Generally, autophagy consists of four distinct phases: (1) initiation, (2) nucleation, (3) elongation, (4) fusion and degradation, which involve many protein complexes required for autophagosome biogenesis, trafficking, fusion with lysosomes, and cargo degradation (1, 20–22).

During the process of selective autophagy, the autophagic machinery is targeted to specific cargos via selective autophagy receptors independent of the energy sensing pathways mTORC1 and AMPK. Selective autophagy can be classified based on different types of cargo into mitophagy (damaged mitochondria), aggrephagy (protein aggregates), xenophagy (intracellular pathogens), and many more (3).

Cargo selectivity is conferred by selective autophagy receptors. The majority of selective autophagy receptors are soluble proteins referred to as sequestosome-like cargo receptors (SLRs). Among them the most prominent ones are SQSTM1/p62, NDP52, next to BRCA1 gene 1 protein (NBR1), Tax1-binding protein 1 (TAX1BP1) or optineurin (OPTN). Selective autophagy and its receptors have been reviewed expertly elsewhere (3, 40, 92–95). In general, their primary function is to create a direct link between the cargo and the autophagic machinery allowing their selective degradation in the lysosome.

These receptor proteins are characterized by the presence of LIR motifs for the interaction with ATG8 family proteins such as LC3 on the surface of phagophores, and many have ubiquitin-binding domains capable of recognizing ubiquitinated cargo, thus establishing a bridge between the cargo and the autophagic machinery (3, 40, 92, 93). Selective autophagy receptors such as NDP52, TAX1BP1 and p62 are also capable of directly recruiting the ULK1/2 complex through binding to FIP200 (subunit of the ULK1 complex) to initiate de novo autophagosome biogenesis at the site of the cargo (40, 42, 43). Moreover, OPTN can directly recruit ATG9A or TANK-binding kinase 1 (TBK1), while NIX can recruit WIPI2, resulting in mitophagy induction (96–98).

A special role during the process of selective autophagy is allocated to TBK1. Initially it was thought that the main role of TBK1 during selective autophagy was facilitating the interaction between ATG8 family protein members and the selective cargo via selective autophagy receptors. We know now that the role of TBK1 in regulating selective autophagy is much broader than previously thought with recent studies implicating TBK1 in the direct recruitment of cargo receptors as well as the early autophagic machinery to intracellular bacteria and mitochondria, which will be described in more detail below (40, 99).

Under physiological conditions, DNA is exclusively found in the nucleus and in mitochondria. However, certain types of cellular stress can lead to the accumulation of DNA in the cytosol, where it is then recognized by cGAS. Cytosolic DNA may originate from DNA viruses, intracellular bacteria or RNA viruses causing the release of mitochondrial DNA (mtDNA) (101, 107–111). In addition, cGAS can recognize not only foreign DNA but also endogenous DNA that is released into the cytosol from mitochondria or micronuclei resulting from mitotic defects ( Figure 2 ). Extracellular DNA originating from dead cells can also be taken up by cells and stimulate the cGAS-STING pathway (111–116). cGAS is also found in the nucleus but is kept inactive due to nucleosome binding and chromatin tethering (117). However, nuclear cGAS can be activated by human immunodeficiency virus (HIV) in dendritic cells and macrophages (118).

Structurally, cGAS can bind double-stranded DNA as well as single-stranded DNA with secondary structures and has even been shown to bind RNA-DNA hybrids (119–122). Even though both small and large DNA fragments can be bound, cGAS activation requires a certain minimal fragment size and has been shown to improve with increasing DNA length (120–124). A very recent publication proposes that mechanical flexibility of DNA, which is increased in the case of DNA damage and higher AT content, promotes the binding and activation of cGAS (125).

Upon DNA binding, cGAS undergoes a conformational change, leading to activation and dimerization with two DNA molecules sandwiched in between (124, 126–128). Those dimers can undergo liquid-liquid phase separation in a DNA length-dependent manner, forming condensates that act as “minireactors” and concentrate cGAS as well as substrates to increase the catalytic activity of cGAS. This could also explain why longer DNA fragments activate cGAS more efficiently (129).

Once activated, cGAS uses GTP and ATP as substrates to produce the secondary messenger cGAMP, which can then be recognized by the ER transmembrane protein STING (130–132). Additionally, cGAMP can diffuse into neighboring cells via gap junctions or be released into the local microenvironment to activate STING signaling in surrounding cells (133–137).

STING resides at the ER membrane as a dimer, where two C-terminal ligand-binding domains form a binding pocket for cGAMP. Binding of cGAMP induces a conformational change that facilitates oligomerization of STING dimers and STING activation (138–140).

Active STING is translocated from the ER to the ER-Golgi intermediate compartment (ERGIC) and Golgi via COPII vesicles, which requires secretion-associated and Ras-related GTPase 1A (SAR1), SEC24 homolog C (SEC24C) and ADP-ribosylation factor (ARF) family members (141).

At the Golgi and in post-Golgi compartments, TBK1 is recruited and then activated by STING (142). Activated TBK1 phosphorylates the C-terminal tail (CTT) of STING at several residues including Ser366, which is part of a highly conserved motif (pLXIS) (143, 144). This motif recruits IRF3, which is then phosphorylated by TBK1, inducing its dimerization. Phosphorylated IRF3 dimers can translocate into the nucleus where they act as transcription factors for type I interferons and other immune-related proteins (143–145) ( Figure 2 ). After activation of IRF3, cGAMP-bound STING oligomers are transported through the trans-Golgi and sorted into RAB7-positive late endosomes that travel to lysosomes for degradation via multivesicular bodies (141).

In addition to its function as an activator of type I IFN synthesis and autophagy, STING can also activate the NF-κB pathway in a TBK1-IKKϵ-dependent or TBK1-independent manner that involves DNA damage response factors ataxia telangiectasia mutated (ATM) and poly(ADP-ribose) polymerase 1 (PARP1) (146–148).

The interplay between autophagy and cGAS occurs through (i) cGAS-Beclin-1 interaction that promotes autophagic degradation of cytosolic dsDNA, (ii) cGAS-LC3 interaction that promotes micronucleophagy, and (iii) degradation of cGAS through p62-mediated selective autophagy ( Figure 3 ).

Robust activation of the enzymatic function of cGAS is dependent on the length of dsDNA fragments and longer fragments (>45bp) promote robust activation of cGAS (123, 149). Conversely, small cytosolic dsDNA fragments decrease cGAS-STING signaling by competing with long dsDNA fragments for cGAS binding. Binding of small dsDNA fragments to cGAS promotes the binding of cGAS to Beclin-1. By binding to Beclin-1, cGAS outcompetes run domain Beclin-1 interacting and cysteine-rich containing protein (Rubicon) (150). Rubicon negatively regulates maturation and fusion of the autophagosome with the lysosome and endocytic trafficking by suppressing VPS34 kinase activity (151–153). Binding of cGAS to Beclin-1 leads to the release of Rubicon, thereby enabling autophagic degradation of cytosolic dsDNA and resulting in decreased cGAS-STING signaling ( Figure 3 ). Moreover, Beclin-1 also suppresses the NTase activity of cGAS to decrease the production of the secondary messenger cGAMP and type I IFN production (150, 154).

Furthermore, cGAS itself has been proposed as a potential selective autophagy receptor for autophagic degradation of micronuclei also known as micronucleophagy (155). Overexpression of cGAS significantly reduces the number of micronuclei in an autophagy-dependent manner independent of its role in cGAS-STING signaling. cGAS was shown to directly interact with the ATG8 family protein LC3 and colocalize at micronuclei, thereby facilitating the recruitment of the autophagic machinery to enable lysosomal degradation of micronuclei ( Figure 3 ). Moreover, cGAS-driven micronucleophagy also dampens cGAMP production and altogether exerts a negative effect on cGAS-STING signaling and innate immunity (155).

In another report, it was shown that K48-linked ubiquitinated cGAS is directly sequestered into the lysosome for degradation by interacting with the selective autophagy receptor p62 ( Figure 3 ). Upon viral infection, the ISG tripartite motif 14 (TRIM14) directly binds to the C-terminus of cGAS and recruits ubiquitin-specific peptidase 14 (USP14) to prevent K48-linked ubiquitination and degradation of cGAS. Consequently, cGAS evades p62-mediated autophagic degradation and can stimulate type I IFN signaling after viral infection (156).

STING and autophagy are connected on four levels: (i) STING facilitates non-canonical induction of autophagy, (ii) microautophagy and selective macroautophagy can induce STING degradation, (iii) ULK1 negatively regulates STING-IRF3 signaling by phosphorylating STING, and (iv) non-canonical RAB22-mediated autophagy can promote intercellular spreading of activated STING ( Figure 4 ).

Autophagy induction has been proposed to be the primordial function of cGAS-STING signaling before interferon induction evolved. STING can induce autophagy independent of TBK1 and IRF3 via an LR motif at the C-terminus of STING (141, 157). This motif can also be found in other organisms such as Nematostella vectensis that are lacking the C-terminal activation domain required for interferon production. Upon cGAMP binding, STING moves to the ERGIC in a COPII- and ARF GTPase-dependent manner. The ERGIC can then serve as a membrane source for LC3 lipidation, which is dependent on the adaptor protein WIPI2 and ATG5 but does not require ULK1 or VPS34, indicating that this non-canonical STING-dependent induction of autophagy can “skip” early steps of the canonical autophagy mechanism (141). STING directly recruits WIPI2 to STING-positive vesicles and competes with PI3P (canonical bulk autophagy) for binding to WIPI2, adding another layer of complexity to the feedback mechanism between autophagy and cGAS-STING signaling (158) ( Figure 4 ). STING-induced non-canonical autophagy was shown to inhibit herpes simplex virus (HSV-1) (159).

Myristic acid has recently emerged as a regulator of STING degradation via N-myristoylation of ARF1, which is involved in directing the membrane trafficking and autophagic degradation of STING, thereby limiting cGAS-STING-induced interferon response. Upon viral infection the production of myristic acid is reduced due to a disruption in the cellular lipid metabolism, which alleviates autophagy-dependent STING degradation (160).

Another recent report postulated that STING-induced LC3 lipidation happens independent of WIPI2 at perinuclear single membrane vesicles, rather than double-membrane autophagosomes, and requires both the V-ATPase as well as the WD40 domain of ATG16L1 required for interaction with V-ATPase (161, 162). Autophagy-independent LC3 lipidation occurs during influenza A virus infection and mouse models lacking the WD40 domain of ATG16L1 show a reduction in major histocompatibility complex class II (MHC-II)-dependent antigen presentation in dendritic cells (162). The V-ATPase-ATG16L1 axis was also shown to initiate intracellular clearance of bacteria (163). In line with this, STING was recently shown to have proton channel activity that is essential for both non-canonical LC3 lipidation and inflammasome activation. This would allow for mechanistic decoupling of IFN induction via phosphorylated STING and proton channel-dependent LC3 lipidation and inflammasome activation (164). In conclusion, STING can induce non-canonical autophagy in a WIPI2-dependent manner at the ERGIC and autophagy-independent LC3 lipidation at perinuclear single membrane vesicles facilitating defense against viruses and bacteria.

While STING can induce autophagy, autophagy can degrade STING through multiple mechanisms including microautophagy and selective macroautophagy. In microautophagy, cytoplasmic cargo is directly engulfed by the lysosome. STING is transported by clathrin-coated vesicles and recycling endosomes (141) and directly encapsulated in lysosomal associated membrane protein 1 (LAMP-1) positive compartments through the ESCRT components tumor susceptibility gene 101 (TSG101) and vacuolar sorting-associated protein 4 (VPS4) (165). STING packaging from the trans-Golgi into clathrin-coated vesicles is mediated by the cargo adaptor complex 1 (AP-1), which also controls the termination of STING-dependent immune activation by an as of yet unknown mechanism (166) ( Figure 4 ). In addition, basal STING levels seem to be kept under control by the ER-associated degradation (ERAD) system. The ERAD protein complex SEL1L-HRD1 directly interacts with and mediates targeted degradation of basal inactive STING thereby limiting the activatable STING pool and preventing overactivation of the immune response (167). Alternatively, STING may be degraded through selective macroautophagy, whereby TBK1-mediated phosphorylation of p62 at S403 increases its affinity for K63-ubiquitinated STING (168) and ubiquitously expressed transcript (UXT) may facilitate the interaction between p62 and STING (169). Taken together, microautophagy and p62-mediated selective autophagy regulate STING degradation and thereby keep the interferon response in check.

Autophagy has also been shown to have a more direct function in negatively regulating STING-dependent signaling. ULK1 phosphorylates STING at S366, thereby inhibiting IRF3- but not NF-κB-dependent interferon production upon stimulation with dsDNA (170).

A non-canonical autophagy pathway that relies on RAB22A was shown to facilitate the spreading of activated STING to neighboring cells. RAB22A activates phosphatidylinositol 4-phosphate 3-kinase C2 domain-containing subunit alpha (PI3K2A) to produce phosphatidylinositol 4-phosphate (PI4P) that recruits ATG12-ATG5-ATG16L1. The ATG16L1 complex then induces the formation of ER-derived non-canonical autophagosomes containing activated STING. RAB22A-induced autophagosomes fuse with early endosomes, while inhibiting RAB7-mediated fusion with lysosomes. Secretion of the RAB22A-induced extracellular vesicles enables intercellular transfer of activated STING and the spreading of interferon induction to neighboring cells similarly to cGAMP (171) ( Figure 4 ).

In summary, STING can induce non-canonical formation of autophagosomes via WIPI2 using the ERGIC as a source for autophagosome lipidation and promotes autophagy-independent LC3 lipidation. However, none of these pathways seem to mediate the degradation of STING. Instead, STING is degraded by direct microautophagic uptake by lysosomes after being packaged into clathrin-coated vesicles, which is mediated by AP-1. In addition to microautophagic degradation, p62-dependent degradation of STING might also contribute to this phenomenon. Lastly, STING can be taken up by non-canonical RAB22A-induced autophagosomes that do not fuse with lysosomes but are secreted to surrounding cells, spreading active STING. Overall, STING positively regulates and induces different types of autophagy, while autophagy negatively impacts STING signaling ( Figure 4 ).

The kinase TBK1 is known to fulfill a broad range of functions in our cells ranging from IFN signaling, innate immunity and inflammation to autophagy. It is likely that distinct functions of TBK1 in different cellular pathways are due to its cellular localization and association with different adaptor proteins (99, 172). Moreover, TBK1 may simultaneously regulate autophagy and IFN signaling by controlling starvation-induced autophagy and different forms of selective autophagy (e.g., mitophagy or xenophagy). TBK1 may regulate the clearance of dsDNA in the cytosol and thereby modulate cGAS-STING activation and IFN production. TBK1 can regulate autophagy by: (i) activating or inhibiting mTORC1, (ii) inhibiting AMPK, (iii) promoting mitophagy, and (iv) promoting xenophagy ( Figure 5 ).

During starvation-induced autophagy, TBK1 can positively and negatively regulate autophagic flux by inhibiting or activating mTORC1 respectively (173, 174). TBK1 can phosphorylate and activate regulatory-associated protein of mTOR (RAPTOR), which decreases activity of mTORC1 and promotes autophagy (173). Besides inhibition of mTORC1, TBK1 can directly activate mTORC1 by phosphorylation of mTOR on S2159 to control innate immunity (174) ( Figure 5 ). Whether TBK1 activates or inhibits mTORC1 seems to be cell-context dependent. TBK1 can also initiate autophagy under starvation conditions by phosphorylating and activating the autophagosome-lysosomal fusion protein syntaxin 17 (STX17). Phosphorylated STX17 translocates from the Golgi to the mammalian PAS and controls the formation and assembly of the ULK1 complex, thereby promoting initiation of the autophagic process (175).

The energy sensor AMPK is also regulated by TBK1. Mice with high fat diet have high protein levels of TBK1, which negatively regulates AMPK activity. At the same time, active AMPK indirectly enhances TBK1 activity by promoting ULK1-mediated phosphorylation of TBK1. This creates a negative feedback loop whereby AMPK activates TBK1, which downregulates AMPK, possibly explaining diminished obesity in TBK1 KO mice (176). Since AMPK is known to promote autophagy via direct activation of ULK1 (24), this negative feedback loop between TBK1 and AMPK can negatively affect autophagic flux ( Figure 5 ).

TBK1 is also involved in the regulation of mitophagy, especially in the early steps, where it is instrumental for initiating autophagosome biogenesis (43, 97, 177, 178). Mitophagy is activated by PTEN-induced kinase I (PINK1), which upon mitochondrial depolarization accumulates at the outer mitochondrial membrane (OMM) followed by the recruitment of PARKIN and ubiquitination of mitochondrial proteins (40, 179). During PINK1-PARKIN-induced mitophagy, the selective autophagy receptors OPTN and NDP52 redundantly activate TBK1, which in turn phosphorylates the mitophagy receptors OPTN, NDP52 and p62 to ensure their retention at damaged mitochondria and to facilitate the recruitment of ATG8 family proteins (180, 181). TBK1-mediated phosphorylation of NDP52 facilitates direct interaction with the FIP200/ULK1 complex thus enabling spatial proximity to ubiquitinated cargo to drive autophagosome biogenesis (43). OPTN-induced mitophagy is not necessarily dependent on the binding to the FIP200/ULK1 complex as shown for NDP52. In fact, TBK1 takes over the functions of the ULK1 complex and directly interacts with the PI3K complex I to initiate mitophagy (97). TBK1-mediated phosphorylation of OPTN supports mitophagy by stabilizing the binding of OPTN to ubiquitinated cargo (181). Reciprocally, OPTN-mediated TBK1 relocalization to mitochondria is crucial for proper TBK1 activation (182, 183). TBK1 also phosphorylates RAB7A, which promotes the recruitment of ATG9-positive vesicles to damaged mitochondria enabling efficient mitophagy (184) ( Figure 5 ). Recently, it was shown that mitochondria can be cleared upon PINK1-PARKIN mediated mitophagy in the absence of the mATG8-conjugation machinery and independent of lysosomal degradation via a process called autophagic secretion of mitochondria (ASM). ASM promotes cGAS-STING signaling in recipient cells. The exact mechanism of how mtDNA-containing extracellular vesicles get in contact with cytosolic cGAS of recipient cells remains to be elucidated (185).

Xenophagy is another form of selective autophagy responsible for elimination of pathogens or clearance of invading bacteria which, like mitophagy, relies on TBK1 (3). During early xenophagy, Galectin-8 (GAL8) recognizes exposed glycans on ruptured vacuoles and subsequently recruits the selective autophagy receptor NDP52. Binding of NDP52 to the two adaptor proteins NAP1 and SINTBAD enables recruitment of TBK1 and the formation of an NDP52-ULK1-TBK1 super complex, which is essential for WIPI2-positive phagophore formation and for the recruitment of LC3 (42) ( Figure 5 ). In addition, the ubiquitin ligase RNF213 directly ubiquitylates lipopolysaccharides (LPS) upon vacuole rupture, followed by the recruitment of linear (M1) Ub assembly complex (LUBAC), which is responsible for the assembly of M1 ubiquitin chains. The recruitment of the selective autophagy receptor OPTN and the immune adaptor NF-κB essential modulator (NEMO) is dependent on LUBAC function, while the recruitment of the selective autophagy receptors p62 and NDP52 requires only RNF213 (186). TBK1 directly phosphorylates S403 of p62, which is required for the autophagic function of p62 in driving autophagosome maturation (187). OPTN is also directly phosphorylated by TBK1, thereby enhancing LC3 binding and autophagic clearance of ubiquitin-coated bacteria (188) ( Figure 5 ).

TBK1 is known to be tightly regulated by multiple post-translational modifications to prevent sustained and prolonged immune activation (99, 189–191). The E3 ubiquitin ligase neural precursor cell-expressed developmentally downregulated gene 4 (NEDD4) controls the stability of TBK1 through K27-linked polyubiquitination. The selective autophagy receptor NDP52 binds polyubiquitinated TBK1 and elicits autophagic degradation, resulting in decreased type I IFN signaling (189) ( Figure 6 ). Lysosomal degradation of TBK1 is also promoted by ubiquitin specific peptidase 19 (USP19), which interacts with the TBK1-HSPA8 complex via the chaperone-mediated autophagy (CMA) motif of TBK1 (190) ( Figure 6 ). Another way in which autophagy negatively regulates TBK1 stability is through the autophagy protein ATG4B, which serves as an adaptor for the recruitment of TBK1 to gamma-aminobutyric acid receptor-associated protein (GABARAP) via its LIR motif. Subsequently, TBK1 is sequestered into the lysosome and degraded ( Figure 6 ). Pharmacological inhibition of ATG4B-dependent autophagic degradation of TBK1 was shown to increase the overall antiviral response (192). Taken together, TBK1 stability is regulated by autophagy to prevent sustained type I IFN production and prolonged immune reaction.

Autophagic degradation of IRF3 is another mechanism of keeping immune activation in check. NDP52 mediates the selective degradation of IRF3 upon viral infection ( Figure 7 ). Under homeostatic conditions, the deubiquitinase PSMD14 cleaves K27-linked polyubiquitin chains at K313 of IRF3, thereby preventing autophagic recognition and degradation (193). Furthermore, OUT deubiquitinase 7B (OTUD7B) promotes autophagic degradation of IRF3 upon viral infection by enhancing the association between IRF3 and p62, thereby negatively regulating the immune response. OTUD7B directly interacts with IRF3 and activates p62 by removing K63 polyubiquitin chains leading to enhanced p62 oligomerization. This allows more efficient degradation of IRF3 in a p62-dependent manner (194) ( Figure 7 ). The histone deacetylase HDAC10 can bind to IRF3 in a deacetylase-independent manner and inhibit its TBK1-dependent phosphorylation at S396. Viral infection induces targeted autophagic degradation of HDAC10 resulting in increased phosphorylation of IRF3 and enhanced interferon response (195).

Genomic instability resulting from defects in DNA damage response, chromosomal instability or defects in chromosome segregation can lead to the formation of micronuclei and generation of aberrant cytosolic dsDNA fragments (113, 201, 202). Micronuclei are small nuclei-like structures containing chromatin encapsulated by the nuclear envelope, which arise from chromosomal instability (203). Micronuclei can rupture and release genomic dsDNA into the cytosol, which is directly engaged by cGAS (113, 204). In addition to genomic dsDNA, dsDNA released from mitochondria can also directly activate the cGAS-STING pathway (205). Another way to activate cGAS is through dying tumor cells, which release DNA into their surroundings. This DNA can be taken up by dendritic cells and activate the cGAS-STING pathway (206).

One of the major concepts in cancer biology is the cancer-immunity cycle, a cyclic self-propagating process that leads to the initiation and amplification of T-cell responses against cancer cells (197). Functional STING signaling affects multiple stages of the cancer-immunity cycle and can enhance tumor-specific T-cell responses (197, 198, 207). Type I IFNs produced by activation of the cGAS-STING pathway instigate the differentiation and maturation of antigen-presenting cells (APCs) and the expression of MHC molecules, thereby facilitating enhanced tumor antigen presentation (208–211). Consequently, cGAS-STING signaling is important for priming, activation and infiltration of tumor-specific cytotoxic T cells (212–214). Active cGAS-STING signaling has also been linked with the production of certain chemokines that recruit T cells to the tumor site such as CXCL10, as well as induction of T-cell stemness (197, 215). Generally, the induction of the type I IFN response targets different immune cells from antigen-presenting dendritic cells, T cells to natural killer (NK) cells (200). In addition to the production of type I IFNs, STING activation can induce the NF-κB pathway and the expression of various proinflammatory cytokines such as interleukin-6 (IL-6) and tumor necrosis factor-a (TNF-α) (200). Apart from its intrinsic function in tumor and immune cells, cGAS-STING can signal intercellularly by cGAMP spreading to neighboring cells via gap junctions (134, 216, 217). This process is an active field of study and has recently been reviewed (217). Conventional cancer therapies including radio- and chemotherapy can induce cGAS-STING and anti-tumor immunity alone or in combination with STING agonists and immune checkpoint blockade (198, 218, 219).

In contrast to its anti-tumorigenic function, cGAS-STING signaling has also been implicated in oncogenesis (200, 220–222). The major culprits seem to be chronic type I IFN production and inflammation, which promote cancer cell invasion, metastasis and cancer cell stemness (200, 223–225). Recently, active STING signaling has been linked to the survival of chromosomally unstable cancers in an IL-6 dependent manner involving both NF-κB and signal transducer and activator of transcription 3 (STAT3) activation (226). Additionally, chromosomal instability can promote metastasis via STING-dependent NF-κB activation (227).

Overall, the relationship between pro- and antitumorigenic functions of the cGAS-STING pathway are complex and highly dependent on the genetic makeup of the tumor as well as the tumor microenvironment. Recent therapeutic approaches have mostly focused on the induction of the pathway with some promising results that seem to outweigh the negative effects. However, preventing chronic induction of type I interferons by using short, focused bursts of interferon induction might prove valuable in the future in preventing possible side effects.

Similar to cGAS-STING signaling, the function of autophagy in cancer development and therapy is complex. The effect of autophagy on cancer is strongly dependent on the stage of the disease. Initially, autophagy is involved in the control and prevention of malignant transformation. Since autophagy mostly acts in a cytoprotective fashion, this also applies to cells that have already turned to malignancy, providing them with multiple tools to avoid detection and destruction (1, 19, 228).

Cancer cells are part of the tumor microenvironment (TME), which includes non-cancerous cells present in the tumor as well as surrounding stroma and various cytokines and chemokines that are secreted. The TME generally tends to promote cancer progression as well as therapy resistance (255, 256). One of the players shaping the TME and its tumorigenic effects is hypoxia. Due to their fast proliferation, tumors can rapidly outgrow their vasculature leading to low oxygen levels. To compensate for their lack of oxygen, cells utilize hypoxic signaling mediated mainly by HIF1 (257). HIF1 is a transcription factor that can bind to hypoxia response elements in promoters, which leads to the transcription of genes involved in cell survival, proliferation and autophagy (258–260). HIF1 is a heterodimer that consists of HIF1α and HIF1β. Both subunits are constitutively expressed, however, during normoxia HIF1α is polyubiquitinated by the VHL complex mediating its degradation by the proteasome. Upon induction of hypoxia, HIF1α is no longer polyubiquitinated and can form a functional heterodimer with HIF1β, allowing it to translocate to the nucleus and induce transcription (14, 261).

BCL2-interacting protein 3 (BINP3) and BCL2-interacting protein 3-like (BINP3L) are induced by HIF1. Under normoxia, B-cell lymphoma 2 (Bcl-2) forms low-affinity complexes with Beclin-1, thereby decreasing the overall rate of autophagy. Under low oxygen, BINP3 and BNIP3L are upregulated, bind to Bcl-2 in place of Beclin-1, freeing it from the complex and leading to higher rates of autophagy (262).

Another way by which hypoxia can lead to the induction of autophagy is via ER stress followed by the unfolded protein response (UPR). Prolonged hypoxia leads to ER stress, which can be sensed by protein kinase R-like endoplasmatic reticulum kinase (PERK) (263). PERK induces the translation of ATF4, which induces the expression of the transcription factor C/EBP homologous protein (CHOP) (264). CHOP regulates the expression of autophagy genes and the switching between autophagy and apoptosis (265).

Since severe hypoxia also goes hand in hand with lack of nutrients, AMPK, which is one of the main regulators of metabolic stress, is often activated under hypoxia and can regulate autophagy directly by phosphorylation of ULK1 as well as indirectly by inactivation of mTOR (266).

Given that cGAS-STING signaling is intricately connected with autophagy, the following section will provide insights into how autophagy affects cGAS-STING signaling in the context of cancer.

In breast cancer cell lines, irradiation-induced Bcl-2-associated X (BAX)-dependent permeabilization of the mitochondrial membranes results in the release of mtDNA into the cytosol, thus increasing cGAS-STING dependent type I IFN production. Autophagy negatively regulates this effect by clearing cytosolic dsDNA and thereby promoting radio-resistance. Yamazaki et al. further identified cytosolic DNA originating from mitochondria as the main driver of the abscopal response, a phenomenon whereby distant tumor sites are affected by irradiation of the main malignant lesion. In autophagy-deficient mice, irradiation shows a strong induction of type I interferons, which are responsible for the regression of abscopal tumors. The authors linked this to clinical outcomes in breast cancer patients; autophagy rates inversely correlate with the amount of mtDNA and type I IFN response and patients with higher amounts of mtDNA show improved disease-free survival and overall survival (205).

In esophageal squamous cell carcinoma (ESCC), however, disruption of mitochondrial biogenesis through transcription factor A, mitochondrial (TFAM) downregulation strongly induces autophagy and promotes ESCC cancer growth, in line with the poor overall survival rate of ESCC patients with low expression of TFAM. Mechanistically, TFAM downregulation promotes the release of mtDNA, thereby promoting non-canonical autophagy through the activation of the cGAS-STING pathway. In TFAM-deficient ESCC cells, ISGs are significantly downregulated supporting the idea that STING degradation through autophagy attenuates the expression of type I IFN genes, thereby promoting cancer cell progression (267). Blockage of autophagy using the late-stage autophagy inhibitor bafilomycin A1 was shown to reduce STING degradation leading to enhanced downstream signaling and improved anti-tumor response (268).

Further building on the inverse relationship between autophagy and cGAS-STING signaling, inhibition of autophagy has been shown to increase the cGAS-STING signaling activity by promoting the accumulation of irradiation-induced dsDNA in the cytosol. Moreover, inhibition of autophagy combined with irradiation induces a potent antitumor effect in PD-L1-deficient lung cancer mouse models via cGAS-STING-mediated T-cell activation, suggesting a possible combination therapy of irradiation, anti-PD-L1 immune checkpoint blockade and autophagy inhibition in patients with lung cancer (218).

Furthermore, inhibition of the deubiquitinase TRAF-binding domain-containing protein (TRABID) that regulates cell division leads to accumulation of micronuclei through defects in cell division and autophagy. Moreover, cGAS is no longer autophagically degraded and can recognize micronuclei generated by mitotic defects, thereby inducing cGAS-STING signaling (269).

In addition to the induction of type I IFNs via phosphorylated IRF3, the activated cGAS-STING signaling cascade also induces the NF-κB signaling pathway, leading to the production of inflammatory cytokines like IL-6 or TNF-α, chemokines such as CXCL10 and cell cycle regulators (104, 270–272). cGAS-STING-dependent IL-6 production contributes to the survival of chromosomally unstable cancers in an IL-6-STAT3 dependent manner, at the same time inhibiting STAT1 signaling, preventing apoptosis and promoting cell survival and growth (226). In addition to cell survival and proliferation, STAT3 has been linked to various other protumorigenic functions such as angiogenesis and immune escape and is often categorized as immunosuppressive (273, 274). Inhibition of hypoxia-induced autophagy was shown to disrupt pSTAT3 signaling through accumulation of p62, which selectively induces pSTAT3 degradation by the proteasome system and thereby restores cytotoxic T-cell mediated killing of tumor cells (273). In triple negative breast cancer cell lines, myc overexpression was shown to downregulate STING signaling and its downstream effectors in a STAT3-dependent manner (274).

In colorectal cancer, IL-6 induces Beclin-1 phosphorylation at Y333 by the JAK2 kinase, which increases Beclin-1 affinity for VPS34. This enhances autophagic rates and promotes chemoresistance in a STAT3-independent manner. Patients with Beclin-1 Y333 phosphorylation also display increased chemotherapy resistance and poor prognosis (275). Recently, it was shown that IL-6-induced autophagy can also contribute to cancer cachexia, a condition characterized by severe loss of body weight and increased muscle wasting and is associated with an increased risk of cancer treatment failure and poor overall survival. Mechanistically, cancer cells capable of inducing cancer cachexia are secreting IL-6, which subsequently induces autophagy in differentiated muscle cells via IL-6 signaling and therefore might be relevant for the characteristic muscle and weight loss during cancer cachexia (276).