Introduction

The process of danger sensing and signalling is essential for the eradication of infections, the resolution of sterile tissue damage and the elimination of neoplasia. It allows the organism to eliminate microbial threats, initiate tissue repair and to re-establish homeostasis. Danger-induced responses include the expression of type I and type III interferons, the induction of pro-inflammatory genes, the activation of inflammasome cascades, the initiation of autophagy and the engagement of cell death pathways1. One major danger signal is abnormal subcellular localization of DNA2 (Box 1). DNA is normally compartmentalized to the nucleus and mitochondria, yet high levels of DNA in the cytoplasm or within the endolysosomal niche stimulate danger signalling. Normally, low levels of DNA occur in these compartments, for example, during genome replication, retrotranscription of endogenous retroelements and phagocytosis of apoptotic cells. However, DNases located in the extracellular space (such as DNase I), endosomes (such as DNase II) and the cytoplasm (such as DNase III or three prime repair exonuclease 1 (TREX1)) digest mislocalized DNA to levels below the threshold for danger signalling3,4,5. Nevertheless, with the appearance of high amounts of exogenous DNA, for example, during infections, certain thresholds seem to be passed so that DNA can initiate sensing and associated signalling cascades (Fig. 1). This is also observed if DNases have reduced activity or if this machinery is overwhelmed by endogenous substrates. In the past few years, it has emerged that the appearance of cytoplasmic DNA can also engage different types of programmed cell death (PCD) pathways, and we are now beginning to understand the molecular basis for the interplay and crosstalk between cell death and other responses following the sensing of DNA as a danger signal. In this Review, we first describe key DNA sensors and how they induce cytokine responses. This is followed by a detailed description of the current knowledge on DNA-stimulated cell death and a discussion of how this may impact on immune defence and pathology during infections, inflammatory diseases and cancer.

Fig. 1: Hallmark cytokine responses induced by DNA-sensing pattern recognition receptors.
figure 1

DNA originating from microorganisms, mitochondria or phagocytosed apoptotic cells or leaked from the nucleus can accumulate in the cytoplasm, where it is sensed by pattern recognition receptors. This initiates danger signalling. The production of type I interferons and IL-1β is the most well-studied immunological response downstream of cytosolic DNA. Toll-like receptor 9 (TLR9) senses CpG-rich DNA in endosomes and signals via myeloid differentiation primary response protein 88 (MYD88) to activate nuclear factor-κB (NF-κB) and interferon regulatory factor 7 (IRF7), the latter being abundantly expressed in plasmacytoid dendritic cells. This leads to strong activation of genes encoding tumour necrosis factor (TNF) and IFNα. Double-stranded DNA (dsDNA) in the cytoplasm is also sensed by absent in melanoma 2 (AIM2), which promotes assembly of inflammasomes with caspase 1 activation and downstream cleavage of pro-IL-1β to the bioactive cytokine. Additionally, dsDNA is sensed by cyclic GMP–AMP (cGAMP) synthase (cGAS), leading to enzymatic activation and synthesis of the stimulator of interferon genes (STING) agonist 2'3'-cGAMP, thus leading to activation of the kinase TANK-binding kinase 1 (TBK1), phosphorylation of the transcription factor IRF3 and transcriptional activation of the promoters for the type I (IFNα and/or IFNβ) and type III (IFNλ) interferon genes. ASC, the adaptor protein ASC; IKK, IκB kinase; IRAK, IL-1R-associated kinase; TREX1, three prime repair exonuclease 1.

Full size image

Immune stimulation by DNA

The immunostimulatory properties of DNA were first reported >50 years ago6. In particular, the ability of DNA to induce the production of type I and type III interferons and IL-1β is a hallmark of DNA-stimulated immune responses. DNA is sensed in endolysosomes by Toll-like receptor 9 (TLR9) and in the cytoplasm by cyclic GMP–AMP (cGAMP) synthase (cGAS) and absent in melanoma 2 (AIM2) and under some circumstances also by RNA polymerase III7,8,9,10,11,12,13,14. Below, we summarize the key DNA sensors that initiate interferon and IL-1β production.

DNA sensing by TLR9

TLR9 was the first DNA-sensing pattern recognition receptor (PRR) to be identified7. Like other nucleic acid-sensing TLRs, it localizes to the endolysosomal compartment and recognizes DNA at this site15. The most potent ligand for TLR9 is unmethylated CpG DNA, which occurs abundantly in bacteria7,16. TLR9 signals through the common TLR adaptor myeloid differentiation primary response protein 88 (MYD88) to drive nuclear factor κB (NF-κB)-mediated transcription and can also activate the transcription factor interferon regulatory factor 7 (IRF7), leading primarily to IFNα production17 (Fig. 1). TLR9 is expressed in a very cell type-restricted manner in humans, with B cells and plasmacytoid dendritic cells (pDCs) selectively expressing high levels of this PRR18. As pDCs express high levels of IRF7, TLR9 is a potent inducer of type I interferon in this cell type19.

DNA sensing by AIM2

The second key DNA sensor to be identified was the pyrin and HIN domain (PYHIN) family protein AIM2, which activates the inflammasome pathway11,12,13,14. The inflammasome is a protein complex that initiates the cleavage of specific proteins to activate and regulate immune reactions. By far, the best-studied inflammasome substrate is pro-IL-1β, which needs to be cleaved to render bioactive IL-1β — a central pro-inflammatory cytokine20,21. AIM2 activates the inflammasome pathway in response to exogenous and endogenous DNA challenge11,12,13,14 (Fig. 1). Following DNA sensing, AIM2 recruits the adaptor protein ASC, which in turn forms a filamentous structure that serves as a recruitment platform for pro-caspase 1. The recruitment of pro-caspase 1 to this structure results in its activation and thereby enables this protease to process its substrates21,22,23. There are reports showing that another PYHIN protein, IFI16, as well as NOD-, LRR- and pyrin domain-containing 3 (NLRP3), can also induce caspase 1 activation in response to DNA24,25, the latter possibly through sensing of DNA-stimulated cell death, as discussed below26. These results may suggest species-specific and cell-specific pathways for IL-1β processing and release in response to DNA detection.

DNA sensing by cGAS

Both of the above-mentioned DNA sensors are largely confined to the myeloid compartment; hence, they cannot explain key parts of the immune response, for example, the sensing of DNA in virus-infected cells of non-myeloid origin. This notion led to an intense effort to identify cytosolic DNA sensors stimulating interferon expression, which led to the identification of cGAS8,27. Other cytosolic DNA sensors were also proposed but were either not independently confirmed28,29, were found to play more subordinate roles9,10,30 or were shown to be accessory proteins in the cGAS signalling pathway31,32,33. Upon DNA binding, the enzymatic activity of cGAS is initiated, leading to production of the cyclic dinucleotide (CDN) 2ʹ3ʹ-cGAMP8,27. This CDN is a ligand for the adaptor protein stimulator of interferon genes (STING), which in turn recruits TANK-binding kinase 1 (TBK1)27. This leads to TBK1-mediated phosphorylation at serine 366 in the carboxy-terminal tail (CTT) of STING34. IRF3 is then recruited to phosphorylated S366 of STING, positioning the latent transcription factor for TBK1-mediated phosphorylation and activation, eventually driving expression of IFNA and IFNB genes35 (Fig. 1). Consequently, the CTT of STING is essential for cGAS–STING-induced interferon expression. In addition to the IRF–interferon pathway, cGAS–STING signalling leads to the activation of NF-κB and the induction of inflammatory cytokines26,36,37,38. The mechanism of STING-mediated activation of NF-κB remains poorly understood. Finally, it has recently emerged that the AIM2 pathway negatively regulates cGAS activity by stimulating caspase 1-mediated cleavage of cGAS as well as by depleting intracellular potassium39,40. Thus, there is crosstalk between the DNA-stimulated pathways with an impact on cytokine responses but also between PCD pathways, as is discussed below.

Biological importance of the DNA–interferon and DNA–IL-1β axes

It is well documented that DNA-driven cytokine responses are of critical biological importance. Mechanistically and functionally, one can distinguish DNA-driven type I interferon responses from DNA-triggered IL-1β maturation. For instance, control of herpes simplex virus 1 (HSV1) infection in the central nervous system is dependent on viral sensing and signalling by microglia through the cGAS–STING axis leading to type I interferon production41,42,43. For bacterial infections, the role of STING-driven interferon responses differs substantially depending on the bacterial species. For instance, although both Legionella pneumophila and Listeria monocytogenes induce type I interferon expression in a STING-dependent manner, the downstream activity of type I interferons is protective in the case of L. pneumophila infection but deleterious in the case of L. monocytogenes infection in mice44,45,46,47. There is also strong evidence for the critical role of the interferon response triggered by endogenous DNA through the cGAS–STING pathway. For instance, individuals with TREX1 or DNase II deficiency develop autoinflammatory diseases48,49, and murine studies have demonstrated this to be dependent on STING and the type I interferon receptor4,50,51,52. Recently, it was reported that the release of endogenous DNA during acute conditions, such as myocardial infarction, also triggers STING-dependent interferon responses, thereby augmenting the inflammatory pathology and decreasing survival53. Finally, DNA-induced interferon responses also seem to promote the control of cancer. The mechanisms include the activation of antigen-presenting cells by tumour cell DNA, which promotes antitumour T cell responses54,55.

DNA-induced IL-1β is also likely to play important roles in both defence and tissue damage. Studies from mouse disease models have shown similar phenotypes in Aim2−/− and Il1r−/− mice in several cases. For instance, mice deficient in AIM2 have reduced expression of IL-1β and exhibit elevated susceptibility to Francisella tularensis infection, as is also observed in Ilr1−/− mice56,57,58,59. Also, host DNA released from injured cells in the lungs during influenza A virus infection induces early AIM2-dependent IL-1β production60,61. Accordingly, AIM2-deficiency ameliorated symptoms and lung pathology, similar to the phenotype of IL-1R-deficient mice61,62. However, another study reported that AIM2-deficiency led to elevated inflammatory response to influenza A virus infection60. This issue remains to be resolved. Regarding autoinflammatory diseases, there is evidence supporting the idea that the pathological IL-1β response in psoriasis is driven by the AIM2 inflammasome63. Finally, chemotherapy causes severe gastrointestinal tract toxicity in a manner dependent on DNA damage and AIM2-induced IL-1β production, although the total dependence on AIM2 is somewhat surprising given the presumed release of multiple damage-associated molecular patterns (DAMPs) during cell lysis64. Collectively, the DNA–interferon and DNA–IL-1β axes play important roles in host defence and inflammatory diseases.

Pathways for DNA-stimulated cell death

As DNA sensors have primarily been regarded as PRRs, the main focus has been put on their cytokine outputs and other outcomes that depend on de novo gene expression. However, in recent years, it has emerged that DNA sensing also induces additional cellular events, such as autophagy65,66,67 (Fig. 2a). Moreover, it has been shown that cytosolic DNA stimulates signalling pathways that lead to different types of PCD (Box 2). These include pathways that induce apoptosis, pyroptosis, necroptosis and lysosomal cell death12,14,68,69,70. The focus of this Review is to describe the current knowledge on DNA-driven cell death (Fig. 2b), highlighting the pathways involved and their physiological relevance.

Fig. 2: Cytoplasmic DNA activates a broad range of stress responses.
figure 2

a | In addition to the production of type I interferon and IL-1β, several other cellular responses are induced by cytoplasmic DNA. These include nuclear factor-κB (NF-κB)-induced gene expression, autophagy and several programmed cell death (PCD) pathways, such as apoptosis, necroptosis and pyroptosis. b | An overview of the current knowledge on PCD pathways induced by cytoplasmic DNA, including the interactions between the pathways, is shown. Blue arrows indicate interactions dependent on de novo gene expression. AIM2, absent in melanoma 2; cGAMP, cyclic GMP–AMP; cGAS, cyclic GMP–AMP synthase; ER, endoplasmic reticulum; GSDMD, gasdermin D; IRF3, interferon regulatory factor 7; MLKL, mixed-lineage kinase domain-like protein; NLRP3, NOD-, LRR- and pyrin domain-containing 3; RIPA, RLR-induced IRF3-mediated pathway of apoptosis; RIPK3, receptor-interacting serine/threonine-protein kinase 3; STING, stimulator of interferon genes; TNF, tumour necrosis factor; ULK, Unc-51-like autophagy activating kinase.

Full size image

The AIM2 pathway in pyroptosis and apoptosis

One of the first reports linking DNA recognition to the induction of cell death was provided by Stacey and colleagues71. These authors observed a dramatic decline in the viability of murine macrophages upon electroporation with different types of DNA molecules. Interestingly, the observed cell death did not show hallmarks of apoptosis, indicating that another cell death modality might be at play. In retrospect, this study provided a functional characterization of the first cytosolic DNA receptor to be discovered, namely, AIM2. As outlined above, AIM2 is a cytosolic PYHIN protein that harbours an amino-terminal pyrin domain and a carboxy-terminal HIN200 domain. AIM2 detects cytosolic DNA of >70–80 bp in length in a sequence-independent manner22. This results in the formation of a pyrin domain platform that functions as a seed to recruit the universal inflammasome adaptor molecule ASC and the protease caspase 1, which cleaves gasdermin D to induce pyroptosis (Fig. 3a). In mouse macrophages, cytosolic DNA delivery mainly triggers AIM2–ASC-dependent cell death, with caspase 1-initiated pyroptosis playing the predominant role. However, in the absence of caspase 1 and when there are low amounts of DNA, ASC recruits and activates caspase 8, which then is the apical caspase in a subsequent apoptotic cell death response72 (Fig. 3a). While this switches the quality of the cell death to apoptosis, it still represents an AIM2–ASC-dependent PCD. These findings extend the role of AIM2 and ASC beyond inflammasome activation, which might be relevant in cells types that do not express caspase 1 yet do express AIM2, ASC and caspase 8. The physiological importance of this alternative AIM2-driven PCD pathway has yet to be fully understood.

Fig. 3: Pathways in DNA-stimulated cell death.
figure 3

a | Cell death pathways stimulated by absent in melanoma 2 (AIM2) in response to DNA sensing are shown. b | Cell-autonomous death pathways stimulated through stimulator of interferon genes (STING)-dependent mechanisms are shown. c | Stimulation of apoptosis and necroptosis by STING-dependent paracrine signalling is shown. ASC, the adaptor protein ASC; BAX, BCL-2-associated X gene; cGAMP, cyclic GMP–AMP; cGAS, cyclic GMP–AMP synthase; ER, endoplasmic reticulum; GSDM, gasdermin; IFNαR, IFNα receptor; IRF3, interferon regulatory factor 3; ISGF3, interferon-stimulated gene factor 3; MLKL, mixed-lineage kinase domain-like protein; NF-κB, nuclear factor-κB; NLRP3, NOD-, LRR- and pyrin domain-containing 3; P, phosphorylation; PUMA, p53 upregulated modulator of apoptosis; RIPA, RLR-induced IRF3-mediated pathway of apoptosis; RIPK, receptor-interacting serine/threonine-protein kinase; TNF, tumour necrosis factor; TNFR, TNF receptor.

Full size image

STING and apoptosis

One of the first reports that implemented STING signalling in the triggering of cell death came from studies in T cells infected with human T cell leukaemia virus type 1 (ref.73). Here, it was observed that primary T cells underwent apoptosis that was dependent on the formation of a complex between IRF3 and BCL-2-associated X gene (BAX), which promotes apoptosis through the mitochondrial pathway73,74. Interestingly, this IRF3-mediated pathway of apoptosis (also known as RLR-induced IRF3-mediated pathway of apoptosis (RIPA)) operates independently of the transcriptional activity of IRF3 but still requires the presence of TBK1 (ref.74). This TBK1 requirement, yet the independence of IRF3 phosphorylation, might be attributed to the fact that TBK1-dependent phosphorylation of the adaptor protein (STING or MAVS) is still required to allow IRF3 recruitment34. Further studies revealed that linear polyubiquitylation of IRF3 by linear ubiquitin chain assembly complex (LUBAC) played a critical role in this particular cell death pathway75. In light of the fact that IRF3 is activated downstream of a number of signalling hubs, this type of cell death is not specific to cGAS–STING-dependent DNA recognition but is also seen in the context of RNA-activated RIG-I signalling. All in all, these results imply that cGAS–STING signalling triggers apoptosis independently of transcriptional activity through the mitochondrial apoptosis pathway (Fig. 3b). Of note, most studies reporting on RIPA have been performed in fibroblast cell lines.

Another study that has reported on the capacity of STING to induce apoptosis proposed that the transcriptional activity of STING was required for apoptosis induction in T cells76. To this end, it was shown that the de novo expression of the BH3-only proteins via STING signalling strongly correlated with the sensitivity of cells to succumb to apoptosis. This pro-apoptotic gene expression programme correlated with STING expression levels, serving as a possible explanation as to why mouse T cells (which show high STING expression) but not mouse macrophages (which show low STING expression) die of apoptosis upon STING activation. Again, as the induction of pro-apoptotic BH3-only proteins is also seen for other PRR pathways77, this type of cell death induction is not unique to cGAS–STING signalling (Fig. 3b). Along with these reports, additional studies have reported on the capacity of STING agonists to trigger apoptosis, yet mechanistic studies were not always performed that would allow the inference of the underlying mode of action. For example, before the discovery of STING, several studies documented the pro-apoptotic effects of the tumour vascular-disrupting agent DMXAA (5,6-dimethylxanthenone-4-acetic acid) in mice in vitro and in vivo78,79. With the discovery of DMXAA being a direct and specific agonist for mouse STING80,81, these studies retrospectively imply that STING-dependent effects were responsible for at least part of these observations.

Finally, cGAS–STING-induced type I interferon can stimulate genes that induce apoptosis in a paracrine manner82. Most notably, type I interferon stimulates expression of tumour necrosis factor (TNF)-related apoptosis-inducing ligand (TRAIL; also known as TNFSF10), which acts via death receptor 5 to induce apoptosis82,83 (Fig. 3c). This pathway has been demonstrated to be operative in vivo and to contribute to DNA-stimulated cell death and immunopathology82.

STING and lysosomal cell death

The CTT of STING and its capacity to recruit and activate TBK1 evolved along with the emergence of IRF3 and type I interferons in the vertebrate system84. In light of this observation, it seems plausible that certain STING functions do not require the CTT, the downstream kinase TBK1 or the functionality of the CTT to recruit IRF3 for its activation. For example, the profound lack of memory CD4+ T cells that is observed in patients with gain-of-function mutations in STING appears to be connected to such a CTT-independent STING functionality85. To this end, transgenic expression of a constitutively active STING molecule exerts a strong antiproliferative effect on CD4+ T cells, and this functionality is also seen for a STING construct lacking the majority of the CTT that is incapable of activating TBK1 (ref.86). This indicates that canonical STING signal transduction is not required for this effect. Another, signalling-independent function of STING activation was observed in human myeloid cells. Here, STING activation leads to its translocation to lysosomes and the subsequent permeabilization of lysosomal membranes26 (Fig. 3b). This in turn triggers a cell death pathway that is compatible with the concept of lysosomal cell death. STING-induced lysosomal cell death is observed in cells lacking TBK1 and IκB kinase-ε (IKKε) and when expressing a mutant version of STING that cannot facilitate IRF3 phosphorylation via TBK1. As such, it is clearly distinct from the canonical signalling function of STING. Indeed, in human myeloid cells, STING-dependent lysosomal cell death leads to the secondary activation of the NLRP3 inflammasome, which is sensitive to cell membrane perturbation that is triggered by lysosomal cell death26. These results are insofar intriguing, as human myeloid cells do not utilize the dedicated DNA-sensing AIM2 inflammasome for this purpose. Of note, in mouse embryonic fibroblasts and macrophages, STING is also translocated to the lysosome upon activation87; yet in these cells, lysosomes function to degrade STING but not to induce cell death. As such, it appears that cell type-specific properties dictate the involvement of lysosomes in STING biology.

Finally, although not demonstrated to be dependent on DNA and cGAS, STING is involved in endoplasmic reticulum (ER) stress88. This leads to downstream apoptosis through a mechanism dependent on the ER-resident proline-rich receptor-like protein kinase (PERK; also known as EIF2AK3) and the downstream target transcription factor CCAAT-enhancer-binding protein homologous protein88.

STING and necroptosis

A third type of cell death stimulated by DNA is necroptosis70 (Fig. 3c). In this regard, it was reported that DNA transfection and STING agonists — in the context of pan-caspase inhibition — activate necroptosis in mouse fibroblasts, as well as in bone marrow-derived macrophages70,89. This DNA-induced necroptosis was dependent on STING-induced TNF and type I interferon and could also be triggered by other TNF-inducing and type I interferon-inducing stimuli, such as RIG-I or TLR3 agonists70,89. As such, under these conditions, there seems to be no requirement for a STING-specific signal to induce necroptosis. Nevertheless, it has been shown that tonic type I interferon production, induced through the cGAS–STING pathway, generally lowers the threshold for necroptosis. This is governed by the expression of the key necroptosis executor mixed-lineage kinase domain-like protein (MLKL)90. In this regard, TLR4-triggered necroptosis was also affected by cGAS and STING deficiency, which could be explained by a loss of tonic type I interferon production and an associated decrease in MLKL expression90. Of note, while all these in vitro experiments require the concomitant inhibition of caspases to render cells sensitive to STING-dependent necroptosis, in vivo experiments suggest that necroptosis can indeed be achieved by single STING agonism. To this end, mice treated with high doses of a STING ligand succumbed to lethal shock associated with pro-inflammatory cytokine production, which could in part be rescued by deficiency of receptor-interacting serine/threonine-protein kinase 3 (RIPK3) or MLKL89. Future studies will be required to address the role of this pathway in the context of physiological cGAS–STING activation.

Role of cGAS–STING in autophagy-dependent cell death

Finally, it is also possible that cytosolic DNA can induce autophagy-dependent cell death, which is a form of cell death induced in an autophagy-dependent manner, notably in response to strong inducers of autophagy91. The exact mechanisms that execute cell killing in autophagy-dependent cell death are not fully understood but could give us key information on what governs the switch from healthy to deadly forms of autophagy. It is important to emphasize that there is currently no experimental evidence for DNA-activated autophagy-dependent cell death. However, as DNA strongly activates autophagy through both STING-dependent and STING-independent pathways66,67,92, and bacteria-derived CDNs trigger extensive ER-phagy88, it seems possible that DNA could induce cell death through autophagy under some conditions and in some cell types.

In summary, the accumulation of DNA in the cytoplasm can promote signalling leading to many distinct types of cell death. At this stage, there is limited knowledge on the factors that determine which of the death pathways are activated in a given cell type and how the DNA-stimulated death pathways interact with other DNA-stimulated reactions and host responses in general.

Biological functions of different types of cell death

The different types of cell death differ with respect to how they impact on host responses to danger sensing (Table 1). Most notably, apoptosis is typically understood to be non-inflammatory in nature owing to the lack of release of cellular content by apoptotic cells and their rapid clearance by phagocytic cells. Interestingly, the avoidance of DNA recognition plays an important role in this process. A recent study showed that tissue-resident macrophages are programmed to clear apoptotic cells in an immunologically silent manner owing to several key features of these macrophages, including high expression of apoptotic cell recognition receptors and low expression of TLR9 (ref.93). Moreover, cell-intrinsic properties function to avoid activation of the immune system by DNA in the course of apoptosis. As such, the non-inflammatory nature of apoptosis proceeds despite mitochondrial outer membrane permeabilization leading to the cytoplasmic accumulation of mitochondrial DNA, which could in principle be sensed by cGAS94,95. Indeed, apoptotic caspases block induction of type I interferon through the cGAS–STING pathway following release of mitochondrial DNA94,95, whereas the exact substrates in the cGAS–STING pathway that are targeted by the apoptotic caspases remain to be identified. Thus, apoptotic caspases both induce cell death and prevent the dying cells from activating a DNA-driven interferon response. Despite the generally non-inflammatory nature of apoptosis, it is important to note that apoptotic cells can undergo secondary necrosis96. Thus, if not rapidly eliminated by phagocytes, apoptosis can lead to inflammation through secondary necrosis. Of note, recent studies have blurred the definitions of apoptotic signalling and programmed necrosis. For certain cell types, it has been shown that caspase 3 can cleave gasdermin E (also known as DFNA5), which then results in pyroptotic cell death with the associated release of cytoplasmic content97,98. Therefore, as determined by multiple factors, including the expression of gasdermin E, it appears plausible that apoptotic signalling cascades can evoke mixed types of PCD. Collectively, if apoptotic cells are not properly eliminated or if apoptotic cells do not prevent the cell-autonomous recognition of DAMPs, apoptosis can stimulate inflammation.

Table 1 Features of types of cell death induced by cytoplasmic DNA
Full size table

Although necroptosis and pyroptosis are induced by different pathways, they are both necrotic types of cell death with membrane rupture and release of intracellular content. One important difference between necroptosis and pyroptosis is that the latter is associated with release of inflammasome-cleaved bioactive cytokines, most notably IL-1β20,21. Like pyroptosis and necroptosis, lysosomal cell death is also necrotic with plasma membrane rupture and release of cytoplasmic content, including DAMPs. It remains unknown whether cells dying through these different types of programmed necrosis differentially allow release of intracellular content, thereby resulting in a different repertoire of potential DAMP molecules being released.

Autophagy-dependent cell death is characterized by extensive cytoplasmic vacuolization eventually leading to phagocytic uptake and lysosomal degradation91. As such, this type of cell death is most likely an intermediate between apoptosis and different forms of necrosis with respect to stimulation of inflammation. Thus, DNA-stimulated cell death can contribute to both promotion and resolution of inflammation depending on the biological context.

Biological impact of DNA-stimulated PCD

Regarding the role of DNA-activated cell death pathways in defence and disease, there is now accumulating evidence for these pathways playing important roles in both the promotion and the control of pathology (Fig. 4). For most studies, it is difficult to discern the contribution of DNA-driven cytokine responses from the PCD pathways being activated, as genetic tools to uncouple these two responses from one another are often not available. At the same time, it still remains challenging to ascertain a certain type of cell death in settings in vivo. In the following section, we provide examples in which the distinction between direct PRR signalling effects and PCD cascades was addressed.

Fig. 4: Role of DNA-stimulated PCD in pathology and host protection.
figure 4

An overview of our current understanding of the role of DNA-stimulated programmed cell death (PCD) in different immunological and pathological settings is shown. The figure summarizes how apoptosis, necroptosis and pyroptosis contribute to defence against infections and cancers, promotion of immunopathology and regulation of immune responses. For more details, see the main text. CDN, cyclic dinucleotide; HTLV-1, human T cell leukaemia virus type 1.

Full size image

Microbial infections

In response to infections, human T cell leukaemia virus type 1 induces apoptosis in human macrophages through STING-dependent triggering of IRF3–BAX complexes73. Given the reported role for myeloid cells in spread of virus to CD4+ T cells99, the depletion of the infected monocytes and myeloid precursor dendritic cells (DCs) via DNA-dependent PCD may function to prevent the spread of the virus and hence limit infection. In a study on Mycobacterium tuberculosis variant bovis, it was found that the bacterium induces apoptosis in murine macrophages in a manner dependent on both ER stress and STING–TBK1–IRF3 (ref.100). Importantly, inhibiting TBK1 blocked mitochondrial apoptosis and elevated bacterial replication, thus suggesting DNA induced PCD to block microbial spread. The cGAS–STING pathway was also demonstrated to mediate apoptosis-like cell death in human foreskin fibroblasts infected with HSV1, as measured by poly(ADP-ribose) polymerase (PARP) cleavage101. However, as lysosomal cell death also leads to PARP cleavage, the published data did not resolve whether this occurs through the STING–IRF3–BAX pathway or the STING-induced lysosomal cell death pathway. It will be interesting to learn whether the cGAS–STING-dependent stimulation of apoptosis-like PCD impacts on control of herpesvirus infections. The observation that the HSV1 ubiquitin E3 ligase ICP0 prevents cGAS–STING-dependent PARP cleavage101 suggests this to be the case.

Necrotic types of PCD are also induced during infections through DNA-stimulated pathways. During infections with many bacteria, pyroptosis is induced by AIM2 and/or NLRP3 (refs57,102,103,104,105). This includes infection with L. monocytogenes, L. pneumophila, F. tularensis, Streptococcus pneumoniae and Brucella abortus. There is also evidence for activation of DNA-stimulated pyroptosis by viruses106,107,108 and intracellular parasites109. Although it has been reported that pyroptosis can be a central player in antibacterial defence in vivo110, the importance of DNA-triggered pyroptosis in antimicrobial defence is still not fully established. To this end, it is currently difficult to discern the relevance of inflammasome-dependent cytokine maturation versus the induction of pyroptosis. Regarding viruses, one report has shown an antiviral role of AIM2-stimulated pyroptosis during infection with enterovirus A71 in a neuronal-like cell line106. Many bacteria and viruses evade the DNA-stimulated inflammasomes, thus underscoring the importance of these pathways in antimicrobial defence. Examples of such infections include Francisella tularensis subsp. novicida, which uses a CRISPR–Cas system to maintain membrane integrity and limit AIM2 activation111, and the HSV1 VP22 protein that binds AIM2 and inhibits downstream signalling112. However, pyroptosis can also contribute to infection pathology. For instance, the observation that Casp1−/−Casp11−/− mice are protected against Escherichia coli sepsis whereas mice deficient in both IL-1β and IL-18 are not suggests that pyroptotic cell death amplifies the inflammatory response in a pathological manner113. A second example is provided by HIV, which infects CD4+ T cells and eventually causes T cell depletion and immunodeficiency. Greene and co-workers demonstrated that the AIM2 family member IFI16 senses DNA in cells abortively infected with HIV, thus inducing pyroptosis and potentially contributing to T cell depletion, chronic immune activation and disease pathogenesis107.

Regarding the role of DNA-driven necroptosis during infections, there is only limited information available. However, as outlined above, it has been reported that hyperactivation of STING in mice leads to a shock-like syndrome, which is dependent on TNF and type I interferon, both of which are also required for STING-dependent necroptosis89. These data suggest that STING signalling might also engage necroptosis under physiological conditions, such as infections with herpesviruses, which potently activate cGAS–STING signalling43,114. Importantly, murine gammaherpesvirus MHV68, which is closely related to Epstein–Barr virus and Kaposi sarcoma-associated herpesvirus, induces necroptosis in a murine fibroblast cell line in a manner dependent on STING and TNF70. Because several DNA viruses that are known to trigger the cGAS–STING pathway also encode proteins that can inhibit caspase 8, for example, vaccinia virus115, it is likely that DNA-driven necroptosis is of importance for control of virus infections. Indeed, Ripk3−/− mice are more susceptible to vaccinia virus infection than wild-type mice116.

Sterile inflammatory diseases

Given the potentially pro-inflammatory role of necrotic types of cell death, it would not be surprising if programmed DNA-stimulated cell pathways also contribute to the pathogenesis of sterile inflammatory diseases. This is indicated by the finding that the STING–IRF3–BAX apoptosis pathway is activated in alcoholic liver disease and that Irf3−/− mice but not Ifnar−/− mice are protected from disease117. In many inflammatory diseases where DNA-stimulated pathways have been reported to contribute to the pathogenesis, there is extensive cell death. Examples include systemic lupus erythematosus, Sjögren’s syndrome and STING-associated vasculopathy with onset in infancy (SAVI)85,118,119. However, at present, there are limited data directly demonstrating that the cell death occurring in these pathologies is induced by DNA. Currently, the data available are indirect. For instance, STING agonists induce apoptosis in endothelial cells and disrupt vasculature78, which is seen in the above-mentioned diseases. These diseases are also commonly referred to as type I interferonopathies owing to the high level of type I interferons associated with them. Interestingly, DNA-induced interferon production is subject to negative regulation by AIM2-dependent pyroptosis11,120. Therefore, lack of full AIM2 activity leads to augmented DNA-stimulated type I interferon responses owing to prolonged activation of the cGAS–STING pathway. Two questions emerging from the discussion above are whether type I interferon is in fact the only main driver of the interferonopathies and whether PCD may play a role. In favour of interferon-independent mechanisms also contributing to the pathogenesis, for at least some interferonopathies, is the finding that mice harbouring the SAVI-mimicking mutation N153S are not protected from disease in the absence of the central interferon-inducing transcription factor IRF3 (ref.121). This is, however, in contrast to Trex1−/−Irf3−/− mice, which are protected from cGAS–STING-dependent disease5,51,122. Because accelerated cell death is observed in SAVI85, it would be interesting to test whether STING-N153S mice with selective defects in STING-dependent death pathways are protected against disease.

Despite the lack of solid evidence for DNA-driven cell death in human inflammatory diseases, the concept of DNA-stimulated cell death as a key driver in sterile inflammation has been demonstrated. Flavell and colleagues reported that radiation-induced tissue damage was dependent on DNA-induced cell death123. Importantly, Aim2−/− mice were protected against both subtotal body irradiation-induced gastrointestinal syndrome and total body irradiation-induced haematopoietic failure, and the effect of AIM2 was largely mediated by pyroptotic cell death. As another example of DNA-driven cell death in acute sterile inflammation, it was reported that aortic aneurysm leads to STING-dependent activation of TBK1, RIPK3 and phosphorylated MLKL in smooth muscle cells. STING deficiency or TBK1 inhibition blocked these responses and improved the disease outcome in mice. These data suggest a role for STING-dependent necroptosis in aortic aneurysm124. Similar observations have been made in a mouse model for inflammatory bowel disease, in which STING was found to drive type I interferon and TNF-mediated necroptosis of intestinal epithelial cells125.

Cancer

Cancer is characterized by uncontrolled cell proliferation but also excessive cell turnover. This leads to accumulation of host DNA in the cytoplasm, which can stimulate beneficial and pathological signalling37. In the process of proliferation of cancer cells, micronuclei are frequently formed in the cytoplasm, and these can be detected by cGAS126,127. This could suggest a role for the cGAS–STING pathway and maybe cytosolic DNA sensing in general in tumour immunity. In line with this, many cancer cells have lost STING expression, and clear effects of STING gene deletion or STING agonists on cancer development have been reported54,55,128,129,130. On the other hand, cytosolic DNA may also drive innate immune responses that promote cancer progression37,131.

The ability of the STING pathway to induce apoptosis or apoptosis-like cell death can directly target STING-expressing tumour cells for death following accumulation of DNA in the cytoplasm. In mouse models for T cell leukaemia and multiple myeloma, treatment with STING agonists induces apoptosis in cancer cells, and this promotes tumour control and improves disease outcome76,132. These data reveal a therapeutic potential for CDNs as direct tumoricidal agents against STING-expressing tumour cells. It will be interesting to learn whether cell-autonomous apoptosis induced by endogenous DNA derived from tumour cells is involved in tumour immunosurveillance. On the other hand, as the T cell response is a major player in tumour immunity, apoptosis of lymphocytes may also hamper the antitumour immune response. To this end, it has been reported that STING agonists induce tolerogenic effects on T cell responses, and so there is a need to understand this phenomenon in more detail133.

Many of the current cancer therapies target cancer cells through either irradiation or inhibition of DNA replication. As discussed above, irradiation induces AIM2 activation and pyroptotic cell death123. In addition, chemotherapeutic agents such as cisplatin and etoposide inhibit DNA replication37. Thus, the mechanisms through which conventional anticancer therapies kill cancer cells may involve DNA-activated PCD pathways. Therefore, altered expression of proteins in the DNA-sensing death pathway by cancer cells may not only affect tumour growth but could also impact on the efficiency of therapy.

Mechanistic understanding of DNA-stimulated cell death pathways and their role in human pathologies may allow development of new treatments. There is currently a large interest in development and therapeutic use of agonists and antagonists for the cGAS–STING pathway, and substantial progress has already been made134,135,136,137. The field is thus in a favourable position to rapidly take advantage of such new discoveries to potentially target DNA-triggered PCD in a number of diseases.

Concluding remarks and outstanding questions

In recent years, it has become clear that cells can succumb to cell death through a broad range of programmed mechanisms. At the same time, it has become apparent that these pathways constitute a central part of the immune response to infections and the restoration of homeostasis under sterile conditions. Accumulation of cytoplasmic DNA, which occurs during infections, sterile tissue damage, autoimmune diseases and cancer, is a very potent danger signal that has mainly been reported to induce the production of cytokines such as type I interferons and IL-1β2. However, as outlined in this Review, cytoplasmic DNA also induces several PCD pathways, which we are now starting to understand at the molecular level. Furthermore, there is accumulating evidence for these pathways being of relevance for the beneficial and pathological biological activities of cytoplasmic DNA. With the recent advances in this field, a number of outstanding questions have emerged. For instance, what determines whether DNA stimulates type I interferons, IL-1β, apoptosis, autophagy, necroptosis, lysosomal cell death and/or pyroptosis? Possible influencing factors include the strength of signal induced by DNA recognition, the duration of signalling, the cell type sensing DNA, the length and format of the DNA, whether there is microbial modulation of specific DNA-sensing pathways, the metabolic state of the host and the local cytokine environment. Also, we still do not understand many of the molecular details of the pathways for DNA-stimulated PCD. In addition, much more work and better tools are required to fully characterize the roles of the different DNA-stimulated PCD pathways in diseases.

On a final note, the discovery that cytosolic DNA can induce multiple types of PCD has revealed the ironic phenomenon that the immune system uses ‘the molecule of life’ to induce death in many different ways.