INTRODUCTION

      DNA, in its function as the primary medium for storage of biological information, resides in specialized compartments, separate from the rest of the eukaryotic cell. This separation, in the nucleus or in endosymbiont organelles (i.e., mitochondria and chloroplasts), allows for specialized and compartmentalized control of major DNA functions and cellular processes, including gene expression, genome replication, and repair of damaged DNA. Extending this, cellular responses have evolved to sense, destroy, and signal the presence of foreign DNA, in part by their localization to cellular compartments normally free of DNA. In metazoan cells, these cellular responses are part of an organism’s innate immunity, which can be activated as an alarm signal by many stimuli (e.g., viral and bacterial infections) (Brubaker et al., 2015).

      Similar pathways can also be activated by mislocalized endogenous cytoplasmic DNA (cytoDNA) (see Box 1), which can drive cell-autonomous responses in the absence of infection. Endogenous cytoDNA is a contributor to so-called sterile inflammation, inflammation in the absence of pathogenic infection, which is associated with many chronic age-associated conditions, including cancer, cardiovascular disease, and neurodegenerative disorders (Ferrucci and Fabbri, 2018; Franceschi et al., 2007; Furman et al., 2019). Altered immune function, including sterile inflammation, is considered a hallmark of the aging process, sometimes referred to as ‘‘inflamm-aging’’ (Franceschi et al., 2007; Lo´pez-Otı´n et al., 2013).

       Although the interplay between viral DNA and host defenses is well-studied, the role of cytoDNA in aging and chronic diseases has only recently been appreciated. In this review, we focus on four  species of endogenous cytoDNA: micronuclei, cytoplasmic chromatin fragments, mitochondrial DNA, and retrotransposons.We evaluate the major species of endogenous cytoDNA, illustrating the mechanisms involved in cytoDNA formation, functions, and pathologies.

Micronuclei

      Micronuclei (MN) are nuclear membrane-enclosed whole chromosomes or chromosome fragments in the cytoplasm, separate from the nucleus, either directly shown or often presumed to result from mitotic defects.

Formation of MN

       MN formation occurs in the context of cell division and is associated with two related mechanisms of chromosome missegregation, dependent on: (1) mitotic defects including defective metaphase centromere-kinetochore assembly or defects in anaphase, and (2) DNA misrepair (Fenech et al., 2011; Ye et al., 2019) (Figure 1A). Although mitotic defects are associated with mis-segregation of whole, ‘‘lagging’’ chromosomes in anaphase, defective DNA repair is associated with mis-segregation of chromosome fragments. These fragments can be generated through unrepaired DNA double strand breaks (DSBs) or misrepair of DSBs and telomere ends leading to fusions between chromosome fragments. The former can cause mis-segregation of a chromosome fragment lacking a centromere. The latter can lead to the formation of chromosome bridges (McClintock, 1939; Umbreit et al., 2020), where a fused chromosome with two centromeres can either break during cytokinesis or persist and form a bridge between two newly forming nuclei after anaphase (Liu et al., 2014; Maciejowski et al., 2015). Breakage of this bridge leads to extensive rearrangement of the bridge region by non-homologous end joining (Ly et al., 2017) and formation of chromosome fragments lacking centromeres or with other centromeric defects. These abnormal chromosomes can then generate MN in the next round of cell division (Umbreit et al., 2020). In each case, the mis-segregated nuclear DNA forms a separate nuclear envelope and becomes a MN, with consequences in cell function and fate.

Function of MN

      MN are initially surrounded by a nuclear envelope that compartmentalizes MN away from the cytoplasm; however, the integrity of this nuclear envelope can be spontaneously lost, leading to rupture (Hatch et al., 2013) (Figure 1B). MN have defective nuclear envelope assembly, characterized by deficiency of nuclear pore complexes and attenuated nuclear transport (Liu et al., 2018), which affects many processes in MN, including defective replication and repair of MN DNA (Crasta et al., 2012; Hatch et al., 2013; Umbreit et al., 2020; Zhang et al., 2015). Loss of MN nuclear envelope integrity is associated with MN rupture and sensing of MN DNA by cytoDNA sensors.

      Although many DNA sensors have been described (Dempsey and Bowie, 2015), the best-studied sensor and paradigm for sensing of cytoDNA is cyclic GMP-AMP synthase (cGAS). Upon DNA recognition, activated cGAS generates the second messenger cyclic-GMP-AMP (cGAMP) from ATP and GTP, which binds to and activates stimulator of IFN genes (STING) (Gao et al., 2013a; Li et al., 2013b; Sun et al., 2013; Wu et al., 2013). cGAMP-induced STING oligomerization provides a signaling platform for recruiting and activating TANK-binding kinase (TBK1) (Zhang et al., 2019). STING also functions as a scaffold for IRF3 phosphorylation by TBK1 and subsequent dimerization of IRF3 (Tanaka and Chen, 2012). Dimerized IRF3 enters the nucleus and triggers the type-I interferon response. TBK1 associated with STING also facilitates dsDNA-mediated activation of nuclear factor kB (NF-kB) to promote proinflammatory gene transcription (Abe and Barber, 2014; Ishikawa and Barber, 2008). Knockout studies in mice have demonstrated that cGAS is required for DNA ligand and viral-induced interferon responses, suggesting that cGAS is the predominant cytoplasmic DNA sensor in these contexts (Gao et al., 2013a; Lahaye et al., 2013; Li et al., 2013b; Schoggins et al., 2014). Similarly, cGAS activation due to rupture of the MN nuclear envelope is thought to be the major trigger of an interferon-based inflammatory response (Bartsch et al., 2017; Gekara, 2017; Harding et al., 2017; Mackenzie et al., 2017).

      However, the exact relationship between MN rupture, DNA damage, and cGAS sensing remains unclear. Given that a proportion of cGAS ordinarily localizes to nucleosomes and activation of cGAS by DNA/chromatin binding is highly regulated (Perspective and Conclusion), MN rupture alone is not sufficient for cGAS activation (de Oliveira Mann and Hopfner, 2021). MN rupture, similar to nuclear rupture, is associated with extensive DNA damage (Denais et al., 2016), perhaps due to exposure of the MN to cytoplasmic factors such as the exonuclease TREX1 (Maciejowski et al., 2020) or other nucleases (Umbreit et al., 2020), thus DNA damage might directly or indirectly facilitate cGAS activation. cGAS sensing of MN has context-dependent roles in inflammatory signaling and tumorigenesis. Mackenzie et al. (2017) speculate that cGAS recognition of micronuclear DNA may act as a cellintrinsic immune surveillance mechanism detecting a range of neoplasia-inducing processes (i.e., MN sensing is a tumor suppressor mechanism). Harding et al. (2017) also imply that MN sensing by cGAS/STING is tumor suppressive, perhaps predominantly early in disease progression. Accordingly, cancer cells often inactivate the cGAS-STING pathway to evade immune detection (Xia et al., 2016).

Dysfunction of MN

      Although MN have been proposed to activate as tumor suppressor pathways, MN also play a critical pathogenic role in the evolution of cancer genomes and metastasis by driving genomic instability through chromothripsis (Bakhoum et al., 2018; Umbreit et al., 2020; Ye et al., 2019; Zhang et al., 2015). Chromothripsis is the process of extensive chromosome fragmentation that is thought to occur, in part, through breakage-fusion-bridge cycles (Umbreit et al., 2020) and through accumulation of DNA damage in micronuclei (Crasta et al., 2012; Zhang et al., 2015). Nucleoplasmic bridges formed between daughter nuclei have similar defects in nuclear envelope assembly compared to MN, leading to defects in replication and repair and ultimately rupture (Umbreit et al., 2020). This rupture and subsequent cell division can initiate cycles of MN formation, damage, and reintegration into the genome by non-homologous end joining, further exacerbating chromothripsis and evolving genome instability (Umbreit et al., 2020). Moreover, STING-dependent activation of noncanonical NF-kB signaling, due to ruptured MN, promotes metastasis (Bakhoum et al., 2018).

Figure 1. Mechanisms of micronucleus formation

(A) Micronuclei form through defects in chromosome segregation, due to chromosome missegregation during mitosis or DNA misrepair causing mis-segregation of a chromosome fragment. Formation of abnormal chromosomes may directly lead to formation of a micronucleus or to formation of a nuclear bridge. Rupture and reassembly of nuclear bridges can form MN in the next round of cell division.

(B) Micronuclei are characterized by an unstable nuclear envelope, rupture of which leads to activation of cGAS.

Therapeutic targeting of MN

      MN activation of cGAS-STING signaling is a cell-intrinsic mechanism to surveil genome instability with cancer therapeutic implications. Therefore, activation of signaling pathways downstream of MN, especially STING, presents an attractive target in cancer immunotherapy (Su et al., 2019). However, caution is required because, in some contexts, activation of the cGAS-STING pathway or other candidate MN sensors is pro-tumorigenic. For example, in breast and lung cancers, suppression of chromosomal instability and associated cytoDNA decreases metastasis (Bakhoum et al., 2018). This is perhaps related to context-dependent utilization of the cGAS-STING downstream pathways in altering immune function, the involvement of other cytoDNA sensors, the role of the tissue environment in different contexts, or some combination of these factors. For example, STING activation can also be pro-tumorigenic in the context of chronic STING activation in tumor-adjacent senescent cells (see below and Jiang et al., 2020), suggesting that STING and cGAS pharmacological targeting must also consider other contexts of cytoDNA biology in non-tumor cells.

Cytoplasmic chromatin fragments

      Among the contributing factors to ageassociated sterile inflammation and disease is cellular senescence, which is a cellular response to severe stress primarily characterized by morphological and metabolic changes, permanent exit from the cell cycle, resistance to cell death, and development of an inflammatory senescence-associated secretory phenotype (SASP) (Acosta et al., 2008; Coppe et al., 2008; d’Adda di Fagagna, 2008; Gorgoulis et al., 2019; Krtolica et al., 2001; Kuilman et al., 2008; McHugh and Gil, 2018). Senescent cells have chromatinized micronuclei-like structures in the cytoplasm (Suzuki et al., 2002); however, unlike true micronuclei, many of these observed structures are formed independently of cell cycle progression, by apparent blebbing of chromatin into the cytoplasm from a seemingly intact nucleus, and contain distinct molecular markers (Table 1), suggesting a different mechanism of formation of these cytoplasmic chromatin fragments (CCF) (Ivanov et al., 2013).

Formation of CCF

      CCF are characterized by a suite of chromatin modifications, including the presence of heterochromatin-associated H3K27me3 and relative absence of euchromatin-associated H3K9Ac, suggesting they are formed from heterochromatin, although their genome locus of origin remains unclear. Formation of CCF is associated with loss of nuclear membrane integrity in senescent cells (Ivanov et al., 2013). Nuclear membrane integrity is, in part, dependent on the nuclear lamina, a meshlike network of lamin A, B1, B2, and C intermediate filament proteins, which provides support to the nuclear membrane (Prokocimer et al., 2009). Loss of lamin B1 is a well-known feature of cell senescence (Freund et al., 2012; Shimi et al., 2011). CCF formation is also associated with autophagy, a collection of cellular homeostatic processes for trafficking and degradation of macromolecules and cellular components (Aman et al., 2021), and specifically autophagic degradation of lamin B1 in senescent cells. However, whether downregulation of lamin B1 is a cause or consequence of CCF formation remains unclear. Knockdown of lamin B1 is known to drive CCF-like foci formation (Hatch et al., 2013), whereas the overexpression of lamin B1 delays onset of senescence (Shimi et al., 2011), suggesting that lamin B1 loss may be upstream of CCF formation. However, CCF formation in senescent fibroblasts requires interaction of lamin B1 with autophagy substrate delivery protein LC3B in the cell nucleus, and this is followed by autophagic degradation of lamin B1-containing CCF in the cytoplasm (Dou et al., 2015), suggesting that, at least in part, CCF are a vehicle for degradation of lamin B1. In addition, there is evidence that lamin B1 mRNA levels are also decreased in senescent cells (Freund et al., 2012), suggesting that multiple mechanisms may regulate expression of lamin B1 in senescence. Clarification of the dynamics of lamin B1 expression and degradation in senescent cells can provide important insights into the relationship between CCF formation and autophagy.

      CCF are strongly positive for the DNA damage marker gH2AX but lack the typically colocalized DNA repair protein 53BP1, which acts as a negative regulator of CCF formation (Ivanov et al., 2013; Vizioli et al., 2020). The observation that CCF are strongly positive for gH2AX suggests that DNA damage plays a role in CCF formation. DNA damage, particularly the formation of DSBs, is a potent inducer of cellular senescence, in part through a positive feedback circuit linking altered mitochondrial function in senescent cells to persistent DNA damage via nucleus-mitochondria (anterograde) and mitochondria-nucleus (retrograde) pathways (Passos et al., 2010). Compared to normal proliferating or quiescent cells, senescent cells have increased mitochondrial mass, decreased mitochondrial membrane potential, potentially altered mitophagy, and increased levels of mitochondrial reactive oxygen species (Chapman et al., 2019). Interestingly, mitochondrial ablation from senescent cells, or inhibition of mtROS formation, inhibits the SASP and prevents the formation of CCF, suggesting that mito-nuclear communication plays a role in CCF formation (Correia-Melo et al., 2016; Vizioli et al., 2020). Mechanistically, mtROS has been reported to activate the stress-activated kinases JNK1/2, which interacts with the CCF suppressor 53BP1. 53BP1 is a DNA damage-associated scaffold protein that recruits DNA repair factors to DNA DSBs and inhibits the resection of DSB ends (Mirman and de Lange, 2020), suggesting that DNA DSB end resection is required for CCF formation. Consistent with this hypothesis, inhibition of MRE11 with the small molecule mirin prevents CCF formation (Vizioli et al., 2020) (Figure 2). Similar pathways have recently been described in human brain and Drosophila models of genome instability (Aguado et al., 2021; Joy et al., 2021).

Function of CCF

      Once CCF enter the cytoplasm, they are sensed by the cGASSTING pathway (Dou et al., 2017; Glu¨ck et al., 2017; Takahashi et al., 2018; Yang et al., 2017) (Figure 2). However, whether the DNA in CCF is readily accessible to cGAS and by default is a good substrate for cGAS activation remains to be resolved. cGAS-STING activation leads to NF-kB activation of SASP genes downstream of STING instead of an interferon gene program, potentially due to p38-mitogen-activated protein kinase (MAPK) repression of interferon genes in senescent cells (Dou et al., 2017). Under conditions of acute stress, senescent cells act as a potent tumor suppressor mechanism by arresting cell growth and promoting clearance of potentially malignant cells through SASP-dependent activation of the immune system. The observation that lamin B1 in CCF is degraded by autophagy (Dou et al., 2015) suggests that CCF may function as a reservoir for nuclear autophagy substrates. Other substrates of nuclear autophagy have been identified in senescent cells (Xu et al., 2020); however, the direct role of CCF in this process is unclear. Indeed, CCF contain many chromatin-associated proteins (Zhao et al., 2020a), but their fate in the cytoplasm, for example autophagic degradation, cGAS-independent signaling, or even secretion from the cell (Takahashi et al., 2017) is poorly understood. Lastly, the association gH2AX and CCF suggest that CCF form as by-products of DNA repair processes. Most DNA repair factors are downregulated in senescent cells resulting in decreased repair activity, especially homologous repair pathway activity (Mao et al., 2012; Seluanov et al., 2004). This suggests that expulsion and degradation of failed, potentially unstable, repair intermediates such as CCF may be a mechanism to preserve genome integrity and/or suppress excessive DNA damage response (DDR) signaling.

Therapeutic considerations for CCF

      In addition to the functions of CCF and SASP noted above, these signals can also have harmful effects on human health. Under some conditions, including old age, failed clearance of senescent cells leads to their accumulation and chronic SASP inflammation (Ovadya et al., 2018). Cumulative evidence supports that accumulation of senescence and a chronic SASP are major contributors to aging and age-associated disease. Accordingly, removal of senescent cells by genetic or pharmacological ‘‘senolytic’’ approaches extends median lifespan in old mice and prevents the onset of age-associated disease (Baker et al., 2011, 2016; Xu et al., 2018; Zhu et al., 2015). However, chronic administration of at least some senolytics in humans can be associated with toxic side effects (Kaefer et al., 2014; Wilson et al., 2010) that may be due to off-target killing of non-senescent cells. As an alternative approach, reduction of the SASP through ‘‘senostatic’’ or ‘‘senomorphic’’ approaches may be less toxic and has been shown to be effective in animal models (Birch and Passos, 2017; Georgilis et al., 2018; Laberge et al., 2012, 2015; Niedernhofer and Robbins, 2018; Xu et al., 2015; Yoshimoto et al., 2013).

      Targeting CCF formation through JNK signaling, or perhaps downstream signaling through use of cGAS-STING pathway inhibitors, can provide additional targets for development of senomorphic drugs. It has recently been shown that histone deacetylase inhibitors (HDACis) including vorinostat, an HDACi approved for use in humans to treat some cancers, prevent CCF formation and SASP indirectly through activation of mitochondrial function in senescent cells and in mouse liver, suggesting that this pathway is therapeutically targetable in vivo (Vizioli et al., 2020). Further research into the mechanisms involved can identify additional targets (e.g., JNK1/2 and MRE11) to reduce SASP, chronic inflammation, and associated chronic disease in humans. Given the role of senescent cells in promoting age-associated disease, it will be important to understand the interaction between mitochondrial function, autophagy and cytoDNA species in senescence and the therapeutic potential of modulating these interactions (see Perspective and Conclusions).

Other forms of DNA damage-associated cytoDNA

      In addition to MN formation in dividing cells and CCF formation in senescent cells, several other cytoDNA species have been labeled as MN or CCF, but with unclear mechanisms of formation. These include cytoDNA in human patient-derived cells and mouse models of ataxia-telangiectasia (Song et al., 2019), Dnase2a deficiency (Lan et al., 2014, 2019), telomere damage crisis-induced CCF-like structures (Nassour et al., 2019), HGPS nuclear abnormalities (Kreienkamp et al., 2018), Exo1 hyperexcision-driven chromosomal instability (Guan et al., 2021), and small DNA-damage-associated cytoDNA fragments (Jiang et al., 2019; Wang et al., 2021). In each case, formation of these cytoDNA is directly or indirectly associated with the DNA damage marker gH2AX; however, it is unknown how these mechanisms of formation are related to the specific mechanisms of MN and CCF formation described above. Despite these apparent differences, these DNA damage-associated cytoDNA species have been connected to cGAS-STING pathway activation, with the exception of cGAS-STING-independent KU complex cytoDNA sensing in aged T cells (Wang et al., 2021), and likely play important roles in physiology and disease.

Mitochondrial DNA

      Mitochondrial DNA (mtDNA) is present at thousands of copies per cell and is densely packaged into nucleoids (Spelbrink, 2010), which consist of one copy of mtDNA and different proteins, such as mitochondrial transcription factor A (TFAM), a protein responsible for nucleoid architecture, abundance, and segregation (Kanki et al., 2004; Kasashima et al., 2011). Typically located in the inner mitochondrial matrix, mtDNA encodes for 37 genes: 13 mRNAs that are translated into subunits of components of the oxidative phosphorylation system, 2 ribosomal RNAs (rRNA), and 22 transfer RNAs (tRNAs) (Taanman, 1999).

Mechanisms of mtDNA cytosolic release

      The presence of mtDNA in the cytosol has been associated with the process of apoptosis. Upon receiving apoptotic signals, proapoptotic proteins BAX and BAK become activated and form macropores in the outer mitochondrial membrane, inducing mitochondrial outer-membrane permeabilization (MOMP) (Tait and Green, 2013) (Figure 3). Following MOMP, BAX and BAK pores gradually widen, allowing the extrusion of the inner mitochondrial membrane into the cytosol, where it becomes permeabilized, facilitating mtDNA release (McArthur et al., 2018; Riley et al., 2018). MOMP was originally thought to be an all-or-nothing event, spreading to all mitochondria in a cell, whereupon proapoptotic proteins, such as cytochrome c are released from the intermembrane space, ultimately leading to cell death. However, it has been shown that in conditions of sub-lethal stress, a small subset of mitochondria undergo MOMP without inducing cell death, a phenomenon called minority MOMP (miMOMP) (Ichim et al., 2015). Therefore, mtDNA can also be present in the cytosol of cells that do not undergo apoptosis. Other mitochondrial stresses can also contribute to cytosolic mitochondrial DNA leakage. For example, reduced TFAM expression, which results in aberrant mtDNA packaging, organization, and distribution, leads to release of mtDNA into the cytosol (West et al., 2015).

      Aside from miMOMP and TFAM-deficiency, another mechanism that could mediate mtDNA release into the cytosol is through opening of the mitochondrial permeability transition pore (mPTP) (Patrushev et al., 2004, 2006) (Figure 3). The mPTP is a transmembrane protein located in the mitochondrial inner membrane and normally opens in response to calcium accumulation in the mitochondrial matrix, oxidative stress, and other stressors (Briston et al., 2017). Although it has been shown that the mPTP does not play a role in mtDNA efflux during apoptosis (Riley et al., 2018), one study has shown that fragments of mtDNA are released from mitochondria in brains of irradiated mice through transient opening of the mPTP (Patrushev et al., 2006). Another study revealed that inducing oxidative stress in mitochondria isolated from rat hepatocytes caused release of mtDNA fragments that were partly mediated by mPTP opening (Garcı´a et al., 2005). The aforementioned studies showing that only fragments of mtDNA are released through the mPTP are in line with the fact that the mitochondrial permeability transition pore only allows the transport of molecules smaller than 1.5 kDa, which is smaller than a mtDNA nucleoid (Halestrap et al., 2002). However, it is also possible that constitutive opening of the pore might cause mitochondrial swelling, leading to disruption of the inner membrane and consequently mtDNA cytosolic release, although further studies are needed to determine this (Riley and Tait, 2020). Interestingly, a recent study has proposed a role for voltage-dependent anion channel (VDAC) oligomerization in mediating mtDNA release into the cytoplasm (Kim et al., 2019). It was shown that mouse fibroblasts deficient in the mitochondrial endonuclease g, which display increased oxidative stress, had higher levels of cytosolic mtDNA and this could be attenuated by knocking down VDAC 1 and 3 (Kim et al., 2019).

      Changes in mitochondrial network dynamics can also contribute to release of mtDNA into the cytosol. TFAM-deficient fibroblasts, which have increased cytosolic mtDNA, display elongated and hyperfused mitochondria (West et al., 2015). Mitochondria fission is important to ensure proper nucleoid distribution and removal of damaged mtDNA. Consistent with this, the authors showed that increased mitochondria fission, due to knockdown of mitofusin 1 (Mfn1) in TFAM-depleted cells, was associated with reduced expression of interferon-stimulated genes (ISGs), suggesting that disruptions in the mitochondria network also influence mtDNA-induced expression of interferon genes (West et al., 2015). In line with this, a recent study demonstrated that knockdown of MxB, a GTPase located in the inner mitochondrial membrane, leads to mitochondria fragmentation, disruption of the inner mitochondrial membrane, and increased cytosolic mtDNA (Cao et al., 2020), supporting the notion that imbalance in mitochondria network homeostasis is a contributor to mtDNA release into the cytoplasm.

Function of cytosolic mtDNA

      MtDNA shares many features with bacterial DNA. For instance, it has been shown that mtDNA is hypomethylated compared to nuclear DNA (Hong et al., 2013; Mechta et al., 2017). Unmethylated CpG motifs in bacterial DNA are potent triggers of inflammation (Klinman et al., 1997; Krieg et al., 1995). Interestingly, the similarities between mtDNA and bacterial DNA at the structural level likely reflect the bacterial origin of mtDNA (Roger et al., 2017) and even now appear to be a basis for shared roles in immune signaling. MtDNA is recognized as foreign when found outside of mitochondria and can trigger inflammation (Maekawa et al., 2019; Shimada et al., 2012; Zhang et al., 2010). For example, in TFAM-depleted cells, the presence of mtDNA in the cytosol was associated with increased expression of ISGs in a cGAS STING-IRF3-dependent manner (West et al., 2015) (Figure 3). Moreover, cytoplasmic mtDNA that accumulates in response to mitochondrial endonuclease g deficiency has also been shown to induce expression of ISGs in mouse fibroblasts (Kim et al., 2019), suggesting that cytosolic mtDNA is a major trigger of inflammatory pathways.

      Cytosolic mtDNA has also been shown to play a role in response to viral infections. For example, RNA viruses can activate the cGAS-STING pathway despite the specificity of cGAS for cytoplasmic DNA over RNA (Gao et al., 2013b; Schoggins et al., 2014). One study demonstrated that infecting human cells with the dengue virus (DENV-2), which has an RNA genome, resulted in increased mtDNA in the cytoplasm, which coincided with cGAS activation, thus providing a mechanism through which RNA viruses activate innate immune signaling (Sun et al., 2017). Consistent with a role for mtDNA in priming the anti-viral immune response, TFAM-deficient cells display heightened expression of type I interferon and ISG upon infection with herpes simplex virus 1 (HSV-1) and are more resistant to infection compared to wild-type cells (West et al., 2015). Anti-viral priming and resistance to viral infection could be ablated by treating TFAM-deficient fibroblasts with dideoxycytidine (ddC), which inhibits mtDNA replication and reduces mtDNA stress in these cells, suggesting that cytosolic mtDNA potentiates antiviral innate immunity (West et al., 2015). Interestingly, another recent study showed that microbial infection induced miMOMP and BAX/BAK-dependent mtDNA cytosolic release, leading to cGAS-STING activation and secretion of pro-inflammatory cytokines (Brokatzky et al., 2019). Blocking sub-lethal apoptotic signaling impaired the ability of cells to control pathogenic infection, implicating BAX/BAK-mediated mtDNA cytosolic release as a mechanism for efficient immune defense (Brokatzky et al., 2019).

      As well as triggering inflammation via recognition by cGASSTING, cytosolic mtDNA can activate other sensors within the cell that culminate in downstream inflammation. For example, mtDNA has been shown to activate the NLRP3 inflammasome. One study showed that treating cells that have dysfunctional mitochondria with lipopolysaccharide (LPS) or ATP resulted in increased levels of cytosolic mtDNA, which contributed to secretion of interleukin (IL)-1b and IL-18 via the NLRP3 inflammasome (Nakahira et al., 2011). The authors also suggested that as well as acting as a downstream sensor of cytoplasmic mtDNA, NLRP3 also contributes to mtDNA release by facilitating mPTP formation in the mitochondria inner membrane (Nakahira et al., 2011). Another study showed that during macrophage apoptosis, oxidized mtDNA is released into the cytosol, where it binds to and activates the NLRP3 inflammasome (Shimada et al., 2012). Furthermore, a more recent study also demonstrated that priming of macrophages generates newly synthesized, oxidized mtDNA fragments that activate the NLRP3 inflammasome, suggesting that NLRP3 preferably binds oxidized mtDNA (Zhong et al., 2018).

Therapeutic considerations of cytosolic mtDNA

      The presence of mtDNA circulating in plasma has also been reported in several specific age-associated diseases. For example, mtDNA was shown to be present at significantly higher levels in plasma and synovial fluid from patients with rheumatoid arthritis, a condition characterized by chronic inflammation and that is more prevalent in older age (Hajizadeh et al., 2003). Patients with type 2 diabetes mellitus also display increased levels of the pro-inflammatory factor tumor necrosis factor alpha (TNF-a) and mtDNA in the circulation, and the latter correlates with insulin resistance (Yuzefovych et al., 2019). In addition, circulating mtDNA has been identified in patients with ovarian (Zachariah et al., 2008), testicular (Ellinger et al., 2009), prostate (Ellinger et al., 2009), and lung cancer (Hou et al., 2013). Elevated concentration of mtDNA was also found in plasma and bronchoalveolar lavage from patients with idiopathic pulmonary fibrosis (IPF) (Ryu et al., 2017). The level of plasma mtDNA was associated with disease progression and was shown to predict mortality (Ryu et al., 2017). Moreover, normal human lung fibroblasts treated with isolated mtDNA expressed high levels of a-smooth muscle actin, which is a characteristic of fibrotic disease, suggesting that mtDNA plays a role in the pathophysiology of IPF (Ryu et al., 2017).

      Although it is not known whether circulating mtDNA is causal in age-associated diseases, given its strong pro-inflammatory potential, it is likely that cytoplasmic and extracellular mtDNA contribute to some extent to disease development. Therefore, mtDNA and the mechanisms facilitating its release might represent a potential therapeutic target to improve or prevent ageassociated pathologies. For example, it has been shown that Bax expression is increased in alveolar epithelial cells of patients with IPF (Plataki et al., 2005), suggesting an involvement of MOMP in the etiology of the disease. Consistent with that, treatment with BAX-inhibiting peptide V5 (BIP-V5), which inhibits BAX activation in the cytosol, reduced inflammation and improved lung pathology and survival in a mouse model of IPF (Suzuki et al., 2017). Furthermore, deletion of Bax has been shown to extend ovarian function in aged female mice, and also resulted in significant improvements in a number of age-related phenotypes, such as increased bone mineral density and strength, better retention of lean body mass, and decreased anxiety (Perez et al., 2007). Another study demonstrated that inhibition of BAX by two small molecule BAX channel inhibitors, Bci1 and Bci2, confers a protective role during brain ischemia in gerbils (Hetz et al., 2005). Administration of BAX inhibitors following reperfusion effectively inhibited MOMP, shown by reduced cytosolic cytochrome c release in neurons, and significantly decreased hippocampal damage (Hetz et al., 2005). These studies suggest that inhibition of MOMP by targeting BAX might represent a valuable strategy to improve age-related conditions and extend health span. Furthermore, inhibiting VDAC oligomerization by using the inhibitor VBIT-4 has been shown to decrease mtDNA release and IFN signaling in a mouse model of systemic lupus erythematosus and ameliorated the disease severity (Kim et al., 2019), suggesting that inhibition of VDAC oligomerization and perhaps mPTP might represent an effective therapeutic strategy to improve diseases involving cytosolic mtDNA release.

      Components of the SASP have been shown to act in a paracrine manner, inducing senescence in surrounding cells, and this is thought to be a mechanism by which senescent cells contribute to age-related tissue dysfunction (Acosta et al., 2013; Baker et al., 2016; Coppe´ et al., 2010; Nelson et al., 2012). IL-1b, a major SASP factor, can induce mtDNA release with subsequent activation of pro-inflammatory pathways (Aarreberg et al., 2019). One possibility is that IL-1b or other SASP components released by senescent cells can stimulate cytosolic mtDNA release in neighboring cells, potentiating inflammation. However, further studies are needed to confirm this hypothesis. A recent study has shown that senescent cells are a major source of cell-free mtDNA that accumulates with aging (Iske et al., 2020). Pharmacological elimination of senescent cells from old donor animals significantly reduced circulating mtDNA, improved age-associated inflammation and prolonged the survival of mice receiving organ transplant from old animals (Iske et al., 2020). Although the mechanism by which mtDNA is released to the extracellular space has not been fully elucidated, this study implicates senescent cells as important contributors to the increase in circulating mtDNA observed with aging. Additionally, it is possible that damaged mitochondria in senescent cells contribute to mtDNA release and age-related chronic inflammation, although this is a topic that warrants further investigation. Elimination of senescent cells has proven to be an effective strategy in improving a number of age-related diseases (Childs et al., 2016; Ogrodnik et al., 2017; Schafer et al., 2017; Xu et al., 2018). Although the role of mtDNA in this context remains to be definitively established, targeting senescent cells and/or mtDNA release from senescent cells is another candidate therapeutic avenue.

Retrotransposons

      Transposable elements (TEs) are mobile DNA sequences that were identified in 1950 by Barbara McClintock (McClintock, 1950). TEs make up
50% of the human genome and contribute significantly to genetic variation at the population level (Maxwell, 2016). Transposable elements are divided into two classes: class II TEs, also known as DNA transposons, encode the protein transposase, which mediates insertion and excision, allowing these transposons to translocate through the genome by a ‘‘cut and paste’’ mechanism (Goodier, 2016). On the other hand, class I TEs, or retrotransposons, do not encode transposase and thus depend on RNA intermediaries to transpose into other sites of the genome. (Goodier, 2016). Retrotransposons are the majority class of TEs present in the human genome and are themselves divided into two types: LTR retrotransposons, which comprise long terminal repeats (LTRs) on bothends, and non-LTRs that do not contain these repeats and are subdivided into long-interspersed nuclear elements (LINEs) and short-interspersed nuclear elements (SINEs) (Goodier, 2016). The latter includes the Alu elements, which are 300 bp in length and are present in more than one million copies, representing 11% of the human genome (Cardelli, 2018). In contrast, the LINE-1 (L1) family can reach up to 6 kb in length, and with more than 500,000 copies, is the most abundant human LINE, comprising 17% of the human genetic material (Cardelli, 2018). In addition, the L1 family includes the only endogenous autonomous retroelements that still retain the capacity to autonomously retrotranspose, and their protein products (ORF2p and ORF1p) are also responsible for mediating retrotransposition of SINE non-autonomous elements (Cardelli, 2018).

Retrotransposon activation and formation of cytosolic retrotransposon cDNA

       Upon activation, LINEs are transcribed by RNA polymerase II, and the resulting mRNA transcript is then exported into the cytoplasm, where it is translated into two proteins: ORF1p, an RNAbinding protein, and ORF2p, which acts both as an endonuclease and a reverse transcriptase (Figure 4). These two proteins form complexes with the mRNA transcript, and are then relocated back to the nucleus, where retrotransposition occurs by a process known as target-primed reverse transcription (TPRT) at the target site of reinsertion (Kazazian and Moran, 2017). During TPRT, it is proposed that the endonuclease encoded by L1 cleaves the bottom strand of the target DNA, which exposes a 30 -hydroxyl group, to serve as a primer for reverse transcription of the L1 mRNA by ORF2p. Following synthesis of a second strand of DNA, the process of reinsertion is completed by mechanisms that are still not fully understood (Kazazian and Moran, 2017). Although TPRT occurs in the nucleus, L1 cDNA has also been found in the cytosol in certain conditions, such as autoimmunity (Stetson et al., 2008; Thomas et al., 2017), and others that will be discussed in more detail below. There are two possible mechanisms by which L1 cDNA might exist in the cytoplasm: one possibility is that accumulation of high amounts of L1 mRNA and L1 proteins in the cytosol might enable reverse transcription to occur without a standard DNA template, resulting in increased cytosolic cDNA elements (Fukuda et al., 2021). Alternatively, it is also possible that during TPRT, the resulting cDNA does not re-integrate into the genome and instead leaves the nucleus into the cytosol by an unknown process (Thomas et al., 2017).

Figure 4. Mechanisms of retrotransposon activation leading to cytosolic DNA generation Upon transcriptional activation, polyadenylated (polyA) bicistronic LINE-1 (L1) mRNA encoding ORF1p and ORF2p is generated, exported to the cytoplasm, and translated. ORF1p and ORF2p bind to the mRNA forming a ribonucleoprotein (RNP) complex. Reverse transcription and L1 nuclear integration occurs via a mechanism known as target-primed reverse transcription (TPRT) that relies on annealing of poly(A) tail to thymidine nucleotides in nuclear DNA. Alternatively, reverse transcription potentially occurs in the cytoplasm. L1 cDNA accumulates in the cytoplasm in aging and disease, where it is sensed by nucleic acid

      DNA damage is associated with retrotransposon activation. Studies have shown that exposing murine and human cells to DNA damaging agents such as etoposide and UV- and g-irradiation results in increased expression of Alu RNA levels as well as increased retrotransposition (Hagan et al., 2003; Rudin and Thompson, 2001). Indeed, new SINE and LINE-1 insertions were detected in mouse germ cells following X-ray irradiation (Asakawa et al., 2004). Although the mechanisms by which DNA damage promotes retrotransposition are still unclear, it is possible that altered transcription factor expression following DNA damage induction facilitates transcription of these elements (Farkash and Luning Prak, 2006). Epigenetic changes that occur as a result of certain genotoxic stresses might also offer an additional mechanism of DNA damage-induced retrotransposition. In the case of L1, methylation of CpG sites at the 50 UTR hinders L1 promoter activity and prevents retrotransposition (Hata and Sakaki, 1997; Woodcock et al., 1997). In support of this, methyl-CpG-binding protein 2 (MeCP2) represses L1 expression and retrotransposition (Yu et al., 2001). Studies have shown that oxidative damage reduces the binding affinity of MeCP2 to damaged methylated DNA (Valinluck et al., 2004); therefore, it is possible that oxidative damage that occurs close to an L1 element might release its repression. Given that accumulation of DNA damage is a feature of both senescence and aging (d’Adda di Fagagna, 2008), it is possible that increased DNA lesions might contribute to reactivation of retrotransposons observed in senescent cells and aged tissues (De Cecco et al., 2013b, 2019; Simon et al., 2019).

      Mounting evidence indicates that extensive epigenetic changes that occur during senescence and aging account for the reactivation of retrotransposable elements. Indeed, a study demonstrated that the increase in L1 transcription observed during cellular senescence was mediated by decreased expression of RB1, a transcriptional repressor that promotes heterochromatinization of L1 elements, as well as increased expression of the transcriptional activator, FOXA1 (De Cecco et al., 2019). Another mechanism of age-related retrotransposon activation is reduced expression of SIRT6, a histone deacetylase that has been shown to regulate longevity (Kanfi et al., 2012). SIRT6 represses L1 activity by binding to the 50 UTR of L1 elements and promotes heterochromatinization by facilitating the interaction of KAP1 with the heterochromatin factor, HP1a, thus rendering L1 elements transcriptionally inactive (Van Meter et al., 2014). However, during senescence and in brains of aged mice, SIRT6 is depleted from L1 loci, allowing L1 expression and retrotransposition to occur (Van Meter et al., 2014). Consistent with this, SIRT6-deficient mice, which display accelerated aging phenotypes, have increased L1 expression in a number of tissues (Simon et al., 2019). Another sirtuin family member, SIRT7, is also responsible for the epigenetic regulation of LINE1 (Vazquez et al., 2019). L1s are preferentially located at lamin-associated domains, which are heterochromatic predominantly at the nuclear periphery (Vazquez et al., 2019). Importantly, SIRT7 mediates deacetylation of H3K18 in L1 elements, thus promoting their association with lamin proteins, maintaining them in a transcriptionally inactive state (Vazquez et al., 2019). One possibility is that during aging, SIRT7 is unable to anchor L1 elements to the nuclear lamina due to the loss of Lamin A/C, allowing the expression and retrotransposition of LINE-1 (Vazquez et al., 2019).

Functions of retrotransposons

      Suppression of retrotransposition is relieved during some physiological processes. For example, during preimplantation development when methylation patterns are erased, retrotransposon expression was shown to be relatively high in late oocytes and early-stage embryos, decreasing in later stages in mice (Peaston et al., 2004). Evidence for L1 retrotransposition during early embryogenesis has also been reported in humans, (Brouha et al., 2002; van den Hurk et al., 2007). In addition, retrotransposition can also occur during neuronal differentiation and influence cell fate (Muotri et al., 2005). Moreover, a number of stressors can induce retrotransposon expression and transposition, such as genotoxic stress, heat shock, viral infections, and heavy metals (Goodier and Kazazian, 2008). Retrotransposons are known to be activated in senescent cells (De Cecco et al., 2019; Gorbunova et al., 2021). In senescent cells, the chromatin of a number of retrotransposon families, such as Alu, SVA, and L1, becomes more relaxed, culminating in increased transcription and, ultimately, retrotransposition of these elements (De Cecco et al., 2013a). Whether or not there is benefit to activation of retrotransposons in all these contexts is not clear. However, it seems feasible that programmed activation of retrotransposons in senescent cells, and consequent activation of a cGAS-STINGdependent IFN-I response and pro-inflammatory phenotype (De Cecco et al., 2019), is of benefit for senescence-associated tumor suppression.

Dysfunctions of retrotransposons

      Unrestricted movement of L1s within the genome can be detrimental, with over 60 human genetic diseases being attributed to retrotransposon insertions (Goodier and Kazazian, 2008). Therefore, retrotransposon transcription is tightly regulated and silenced at multiple levels, including through epigenetic modifications, premature termination of transcription, micro-RNAs (miRNAs) and small-interfering RNAs (siRNAs), and post-transcriptional modifications (Goodier and Kazazian, 2008).

      Several studies have now provided evidence that increased expression of retrotransposable elements is associated with aging and disease (for review, see Gorbunova et al., 2021). For example, an age-related increase in the expression and retrotransposition of different retrotransposons has been shown in C. elegans, budding yeast, D. melanogaster, mice, and human cells (De Cecco et al., 2013a, 2013b; Dennis et al., 2012; Li et al., 2013a; Maxwell et al., 2011; Van Meter et al., 2014). In a model of chronological aging in S. cerevisiae, Ty1 mobility was increased in yeast populations with advanced age, and retrotransposition was associated with chromosome rearrangements (Maxwell et al., 2011). Moreover, expression of the Cer1 retrotransposon increases with age in C. elegans (Dennis et al., 2012), whereas in flies, an age-associated increase in mobilization of gypsy and non-LTR retrotransposons R1 and R2 was associated with loss of neuronal function and cognitive decline (Li et al., 2013a). In mammals, expression of retrotransposons was demonstrated to increase in liver and skeletal muscle of aged mice, resulting in retrotransposition in advanced age (De Cecco et al., 2013b). Interestingly, these events were attenuated by caloric restriction (White et al., 2014), an intervention that has been shown to extend lifespan in a plethora of organisms and decrease the burden of senescent cells (Fontana et al., 2018; Mercken et al., 2012). Retrotransposon activation has also been observed in mammalian cells lacking typical senescence markers, as well as in aged model organisms that do not display cellular senescence, suggesting that retrotransposon activation is an intrinsic and conserved feature of the aging process (Gorbunova et al., 2021).

      Activation of retrotransposons has also been shown to contribute to age-associated chronic inflammation (De Cecco et al., 2019; Simon et al., 2019). Reactivation of LINE-1 elements during late senescence is accompanied by accumulation of L1 cytoplasmic cDNA, due in part to a decrease in the levels of the exonuclease TREX1 (De Cecco et al., 2019), and activation of a cGAS-STING-dependent IFN-I response and pro-inflammatory phenotype (De Cecco et al., 2019). Consistent with this, SIRT6 knockout mouse fibroblasts display L1 activation, accumulation of cytoplasmic L1 cDNA, and increased expression of type I IFN genes (Simon et al., 2019). Immunoprecipitation analysis of these cells showed an increase in the abundance of L1 DNA bound to cGAS, corroborating that L1 DNA in the cytosol binds to cGAS and activates an IFN-I response (Simon et al., 2019). Inhibiting the synthesis of cytoplasmic L1 DNA, either by using nucleoside reverse transcriptase inhibitors (NRTIs) or by short hairpin RNA (shRNA) knockdown of L1, was sufficient to reduce the activation of IFN genes and the magnitude of the SASP, demonstrating that accumulation of L1 cDNA in the cytoplasm of senescent cells is an important contributor to the maintenance of the pro-inflammatory phenotype (De Cecco et al., 2019). Interestingly, increased cytosolic L1 DNA and expression of IFN genes has been observed in a number of tissues of aged mice (De Cecco et al., 2019; Simon et al., 2019).

      Independent of reverse transcription, the transcriptional intermediates derived from retrotransposons have also been associated with the development of age-related diseases. It has been shown that in a form of age-related macular degeneration, known as geographic atrophy (GA), the miRNA-processing enzyme, DICER1, is reduced in retinal pigmented epithelium (RPE) cells, causing an accumulation of Alu RNA (Kaneko et al., 2011). The latter then activates the NLRP3 inflammasome, stimulating cytokine production and MyD88-induced cytotoxicity of RPE cells, thus contributing to the etiology of GA (Tarallo et al., 2012). In addition, L1-derived mRNA was found to be highly expressed in the synovial fluid of rheumatoid arthritis (RA) patients, and the authors alluded to the possibility that L1 activity might play a role in the invasive phenotype of RA (Neidhart et al., 2000).

Therapeutic targeting of retrotransposons

      The most explored therapeutic approach to inhibit L1 retrotransposition has been the use of nucleoside analog reverse transcriptase inhibitors (NRTIs), which are a class of anti-viral compounds that inhibit nucleic acid polymerases, including reverse transcriptase, and have been used in the treatment of HIV (Pau and George, 2014). One study demonstrated by using an in vitro LINE-1 retrotransposition assay that L1 retrotransposition could be suppressed by a number of different NRTIs with different degrees of potencies (Jones et al., 2008). Consistent with this, another group showed that NRTIs effectively inhibit the reverse transcriptase activity of the L1 protein, ORF2p, consequently reducing retrotransposition (Dai et al., 2011). In addition, NRTIs have been shown to abrogate the accumulation of cytoplasmic L1 DNA in senescent cells (De Cecco et al., 2019; Simon et al., 2019). Given that cytosolic L1 DNA triggers an IFN-I response and contributes to age-associated inflammation (De Cecco et al., 2019; Simon et al., 2019), NRTIs could serve as potential therapies for the treatment of age-related pathologies. Indeed, treating aged mice with NRTIs effectively reduced L1 cytoplasmic DNA and the ensuing inflammation in a variety of tissues (De Cecco et al., 2019; Simon et al., 2019) and improved a number of age-related phenotypes, such as tissue macrophage infiltration, glomerulosclerosis of the kidney, and skeletal muscle atrophy (De Cecco et al., 2019). Similar improvements in health span have also been observed in SIRT6 knockout progeroid mice treated with NRTIs (Simon et al., 2019). Furthermore, increased LINE-1 retrotransposition has been implicated in fetal oocyte attrition, a process by which more than two thirds of oocytes are eliminated before birth (Chuma, 2014). It was shown that treating pregnant female mice with the nucleoside analog azidothymidine (AZT) blocked oocyte attrition and improved the oocyte viability in embryos (Malki et al., 2014). This raises the possibility that reverse transcriptase inhibitors, by targeting L1 elements, might also serve as therapies to extend the female reproductive lifespan.

      Although mouse studies suggest that NRTIs are an effective treatment option for L1-driven age-associated inflammation and pathologies, chronic NRTI treatment has been shown to induce adverse side effects in human patients, such as hepatotoxicity, as it also inhibits the mitochondrial DNA polymerase g (Montessori et al., 2003; Wu et al., 2017). Moreover, NRTIs can also inhibit telomerase and are thought to contribute to accelerated aging phenotypes associated with HIV patients (Hukezalie et al., 2012; Leeansyah et al., 2013). Therefore, development of more specific interventions is needed in order to safely and effectively treat conditions arising from uncontrolled L1 activity.

      For example, strategies to promote degradation of cytoplasmic L1 DNA, and thus prevent inflammation, could be an attractive avenue for therapy development.

Perspective and conclusions

      The ability of foreign nucleic acids to stimulate immune responses has been known since the 1960s (Isaacs et al., 1963; Rotem et al., 1963). However, this strategy of pathogen sensing also permits the detection of mislocalized or aberrant endogenous nucleic acids (Barbalat et al., 2011). Recent breakthroughs at the interface of diverse fields of biomedical research—cancer, immunology, aging, DNA damage, and chromatin regulation to name a few—have accelerated progress on understanding molecular mechanisms of endogenous cytoDNA formation and function. Although it is clear that cytosolic localization of endogenous DNA triggers signaling responses with relevance to multiple chronic diseases, much remains unknown. We have identified two primary questions: (1) what are the mechanisms of endogenous cytoDNA species formation and function, and (2) how do interactions between endogenous cytoDNA pathways maintain normal physiology and drive disease? A better understanding of the similarities, differences, and interactions between cytoDNA species and associated pathways can inform therapeutic opportunities in the treatment of human disease.

Common mechanisms of cytoDNA formation and function

      A consideration of common mechanisms can lead to new insights (Figure 5). First, comparison between endogenous cytoDNA mechanisms of formation and signaling can give insight into the roles of cytoDNA in physiology and disease and reveal many similarities. For example, the loss of nuclear membrane integrity in senescent cells (Shimi et al., 2011) and laminopathies, such as HGPS, parallels loss of membrane integrity in MN (Bartsch et al., 2017; Gekara, 2017; Harding et al., 2017; Mackenzie et al., 2017). These parallels include nuclear envelope dysfunction, changes in lamina structure, accumulation of DNA damage, and downstream activation of cGAS (Liu et al., 2018). Loss of nuclear membrane integrity is also a potential mechanism for cytosolic localization of retrotransposons, although the cytoplasmic reverse transcription of L1 is also a possible mechanism (Di Micco et al., 2016; Simon et al., 2019). Second, although the mechanisms involved are poorly understood, CCF and mtDNA are thought to be partially localized in endosomes, where they can be sensed, degraded, or potentially secreted by the cell. Degradation by DNases, such as lysosomal DNase2 or cytoplasmic TREX has been observed for all forms of endogenous cytoDNA, acting as a negative feedback to downregulate DNA sensor activation. Degradation by autophagy has also been proposed for CCF and MN (Dou et al., 2015; Rello-Varona et al., 2012; Sagona et al., 2014). Recent work has shown that cGAS itself can act as an autophagy ligand for MN degradation, suggesting that autophagic degradation may be a common fate of cGAS-associated cytoDNA (Zhao et al., 2021). Multiple forms of cytoDNA may be secreted and act as a cell non-autonomous signal in physiology and disease, which has been proposed as a biomarker for sterile inflammation (Ruffatti et al., 1990). Third, cytoDNA species share similar signaling pathways, most prominently the cGAS-STING pathway, although mtDNA, MN, and other DNA damage-associated DNA fragments are thought to also activate the AIM2 and NLRP3 inflammasome pathway (Di Micco et al., 2016; Motwani and Fitzgerald, 2017; Zhong et al., 2018). It will be interesting to test whether additional DNA sensors are also shared between cytoDNA species. These shared functional mechanisms are consistent with an evolutionarily conserved strategy to detect foreign DNA as a danger signal. Interestingly, many species of bats have evolved to naturally tolerate cytoDNA through decreased cytoDNA sensing (Gorbunova et al., 2020). Fourth, the observation of multiple cytoDNA species in senescent cells echoes the role of foreign DNAs as activators of cellular anti-virus mechanisms and reinforces the links between cell senescence and anti-virus programs. Specifically, senescence can be induced by viral infection (Seoane et al., 2020), senescent cells activate anti-viral pathways (Rai et al., 2017), and replication of several viruses is impaired in senescent cells dependent on the SASP (Baz-Martı´nez et al., 2016; Seoane et al., 2020). Notably, senolytic treatment in animal models of COVID-19 eliminates virus-induced senescent cells and reduces inflammation and mortality, suggesting that senescent cells are a potential therapeutic target in viral infections (Camell et al., 2021; Lee et al., 2021). At the molecular level, key regulators of the senescent cell phenotype are also central to virus defense mechanisms and often targeted by viruses, including tumor suppressors pRB and p53, histone chaperone HIRA, PML nuclear bodies, as well as cGAS, STING, and ISGs. Discovery of additional shared, conserved ‘‘seno-viral’’ mechanisms may inform therapeutic strategies simultaneously targeting multiple cytoDNA species, for example modulation of the SASP in senescent cells that is likely driven or augmented by CCF, mtDNA, and RT.

Distinct mechanisms of cytoDNA formation and function

      Although there are fundamental similarities between endogenous cytoDNA species, detailed examination reveals mechanistic differences and many open questions (Figure 5). Regarding mechanisms of formation, the temporal component of CCF, mtDNA, and RT cytoplasmic localization—as well as the temporal component of cytoDNA sensing—is poorly understood. For example, in senescent cells, CCF formation occurs after cell-cycle arrest, which is consistent with a cell-cycle-independent mechanism of formation (Ivanov et al., 2013); however, the direct role of MN and mtDNA in this process is unknown. Senescent cells also display RT activation (De Cecco et al., 2019); however, this activation only occurs late in senescence (a few months), versus CCF formation (a few days). Because these processes potentially occur within the same cell, are associated with similar formation mechanisms (e.g., nuclear envelope dysfunction, DNA damage), and activate cGAS, it will be interesting to dissect this temporal relationship. We speculate that degradation of unrepaired DNA in senescent cells by CCF formation alters epigenome structure in a progressive manner that is eventually permissive for RT de-repression. Temporal evolution of the senescent cell phenotype has been noted previously in the progression to ‘‘deep senescence’’ (Chen and Ozanne, 2006; De Cecco et al., 2013a; Ivanov et al., 2013; Passos et al., 2010).

      The compartmentalization of cytoDNA within cytoplasmic compartments is poorly understood. Electron microscopy of oncogene-induced senescent cells show that CCF can localize within double membrane-enclosed autophagosomes, and fluorescence microscopy shows that CCF colocalize with lysosomes, dependent on activation of macroautophagy (Dou et al., 2015); however, whether CCF are always incorporated within autophagosomes coincident with their formation or sometimes engulfed after their rupture through a herniated nuclear envelope is not entirely clear (Ivanov et al., 2013; Dou et al., 2015). Whether autophagic degradation is the only fate of CCF is also unclear. Similarly, mtDNA is known to localize to endosomes and activate DNA sensor and pro-inflammatory mediator TLR9 (Zhang et al., 2010). Because endosomes are topologically contiguous with the exterior of the cell, endosomal localization of DNA is typically thought to depend on endocytosis of extracellular DNA (Barbalat et al., 2011). This raises questions as to how cytoplasmic mtDNA gains access to endosomes. The localization of CCF in autophagosomes suggests that trafficking between autophagosomes and other endosomal compartments is another potential localization mechanism. Consistent with this hypothesis, the anti-microbial peptide LL37 has been shown to promote trafficking of aggregated DNA structures to the early endosome, promoting TLR9 activation and interferon production (Lande et al., 2007). Further research into cytoDNA trafficking within the cytoplasm can inform therapeutic strategies to degrade cytoDNA or prevent its secretion, both of which have proposed therapeutic potential in autoimmune diseases.

      Although the vastmajority of studiesfocus on cGAS as a primary sensor of cytoDNA species, the regulation of cGAS appears to be context-dependent. For example, although MN-dependent cGAS activation of interferon genes is well-documented in cancer cells, CCF activation of cGAS in senescent cells is typically associated with an NF-kB-dependent cytokine response, a phenotype that is thought to be partly dependent on p38-MAPK activation in senescent cells (Dou et al., 2017). There are potentially other factors influencing the cGAS pathway—indeed, recent work on cGAS sensing of DNA is revealing a growing network of regulation, including inhibition by phosphorylation (Li et al., 2021), inhibition by nucleosome binding (Boyer et al., 2020; Kujirai et al., 2020; Michalski et al., 2020; Pathare et al., 2020; Zhao et al., 2020b) and chromatin structure (Uggenti et al., 2020), activation by manganese (Wang et al., 2018), activation by coactivators such as TOP1cc in the context of chemotherapeutic DNA damageinduced senescence (Zhao et al., 2020a), and by TFAM in the context of mtDNA sensing (Andreeva et al., 2017). In at least some contexts, localization of endogenous cGAS is thought to be mostly nuclear (de Oliveira Mann and Hopfner, 2021; Volkman et al., 2019), where it binds to nucleosomes in an inactive state— therefore the exact mechanism of cGAS activation (e.g., in rupture of MN) is unclear. At least in the context of exogenous DNA transfection and vaccinia virus infection, there is recent evidence that cytoplasmic trafficking of nuclear cGAS is required for cytoDNA sensing (Sun et al., 2021). However, the relative signaling role of cGAS when ferried as a passenger from the main nucleus into MN and CCF versus cGAS recruitment from the cytoplasm remains to be resolved. Given that protein-DNA complexes play a central role in DNA function, it would not be surprising if cytoplasmic localization of DNA is accompanied by a contextdependent and specific complement of cytosolic proteins. Certainly, cGAS has been reported to be capable of sensing all major cytoDNA species, but detection of cytoDNA by other DNA sensors, proteins specific to cytoDNA origin, and proteins specific to cytoplasmic sub-compartments could also play importantfunctional, therapeutically targetable roles. Indeed, proteomic characterization of CCF has already identified such targets (Zhao et al., 2020a). The genomic regions or specific loci that generate CCF remain unclear, although an enrichment for heterochromatinassociated histone marks suggests that CCF are generated from heterochromatin. Single-cell genome sequencing has greatly improved understanding of the mechanism of MN formation and chromothripsis (Umbreit et al., 2020), suggesting that similar methods could answer this outstanding question in CCF biology. Determining the DNA sequence of CCF may inform therapeutic targeting and mechanisms of formation.

      Last, although cGAS/STING pathway inhibitors are promising strategies in treatment of chronic inflammation and some cancers, use of these drugs must consider the diversity of endogenous cytoDNA functions. For example, inhibiting cGAS/STING signaling downstream of CCF activation may be beneficial in the context of chronic inflammation with age (Decout et al., 2021), but this strategy may also impair signaling of MN-initiated cell-intrinsic immune surveillance and tumor suppression (Harding et al., 2017; Mackenzie et al., 2017). However, CCF and MN are currently difficult to distinguish microscopically and are often confused (Table 1). Therefore, therapeutic approaches must consider specific endogenous cytoDNA species, including mechanism of formation, localization and dynamics, and interactions between different cytoDNA species in contexts where multiple species are present, such as senescent cells.

Conclusion

      DNA localization in the cytoplasm of a cell, whether exogenous or endogenous, acts as a potent danger signal that stimulates an innate immune response. Although this field has historically focused on pathogenic activation of these pathways, more recent work has uncovered endogenous cytoDNA as a major contributing factor in chronic diseases, especially cancer. These chronic diseases—also including cardiovascular disease, neurodegenerative disease, sarcopenia, etc.—share a common variable, the typically advanced age of the patient. Chronic inflammation, driven in part by cytoDNA, is associated with multiple proposed mechanisms and pathologies of aging biology, suggesting that cytoDNA is a therapeutic target in extending human health span. Although preclinical studies targeting cytoDNA show promise, many obstacles remain. Understanding the mechanisms of cytoDNA formation and signaling, especially subtype-specific mechanisms of cytoDNA formation and sensing, will help to realize the potential of cytoDNA-targeted therapies for the treatment of chronic disease.

ACKNOWLEDGMENTS

      We apologize to our colleagues whose excellent work could not be cited due to space limitations. This work was supported by NIH P01 AG031862-13 and NIH R01 AG071861-01 (to P.D.A.), NIH F32 AG066459-02 (to K.N.M.), a Glenn Foundation for Medical Research Postdoctoral Fellowship PD19131 (to N.D.), and a Robert and Arlene Kogod Center on Aging Career Development Award (to S.G.V.). Funding in J.F.P.’s laboratory was provided by NIH R01 AG068048-01 and NIH UL1TR02377-1, the Ted Nash Long Life Foundation, and NIH P01 AG62413.

AUTHOR CONTRIBUTIONS

Conceptualization, K.N.M., S.G.V., J.F.P., and P.D.A. Writing, K.N.M., S.G.V., and N.D. Visualization, H.S. and T.L. Supervision, J.F.P. and P.D.A.

DECLARATION OF INTERESTS

The authors declare no competing interests.

This  article  is excerpted from the Cell 184, October 28, 2021 by Wound World.