effects only in immunocompetent hosts (220).

Intriguingly, MDM2 expression in T cells ensures STAT5 stability in support of tumor-targeting immunity, and disruption of the MDM2–p53 interaction with APG-115 exacerbates this effect by increasing the amount of p53-unbound MDM2 available for binding the STAT5 inhibitor Cbl proto-oncogene (CBL; ref. 221). Thus, disrupting the MDM2–p53 interaction also mediates immunostimulatory effects that originate from immune, rather than malignant, cells. Of note, mitochondrial antiviral signaling protein (MAVS), a signal transducer in cytosolic dsRNA sensing, has been proposed to interfere with the p53–MDM2 interaction and hence stabilize p53 in support of oncosuppression (222). Consistent with this notion, Mavs^−/− mice are more sensitive to DSS/AOM-driven colorectal carcinogenesis than their WT counterparts (222). However, the relative contribution of cell-extrinsic inflammatory pathways versus cell-intrinsic genetic defects to accrued colorectal carcinogenesis in Mavs^−/− mice remains to be clarified. Similar to p53, the p53 interactor tumor protein p53 binding protein 1 (TP53BP1, best known as 53BP1) has been shown to preserve inflammatory homeostasis in physiologic conditions, largely reflecting its ability to inhibit the formation of CGAS-activating cytoplasmic chromatin fragments (CCF; ref. 223). Mitochondrial dysfunction, however, initiates an oxidative pathway culminating with 53BP1 inhibition, CCF accumulation, and CGAS-dependent cellular senescence (223), which is coupled to the secretion of numerous proinflammatory factors. The ultimate effect of such a senescence-associated secretory phenotype on oncogenesis and tumor progression exhibits considerable context dependency (224).

Besides contributing to the cell-cycle arrest driven by DDR signaling (225, 226), the ATR substrate p38^MAPK and its effector MAPK activated protein kinase 2 (MAPKAPK2, best known as MK2) appear to be linked to the regulation of inflammatory responses in a complex manner. p38^MAPK has been involved in the immunogenicity of cancer cells injured by DNA-damaging agents (108). Moreover, the macrophage-specific deletion of Mapk14 limits DSS-driven colitis (227). Similar results have been obtained in Mapkapk2^−/− mice (228), as well as in mice specifically lacking MK2 in the myeloid compartment (229, 230). Furthermore, MK2 has been suggested to promote replication stress as driven by DNA damage (231), which is also expected to promote inflammatory responses (see below). However, whole-body pharmacologic p38^MAPK inhibition as well as Mapk14 deletion from intestinal epithelial cells (IEC) appears to aggravate, rather than limit, the inflammatory and oncogenic effects of DSS/AOM (232, 233). Thus, although p38^MAPK/MK2 signaling in macrophages underlies DSS/AOM-driven oncogenesis, the same pathway may mediate oncosuppressive effects in IECs, potentially by preserving epithelial barrier function (233). Intriguingly, such a protective effect may no longer be relevant during tumor progression, as demonstrated by the fact that deleting Mapk14 in established DSS/AOM-driven carcinomas decelerates, rather than accelerates, tumor growth (233). Although this latter finding has been attributed to a cancer cell–intrinsic mitogenic role for p38^MAPK signaling (233), the aforementioned observations exemplify the complexity of the links between the DDR and inflammation in different cell types.

WEE1, a kinase that inhibits cell-cycle progression by antagonizing the CHEK1 and CHEK2 substrate cell division cycle 25A (CDC25A) and is involved in DDR signaling, suppresses inflammatory responses in various preclinical cancer models. In line with this notion, pharmacologic WEE1 inhibition with AZD1775 potently activates STING1 in mouse SCLC, resulting in tumor infiltration by CD8⁺ T cells and restored sensitivity to PD-L1 blockers (234). Similar effects have been documented in mouse models of ovarian cancer cotreated with WEE1 and ATR inhibitors (198), as well as in models of gastric cancer lacking MUS81 structure-specific endonuclease subunit (MUS81; ref. 235). Moreover, WEE1 inhibitors appear to promote TNF secretion by ALL cells (191). Along these lines, WEE1 inhibition boosts anticancer immunity driven by RT in immunocompetent mice bearing syngeneic TNBC (236) and other tumors (237), at least in some setting synergizing with ICIs (237). Intriguingly, at least part of these effects originates from the ability of WEE1 inhibitors to reverse the G₂—M block induced by RT (237), which has been associated with reduced sensitivity to immune effector molecules (238) similar to the epithelial–mesenchymal transition (another setting in which WEE1 inhibition restores sensitivity to immune effectors; ref. 239). Of note, WEE1 inhibition has also been associated with (i) the activation of ICD in mouse models of melanoma, a process that was exacerbated by concomitant AKT serine/threonine kinase 1 (AKT1) blockage, resulting in restored NK cell–dependent immunosurveillance and ICI sensitivity (240); (ii) the derepression of endogenous retroviruses (ERV), culminating with an interferogenic response driven by cytosolic dsRNA sensors independent of CGAS and STING1 (241); and (iii) the downregulation of baseline PD-L1 expression in pancreatic cancer cells, an effect that was magnified by concomitant ATM inhibition (242). Taken together, these observations exemplify the multipronged immunostimulatory potential of WEE1 inhibitors.

In summary, multiple DDR signal transducers are key regulators of inflammatory responses in both healthy and malignant cells (Supplementary Table S6).

Replication stress occurs when the DNA replication machinery encounters obstacles, resulting in the stalling, collapse, or breakage of the Y-shaped DNA structures where DNA replication occurs, also known as replication forks (Fig. 1; ref. 243). Replication stress is common in cancer cells, owing not only to their hyperproliferative phenotype but also the accumulation of DNA lesions that interfere with replication fork progression (243). Of note, at least some DDR-targeting agents currently investigated for their anticancer activity de facto operate as inducers/aggravators of replication stress. Among others, these agents include ATR, CHEK1, and PARP1 inhibitors (244–246). Fork stalling leads to the accumulation of ssDNA stretches that activate ATR signaling to stabilize the replication fork, hence enabling the completion of DNA replication in support of preserved genomic integrity (247). Processing of stalled replication forks by nucleases such as MRE11 or DNA replication helicase/nuclease 2 (DNA2) can promote the generation of ssDNA fragments with interferogenic effects, especially in the context of SAM and HD domain containing deoxynucleoside triphosphate triphosphohydrolase 1 (SAMHD1) or FA complementation group D2 (FANCD2) defects (114, 248, 249).

PCNA, which is critical for physiologic DNA replication by acting as a DNA-sliding clamp that enhances the processivity of DNA polymerases (250), also prevents the MRE11-dependent production of potentially interferogenic ssDNA species at stalled replication forks (251). Specifically, PCNA phosphorylation on Y211 ensures optimal PCNA processivity in support of suppressed immunogenic signaling and limited NK cell–mediated immune tumor eradication (251). Interestingly, a plasma membrane pool of PCNA has also been proposed to operate as a coinhibitory ligand for the NK cell receptor natural cytotoxicity triggering receptor 2 (NCR2, best known as NKp44; ref. 252). The impact of PCNA phosphorylation on Y211 on such an immunosuppressive activity and the signals that regulate PCNA exposure on the plasma surface of malignant cells remain to be characterized.

Unresolved replication stress can also promote the formation of micronuclei (253), which are prone to rupture and hence can drive CGAS activation (254). In line with these observations, aggravating replication stress with a CHEK1 inhibitor plus low-dose hydroxyurea drives potent anticancer immunity in preclinical models of melanoma, with a major role for natural killer T (NKT) cells and limited synergy with ICIs (255). Similar observations have been made in preclinical models of squamous cell carcinoma and SCLC, where replication stress was induced by treatment with lysine demethylase 4A (KDM4A) or cyclin-dependent kinase 7 (CDK7) inhibitors, respectively, although in these settings anticancer immunity was primarily driven by CD8⁺ T cells and therapeutic effects could be improved by ICIs (256, 257).

The excessive accumulation of R-loops, which are three-stranded nucleic acid structures physiologically involved in transcription, has also been associated with replication stress and consequent accumulation of interferogenic nucleic acids including ssDNA fragments (258, 259) and RNA-DNA hybrids (260). This process, which is negatively regulated by the helicase DEAD-box helicase 41 (DDX41) and ERCC excision repair 1, endonuclease noncatalytic subunit (ERCC1, a regulatory component of NER), appears to be particularly relevant not only for age-related chronic inflammation, at least in the pancreas (259), but also for hematopoietic alterations potentially linked to leukemogenesis (258, 261). In line with this notion, germline DDX41 mutations are linked to an increased susceptibility to myeloid neoplasms in humans (262).

Taken together, these observations suggest that replication stress can be targeted to elicit therapeutically relevant anticancer immunity (Supplementary Table S7).

NER is involved in the resolution of various DNA lesions, including intrastrand adducts and structures that distort the double helix. Damage sensing during NER is mediated by a heterotrimer comprising XPC complex subunit, DNA damage recognition and repair factor (XPC), RAD23 homolog B, nucleotide excision repair protein (RAD23B) and centrin 2 (CETN2). Next, the following factors are recruited: (i) the oligomeric transcription factor IIH (TFIIH), in support of DNA unwinding; (ii) XPA, DNA damage recognition and repair factor (XPA); (iii) RPA; (iv) ERCC excision repair 5, endonuclease (ERCC5, best known as XPG); and (v) an ERCC1/ERCC4 heterodimer (best known as XPF). Ultimately, this results in the incision of the damaged strand on both sides of the lesion by XPG and XPF, followed by DNA repair synthesis (Fig. 1; ref. 124). Only a few NER components have been implicated in the control of inflammatory reactions in cancer.

Defects in XPA have long been known to promote carcinogenesis in both humans and mice, but such an effect was largely attributed to genetic mechanisms (10). However, accelerated oncogenesis as driven by UV exposure and chemical carcinogens in Xpa^−/− mice appears to involve an inflammatory component. Specifically, Xpa^−/− mice exposed to UV light or topical dimethylbenz(a)anthracene (DMBA), a polycyclic aromatic hydrocarbon with pronounced oncogenic potential (263), develop a tumor-promoting dermal inflammatory response encompassing production of TNF, IL10, and prostaglandin E₂ (PGE₂) and accompanied by signs of systemic immunosuppression (264, 265). Along similar lines, Xpa^−/− (but not WT) mice accumulate CXCL1 both dermally and systemically upon UVB exposure, and both CXCL1 neutralization with a specific antibody and systemic administration of an antioxidant limit skin oncogenesis in this model (266).

Finally, lung inflammation elicited by the intratracheal administration of lipopolysaccharide considerably impairs NER along with the downregulation of XPA and ERCC4 (267), providing yet another example of a vicious cycle linking TPI to DDR defects and vice versa (Fig. 2). Lending further support to this possibility, DSS/AOM-driven colitis is associated with a pronounced downregulation of XPF prior to overt oncogenesis (268). The relative contribution of XPF downregulation to inflammation-driven colorectal oncogenesis, however, remains to be formally defined.

In summary, components of the NER pathway regulate inflammatory responses that influence oncogenesis, tumor progression or response to treatment (Supplementary Table S7).

The FA pathway repairs DNA interstrand crosslinks (ICL), toxic lesions that interfere with DNA replication and transcription (269). Schematically, ICLs are detected by FA complementation group M (FANCM) and FA core complex associated protein 24 (FAAP24) resulting in the recruitment of a large multimeric ubiquitin ligase commonly known as the FA core complex to lesioned DNA. The core complex monoubiquitinates FANCD2 and FA complementation group I (FANCI), resulting in their recruitment to DNA repair foci. ICL repair is then carried out by a multitude of DDR factors, including BRCA1 interacting helicase 1 (BRIP1, best known as FANCJ), BRCA1, BRCA2, and PALB2 (Fig. 1; ref. 269). ­Germline mutations in any FA components (including BRCA1 and BRCA2) cause a rare human genetic disease characterized by bone marrow failure, genomic instability, and increased predisposition for acute myeloid leukemia and squamous cell carcinoma (269). Moreover, somatic mutations in FA genes are common in FA-independent tumors, especially squamous cell carcinoma as well as bladder and pancreatic cancer (269). Intriguingly, at least some tumors bearing somatic alterations in FA genes exhibit increased sensitivity to PARP inhibition (269). Whether this is related to the immunostimulatory activity of PARP inhibitors, though, remains to be elucidated.

Although the contribution of dysregulated inflammation to FA remains to be established, multiple FA proteins have been shown to interact with the control of inflammatory responses. For instance, FANCD2, FANCA, FANCC, FANCF, and FANCL (as well as BRCA1 and BRCA2) all appear to be required for the parkin RBR E3 ubiquitin protein ligase (PRKN)-dependent autophagic removal of dysfunctional mitochondria (mitophagy; ref. 270), which otherwise can activate a variety of inflammatory pathways (53). In line with this notion, the loss of FANCD2, BRCA1, or BRCA2 can promote CGAS activation coupled to TNF secretion and simultaneously restore sensitivity to exogenous TNF in multiple human cancer cell lines (137). Although such a proinflammatory effect was accompanied by the accumulation of CGAS-positive micronuclei (137), and at least in some models mechanistically involved the derepression of CGAS-activating type I long interspersed elements (LINE-1; ref. 271), the relative contribution of mitochondrial dysfunction to cytokine production driven by FANCD2 defects remains elusive.

Deletion of Fancd2 (or Fanca) from the mouse bone marrow results in the development of defective Tregs (272), which at least in part may contribute to the immune dysregulation of FA patients. Such a defect may originate from the hypersensitivity of FANCD2-deficient hematopoietic stem cells (HSC) to inflammatory cues, resulting in a rewiring of bioenergetic metabolism toward fatty acid oxidation and oxidative phosphorylation (273). A partially compromised Treg compartment may also account for the hypersensitivity of Fancd2^−/− mice to carcinogen-driven skin oncogenesis (274). In this latter setting, FANCD2 has also been proposed to suppress oncogenesis by stabilizing TP63 (best known as p63; ref. 274). The consequences of this interaction on local inflammation, however, remain unexplored. Further supporting the notion that an intact FA pathway mediates anti-inflammatory effects, FANCA expression in HNSCC cells has been associated with radioresistance along with downregulation of IFN signaling coupled to an accrued secretion of immunomodulatory senescence-associated secretory phenotype (SASP) components (275). Conversely, high levels of FANCI have been linked to tumor infiltration by CD8⁺ and CD4⁺ T cells (but poor prognosis) in a cohort of patients with cervical cancer (276). The reasons underlying such an apparent discrepancy remain to be elucidated and may be linked to FA-unrelated functions of specific FA components.

Fancc^−/− mice exhibit defects in IFNγ secretion by CD4⁺ T cells, presumably linked to the ability of FANCC to physically interact with (and support the function of) STAT1 in response to cytokine stimulation (277, 278). That said, FANCC has also been proposed to interact with heat-shock proteins, hence limiting the detrimental effects of inflammation on HSCs (279). In line with this notion, HSCs from FA patients with defects in FANCC are hypersensitive to IFNγ (280).

In summary, the FA pathway globally resembles other DDR cascades in its functions at the crossroad of genetic stability and inflammatory homeostasis (Supplementary Table S7).

The abundant literature discussed above delineates a leitmotif whereby the DDR, in most (if not all) of its variants, constitutively operates to preserve genetic and inflammatory homeostasis. Thus, DDR defects not only favor the accumulation of genetic lesions that may drive oncogenesis and tumor progression but also initiate inflammatory responses that can influence developing neoplasms in an ambiguous fashion. On the one hand, TPI may support disease progression and resistance to therapy. On the other hand, local inflammation may favor immunosurveillance and instead support sensitivity to (immuno)therapy. In line with this notion, several clinical trials are currently investigating the therapeutic profile of DDR inhibitors optionally combined with ICIs or other immunotherapeutic agents (Table 2). Preliminary findings from some of these studies are encouraging, generally pointing to an acceptable toxicity profile and at least some degree of clinical activity (281–283). Moreover, various ICIs are being used in multiple clinical indications (and extensively tested in clinical trials) together with agents that cause DNA damage along with inflammatory reactions, including several chemotherapeutics (29) as well as RT (30). Although the actual implication of DDR-driven inflammatory responses in the efficacy of these latter combinations remains to be formally demonstrated in patients, abundant preclinical data support this hypothesis. In this context, it is worth noting that cellular senescence driven by DDR signaling appears to be particularly immunogenic (284). Finally, the ability of DNA-damaging agents to elicit tumor-targeting immune responses has been shown to emerge from injured/dying (but hitherto alive) cancer cells (108, 109). Thus, it is tempting to speculate (but remains to be formally demonstrated) that the use of DNA-damaging agents at suboptimal doses (favoring cellular injury or senescence over rapid RCD) may result in superior synergism with ICIs or other immunotherapeutics. Additional work is required to clarify this possibility. At least theoretically, DDR inhibitors also stand out as promising combinatorial partners for exogenously administered PRR agonists (e.g., STING1 agonists, oncolytic viruses). Indeed, although DDR inhibition may per se drive robust PRR signaling in (at least some) malignant cells (Fig. 3), local PRR agonism may also engage nonmalignant components of the TME (which are refractory to DDR inhibitors). With the exception of two studies pointing to a positive interaction between ATM or ATR inhibitors and oncolytic viruses (not necessarily in the context of superior tumor-targeting immunity; refs. 285, 286), this possibility remains largely unexplored.

One of the potential obstacles to the use of DDR inhibitors in cancer patients relates to a somehow nonnegligible potential for secondary, therapy-driven oncogenesis. This has already been documented for DNA-damaging therapeutics including RT (287), and may call for intratumoral (instead of systemic) drug administration to limit exposure for healthy tissues. From a biological perspective, one of the questions that requires immediate experimental attention involves the extent to which modulation of DDR components influences inflammation by (i) controlling the abundance of PRR agonists (as in the case of EXO1; ref. 125), (ii) engage DDR signal transducers (as in the case of p53 activation by CHEK2; ref. 8), or (iii) activate DDR-unrelated pathways (Fig. 4). Potential DDR-independent mechanisms that may be involved in the ability of some DDR components to regulate inflammation include control of redox homeostasis (as in the case APE1; ref. 52), modulation of the gut microbiome (as in the case of MSH2; refs. 69, 70), direct interactions between DDR components and PRRs or signal transducers thereof (as in the case of Ku70; refs. 98, 99), and an ectopic activity as PRR (as in the case of MRE11; Fig. 4; ref. 113). Precisely defining the signaling cascades that connect DDR components to the molecular machinery for inflammation may uncover novel targets that enable improved anticancer immune responses without weakening genetic homeostasis.

Despite these and other unknowns, DDR-targeting agents stand out as promising partners for immunotherapeutic agents including (but not limited to) ICIs, partly owing to their ability to modulate inflammation.

G. Kroemer reports grants from Kaleido, Lytix Pharma, PharmaMar, Osasuna Therapeutics, Samsara Therapeutics, Sanofi, Tollys, and Vascage during the conduct of the study; personal fees from Reithera, Hevolution, and Longevity Vision Funds outside the submitted work; two patents for WO2014020041-A1 and WO2014020043-A1 licensed to Bayer, a patent for WO2008057863-A1 licensed to Bristol Myers Squibb, a patent for WO2019057742A1 licensed to Osasuna Therapeutics, a patent for WO2022049270A1 licensed to PharmaMar, a patent for WO2022048775-A1 licensed to PharmaMar, a patent for EP2664326-A1 licensed to Raptor Pharmaceuticals, a patent for GB202017553D0 licensed to Samsara Therapeutics, and a patent for EP3684471A1 licensed to Therafast Bio; is a scientific cofounder of EverImmune, Osasuna Therapeutics, Samsara Therapeutics, and Therafast Bio; is on the board of directors for the Bristol Myers Squibb Foundation France; and is on the nonremunerated scientific advisory boards for Institut Servier. G. Kroemer's wife, Laurence Zitvogel, has held research contracts with GSK, Incyte, Lytix, Kaleido, Innovate Pharma, Daiichi Sankyo, Pilege, Merus, Transgene, 9 Meters Biopharma, Tusk, and Roche, was on the board of directors for Transgene, is a cofounder of EverImmune, and holds patents covering the treatment of cancer and the therapeutic manipulation of the microbiota. G. Kroemer's brother, Romano Kroemer, was an employee of Sanofi and now consults for Boehringer Ingelheim. A. Ciccia reports grants from the NIH, the Pershing Square Sohn Foundation, the Basser Center, the Mary Kay Foundation, and the Irma T. Hirschl and Monique Weill-Caulier Research Foundation during the conduct of the study. L. Galluzzi reports grants from Lytix, Promontory, and Onxeo, and personal fees from AstraZeneca, OmniSEQ, Longevity Labs, Inzen, the Luke Heller TECPR2 Foundation, Sotio, Onxeo, Noxopharm, Imvax, EduCom, and Boehringer Ingelheim outside the submitted work. No disclosures were reported by the other authors.

V. Klapp is supported by a grant from the Luxembourg National Research Fund (FNR; PRIDE19/14254520/i2TRON). A. Ciccia is supported by two R01 grants (#R01CA197774; #R01CA227450) and one P01 grant from NIH (#P01CA174653), by a Pershing Square Sohn Cancer Research (PSSCR) Award, by a BRCA Award from the Basser Center, by a Cancer Research Grant from the Mary Kay Foundation, and by the Irma T. Hirschl and Monique Weill-Caulier Research Award. G. Kroemer is supported by the Ligue contre le Cancer (équipe labellisée); Agence National de la Recherche (ANR)—Projets blancs; AMMICa US23/CNRS UMS3655; Association pour la recherche sur le cancer (ARC); Cancéropôle Ile-de-France; Fondation pour la Recherche Médicale (FRM); a donation by Elior; Equipex Onco-Pheno-Screen; European Joint Programme on Rare Diseases (EJPRD); Gustave Roussy Odyssea, the European Union Horizon 2020 Projects Oncobiome and Crimson; Fondation Carrefour; Institut National du Cancer (INCa); Institut Universitaire de France; LabEx Immuno-Oncology (ANR-18-IDEX-0001); a Cancer Research ASPIRE Award from the Mark Foundation; the RHU Immunolife; Seerave Foundation; SIRIC Stratified Oncology Cell DNA Repair and Tumor Immune Elimination (SOCRATE); and SIRIC Cancer Research and Personalized Medicine (CARPEM). This study contributes to the IdEx Université de Paris ANR-18-IDEX-0001. The laboratory of L.G. (as a PI unless otherwise indicated) is or has been supported by two Breakthrough Level 2 grants from the US DoD BCRP (#BC180476P1; #BC210945), by a Transformative Breast Cancer Consortium Grant from the US DoD BCRP (#W81XWH2120034, PI: Formenti), by a U54 grant from NIH/NCI (#CA274291, PI: Deasy, Formenti, Weichselbaum), by a STARR Cancer Consortium grant (#I16-0064), by the 2019 Laura Ziskin Prize in Translational Research (#ZP-6177, PI: Formenti) from the Stand Up to Cancer (SU2C), by a Mantle Cell Lymphoma Research Initiative (MCL-RI, PI: Chen-Kiang) grant from the Leukemia and Lymphoma Society (LLS), by a Rapid Response Grant from the Functional Genomics Initiative (New York, US), by startup funds from the Dept. of Radiation Oncology at Weill Cornell Medicine (New York, US), by industrial collaborations with Lytix Biopharma (Oslo, Norway), Promontory (New York, US) and Onxeo (Paris, France), as well as by donations from Promontory (New York, US), the Luke Heller TECPR2 Foundation (Boston, US), Sotio a.s. (Prague, Czech Republic), Lytix Biopharma (Oslo, Norway), Onxeo (Paris, France), Ricerchiamo (Brescia, Italy), and Noxopharm (Chatswood, Australia).