Canonical and noncanonical autophagy in immunity

Summary

Canonical and noncanonical autophagic processes are integrated with innate and adaptive immunity and sterile or pathogen-induced inflammation. In canonical autophagy, double membrane autophagosomes modified by ubiquitin-like ATG8 proteins in a process termed membrane atg8ylation sequester and eliminate intracellular targets such as invading microbes, defunct organelles, aggregates, and inflammatory molecules. Recently, a plethora of noncanonical processes that entail membrane atg8ylation of various intracellular organelles other than autophagosomes have been linked to immunity. This has led to confounding interpretations and conflation of diverse processes as autophagy. Here, we posit that these are divergent manifestations of a common ancestral homeostatic process of membrane atg8ylation and provide an overview of how they affect immunity and inflammation. These relationships are evident in model organisms and are reflected in human genetic predispositions to diseases with immune components. The membrane atg8ylation pathways affect acute and chronic inflammation, infections, autoimmunity, cancer, neurodegeneration, metabolic syndrome, diabetes, and other disorders.

eTOC blurb

Recent studies have expanded the field of autophagy in immunity to include related but incompletely understood noncanonical autophagy processes. Deretic reviews these advancements in the context of immunity offering a unifying framework for canonical and noncanonical autophagy processes in the form of membrane atg8ylation as an overarching principle affecting multiple immunological processes.

Introduction

Canonical autophagy is a conserved intracellular pathway in eucaryotes controlled by a set of the core autophagy factors termed ATGs¹. It entails the formation in the cytoplasm of double membrane organelles termed autophagosomes, generated from several donor intracellular membranes serving as lipid bilayer and phospholipid sources¹. Autophagosomes are purposed to enwrap the cargo recognized by selective autophagy receptors thus sequestering the cytoplasmic targets destined for elimination or recycling¹. The biological outputs of canonical autophagy include immune responses². The canonical autophagy pathway depends on a number of modular systems^1,3 and includes an evolutionarily highly conserved cycle of membrane modifications with mammalian ubiquitin-like ATG8 proteins (mATG8s)⁴. This cycle (Fig. 1A) is referred to as membrane atg8ylation and deatg8ylation, similar to the ubiquitylation-deubiquitylation system⁵. It is driven by the mATG8s protein-lipid conjugation machinery⁴, and is historically associated with the term ‘LC3 lipidation’ emblematic of both canonical and noncanonical autophagic processes.

Fig. 1. Canonical and noncanonical autophagic pathways and their relationship to immune processes.

A. Membrane atg8ylation-deatg8ylation cycle and its components. PE, phosphatidyletholamine, PS, phosphatidylserine. B. Classification of membrane atg8ylation processes with immunological outputs: ACVA (ATG16L1-control of V-ATPase), AMLaS (atg8ylation of multi-lamellar membranous structures), CASM (conjugation of ATG8 to single membranes), EVAC (ER-phagy mediated by the V-ATPase-ATG16L1-LC3C axis), HALD (hemilayer atg8ylation of lipid droplets), LAM (LC3-associated micropinocytosis), LAP (LC3-associated phagocytosis), LANDO (LC3-associated endocytosis), STIL (sphingomyelin-TECPR1-induced LC3 lipidation), and VAIL (V-ATPase-ATG16L1-induced LC3 lipidation). Insets, single membrane lipid bilayers as substrates for membrane atg8ylation in all processes except HALD (where it is a half-bilayer). SM, sphingomyelin. Boxed drawings, immunological outputs. Note that not all molecular details are given and the icons intend to present processes in a stylized manner. Details in the text.

The canonical autophagy pathway affects many aspects of human health⁶ and has strong associations with immune and inflammatory processes^2,7,8. The immune manifestations of canonical autophagy span all levels², from cell-autonomous defense against intracellular microbes^9,10, innate immunity², and immunometabolism¹¹, to development and function of the adaptive immune systems⁷. In parallel to the traditional focus on the canonical autophagy pathway in immunity^7,9,10, a major development in the field has been an emerging understanding that a diverse collection of noncanonical processes engage parts but not all components of the canonical autophagy apparatus^12–15 (Fig. 1B). Whereas canonical autophagosomes classically appear as specialized double membranes enveloping the cargo, noncanonical autophagic processes are different, occur on a variety of pre-existing intracellular organelles, and often do not capture cargo but govern other homeostatic processes such as membrane repair or remodeling. The term ‘autophagic processes’ will be used here to refer collectively to these phenomena (Fig. 1B(i–xii)).

The intent of this review is to update the established immunological roles of canonical autophagy^2,7,8,16,17 and highlight the emerging roles in inflammation and immunity of noncanonical autophagy. The principle of membrane atg8ylation, which unifies all autophagic processes, is a conceptual shift relative to prior reviews of the topic. This framework should be of use to novices and experts alike, especially in the field of immunology, when confronted with a growing variety of acronyms referring to diverse autophagic processes.

Canonical autophagy and its components are linked to immunity

Morphologically, canonical autophagy manifests itself as cytoplasmic membrane crescents (Fig. 1B(i)) termed phagophores which become double membrane autophagosomes upon closure¹. Autophagosomes sequester diverse cytoplasmic targets destined for degradation in autolysosomes with typical cargo being protein aggregates and damaged or surplus organelles¹, invading microbes^9,10,18,19, pathogen-associated patterns (PAMPs), danger associated patterns (DAMPs), and pattern recognition receptors (PRRs)²⁰. Canonical autophagy of mitochondria (mitophagy) maintains and adjusts intracellular pools of these metabolic factories and immune signaling platforms²¹. Autophagy’s immunological outputs include immunometabolism and differentiation of immune cells^11,22,23 (Fig. 1B(i), left), cell autonomous defense against microbes^18,19 (Fig. 1B(i), center), downregulation of acute inflammation²⁰, and, in some cases, establishment of chronic inflammation⁸ (Fig 1B(i), right).

The factors that define the canonical autophagy apparatus¹ can be classified in three principal protein complexes, each with genetic links to inflammatory diseases and immunity: The ULK1(ATG1)-RB1CC1(FIP200)-ATG13-ATG101 complex²⁴ transmits signals from the protein kinases mTOR and AMPK connecting autophagy to immunometabolism and impacts immune cell polarization¹¹. Genetic polymorphisms in ULK1, are associated with severe disseminated tuberculosis²⁵ whereas rare variants in RB1CC1 occur in patients with critical COVID-19 pneumonia²⁶ (Fig. 2(i)). The PI3KC3-C1 complex, class III phosphatidylinositol-3 kinase complex 1²⁷ contains BECN1 (ATG6; Beclin 1)²⁸, the first identified mammalian autophagy factor acting in anti-viral defense²⁸. A polymorphism in BECN1 predisposes to chronic hepatitis B virus infections²⁹ (Fig. 2(i)). The ATG2A/B-ATG9A complex³⁰ acts in lipid transfer³¹ and lipid scrambling³² during phagophore expansion. Polymorphisms in ATG2A³³ are associated with granulomas in Crohn’s disease (CD), a common form of inflammatory bowel disease (Fig. 2(ii)). Autophagy complexes coalesce²⁴ through PI3KC3-C1 phosphorylation of membrane lipids generating phosphatidylinositol 3-phosphate (PI3P). This brings about PI3P binding proteins WIPI1–4 to support initiation and expansion of the phagophores³⁴. WIPI2 is important in antibacterial autophagy (Fig. 1B(i) center)³⁵. Thus, all components of the canonical autophagy factors have genetic links with immunity and inflammation.

Fig. 2. Canonical and noncanonical autophagic pathways in diseases with immunological components and autophagic processes in immune cells.

A. Human data are genetic (italicized risk loci), treatment based, or other clinical information. B. Role of autophagic processes in different cells of the immune system.

RB1CC1 interacts with sequestosome 1-like receptors (SLRs), such as SQSTM1 (p62)³⁶ and CALCOCO2 (NDP52), to initiate selective engulfment of cytosol-invading bacteria³⁷ (Fig. 1B(i) center). RB1CC1 also interacts with innate immunity TRIM proteins (TRIM20 and TRIM21), to downregulate the NLRP3 inflammasome, and IRF3 to quench the type I IFN responses³⁸ (Fig. 1B(i) right). CALCOCO2 variants associate with autoimmune-inflammatory disorders CD and multiple sclerosis (MS)³⁹. Like SLRs, TRIM20 and TRIM21 assemble the ULK1-complex components plus BECN1³⁸. MEFV, encoding TRIM20 (Pyrin), is a risk factor for familial Mediterranean fever whereas TRIM21 (Ro52/SSA1) is an autoantigen in Sjögren’s syndrome and systemic lupus erythematosus (SLE)³⁸. Hence, selective autophagy receptors have a role in inflammation and autoimmunity.

ATG9A is immunologically important^40,41 and spans and coordinates all stages of autophagosome biogenesis. ATG9A interacts with the ATG13-ATG101 complexes during autophagy initiation⁴², works together with ATG2A during membrane expansion³⁰, and interacts with ESCRT proteins during phagophore closure⁴³ to complete the double membrane autophagosome as a prerequisite for their fusion with lysosomes⁴⁴. ATG9A controls activation of type I IFN via STING^40,41. ATG9A inactivation in myeloid cells causes spontaneous mortality in mice⁴³, and it manifests itself in inflammation via canonical and noncanonical autophagic processes^41,45. Immune disease-associated ATG9A polymorphisms are rare⁴⁶. In contrast to ATG9A, ATG16L1 and its interactor IRGM, are well-known as risk factors associated with CD⁴⁷ (Fig. 2(ii)).

In summary, the canonical autophagy pathway is defined by several protein complexes and polymorphisms in their components confer genetic predispositions to immune disorders in humans.

Membrane atg8ylation is a paradigm unifying autophagic processes

Membrane atg8ylation (Fig. 1A) participates in canonical autophagy (Fig. 1B(i)) but also acts in noncanonical processes^5,13,15 with specific immunological outputs (Fig. 1B(ii–xii)). A simple way to grasp the concept of membrane atg8ylation⁵ is to think of it as an equivalent of protein ubiquitylation. In the case of membrane atg8ylation, phospholipids within cellular membranes are getting conjugated to and modified by mATG8 proteins⁴. Membrane atg8ylation occurs during canonical autophagy⁴ or on stressed membrane during inflammation or infection⁵ (Fig. 1A)..

The ubiquitin and mATG8s are related proteins. Both are activated via dedicated conjugation cascades involving ATP, E1, E2 and E3 ligases resulting respectively in protein ubiquitylation or membrane atg8ylation. There are seven mATG8s in humans: LC3A, LC3B, LC3B2, LC3C, GABARAP, GABARAPL1 and GABARAPL2⁴. The factors carrying out membrane atg8ylation (Fig. 1A) include two enzymatic cascades with ATG12—ATG5 and mATG8—aminophospholipid conjugates as their final products⁴. The ATG12—ATG5 protein-protein conjugate combines with additional proteins, such as ATG16L1 or TECPR1, to form E3 ligases that atg8ylate specific membranes⁴. The E3 ligases are guided to membranes by adaptors such as WIPI2⁴⁸ or the V1 domain of the proton pump V-ATPase^49–51 (Fig. 1A). Polymorphisms in the ATG16L1 locus in human populations are associated with resistance to certain infections (Fig. 2A(i)) in addition to the predispositions to CD (Fig. 2A(ii)). Like ubiquitin precursors that are proteolytically processed, mATG8 precursors are first proteolytically activated by ATG4 proteins to expose the C-terminal glycines on mature mATG8s (Fig. 1A) necessary for their entry into the atg8ylation cycle⁵². Similarly to deubiquitylating enzymes, ATG4 isoforms (ATG4A-D) act as membrane deatg8ylases⁵² (Fig. 1A). Polymorphisms in ATG4A and ATG4B are associated with a more severe CD³³ (Fig. 2A(ii)).

To conclude, membrane atg8ylation of various intracellular membranes has a spectrum of immunological outputs (Fig. 1B), With the exception of lipid droplets (LDs), where atg8ylation occurs on a phospholipid monolayer⁵³, atg8ylation of all other membranes, including autophagic phagophores (Fig. 2B(i), inset), occurs on single lipid bilayers.

Manifestations of membrane atg8ylation affect inflammation and immunity

The classification of processes in Fig. 1B takes into account the classical view of canonical autophagy ending in a fusion of autophagosomes with lysosomes to form degradative autolysosomes (Fig. 1B(i)). Other processes (Fig. 1B(ii–xii)) use membrane atg8ylation and sometimes parts of the canonical autophagy pathway but are often nondegradative.

Canonical autophagy involves membrane atg8ylation (Fig. 1B(i)). Although autophagosomes can form in cells devoid of atg8ylation components, there are kinetic delays⁵⁴ and a handicap in selective autophagy of microbes (‘xenophagy’; Fig. 1B(i) center), a form of canonical autophagy assisted by SLRs in targeting microbes to autophagosomes^19,55. SLRs bind to protein complexes forming canonical autophagosomes^36,37 whereas membrane atg8ylation enhances SLRs binding to mATG8-decorated phagophores.

Secretory autophagy^56,57 (Fig. 1B(ii)) uses components of canonical autophagy, membrane atg8ylation, or mATG8s^57–59 but instead of delivering the cargo for degradation leads to its secretion. When perturbed, this leads to the excessive neutrophil activation and degranulation causing inflammation and mortality in a murine model of tuberculosis⁶⁰. Secretory autophagy can enhance viral spread. The enveloped double stranded DNA (dsDNA) herpesviruses exploit membrane atg8ylation for their egress and immune evasion via extracellular vesicles^61,62,63,64. Poliovirus capsids are nonlytically released via phosphatidylserine-enriched atg8ylated vesicles⁶⁵ classified within the broad category of secretory autophagy. Dengue virus (DENV) uses vesicles in human dendritic cells (DCs) for viral transmission via a form of secretory autophagy⁶⁶.

Microautophagy (Fig. 1B(iii))⁶⁷ entails direct invagination of the endolysosomal membranes involving membrane atg8ylation for selective turnover of lysosomal membrane proteins⁶⁸. Microautophagy affects antigen presentation⁶⁹.

Membrane atg8ylation participates in lysosomal and phagosomal membrane repair (Fig. 1B(iv)) countering damage inflicted by invading microbes or mediators of sterile inflammation^70,71. The mechanisms include ESCRT-mediated repair⁷², lipid transfer^73,74, and recruitment of stress granule proteins⁷⁵ that may act as temporary plugs⁷⁶. These processes contribute to protection against invading microbes, membrane-damaging PAMPs and DAMPs, and associated inflammation⁷¹.

A process termed VAIL (V-ATPase-ATG16L1-induced LC3 lipidation; Fig. 1B(v))^51,77 or CASM (conjugation of ATG8 to single membranes)^50,78 occurs during Salmonella invasion of host cells⁴⁹. Here, V-ATPase acts as an adaptor for the ATG12—ATG5-ATG16L1 E3 ligase complex and triggers membrane atg8ylation on compromised phagosomes and lysosomes^49,50,77. Because all atg8ylated membranes, including the canonical phagophores, are modified as single membranes (Fig. 1B, insets) the term VAIL will be used here. VAIL leads to lysosomal membrane tubulation and recruitment of ATG2A via one of the mATG8s, LC3A⁷³, transferring lipids to damaged membranes³¹. VAIL is countered by the Salmonella effector protein SopF (Fig. 1B(v)), supporting Salmonella intracellular replication and virulence in vivo⁷⁹ (Fig. 3(i)). SopF ADP-ribosylates the V-ATPase V0 domain to block ATG16L1 recruitment and curtail VAIL^49,50,77. Thus, the anti-VAIL action of SopF ensures bacterial fitness in the infected host⁷⁹.

Fig. 3. Canonical and noncanonical autophagic pathways in mouse models of diseases with immunological components.

Comparative analysis of murine models of diseases that have both autophagic and immune/inflammatory components. Three representative selections are shown in each grouping. “1.5”, latent autoimmune diabetes in adults with mixed features of T1D and T2D. Acronyms as in Fig. 1 or explained in the text.

VAIL is induced by microbial factors and perturbations of proton gradients in endolysosomes⁸⁰. Bacterial toxin VacA⁸¹ and influenza viroporin M2⁸² (Fig. 1B(v)) dissipate proton gradients and elicit VAIL. DNA viruses, e.g. HSV-1^83,84 or transfected dsDNA ⁵¹ induce VAIL via STING, otherwise known for its role in type I IFN induction via the cGAS-STING-TBK1-IRF3 pathway^51,83. Whereas dsDNA-induced membrane atg8ylation was initially interpreted as autophagy⁸³ recent studies have shown that this is VAIL induced by STING’s intrinsic proton channel activity⁸⁵.

EVAC (ER-phagy mediated by the V-ATPase-ATG16L1-LC3C axis) (Fig. 1B(vi))⁸⁶, is a variation of VAIL in the secretory pathway, starting at the endoplasmic reticulum (ER)⁸⁶ and ending in the trans-Golgi network⁸⁸. The secretory pathway is important for cytokine secretion⁸⁷ (Fig. 1B(vi)). The specific role of EVAC is in quality control⁸⁶, which affects cytokine secretion.

Two related membrane atg8ylation processes, AMLaS (atg8ylation of multi-lamellar membranous structures) and ACVA (ATG16L1-control of V-ATPase) occur during Mycobacterium tuberculosis (Mtb) infection of macrophages^{89 90} (Fig. 1B(vii) and 3(i)). AMLaS depends on V-ATPase and ATG16L1 and causes cyclical bursts of luminal acidification⁸⁹. ATG16L1 recruited to V-ATPase acts not only in VAIL but also regulates V-ATPase proton pump’s activity via ACVA⁹⁰. ACVA works like a cycle (Figs. 1B(viii)): Following endolysosomal membrane damage, ATG16L1 binds and inhibits V-ATPase to block futile hydrolysis of ATP⁹⁰. Then, membrane atg8ylation via VAIL promotes membrane repair⁷³. Next, mATG8s on atg8ylated membranes bind and recruit DMXL1⁹¹. Finally, DMXL1 enhances the assembly of V-ATPase and restarts the pump restoring luminal acidification. ACVA ensures that luminal acidification occurs only after endolysosomes can maintain H⁺ gradient, lest unnecessarily burnt ATP depletes cellular energy charge⁹⁰.

STIL (sphingomyelin-TECPR1-induced LC3 lipidation)⁸⁰ (Fig. 1B(ix)) occurs on stressed endolysosomal membranes and is different from VAIL. During STIL, TECPR1 substitutes for ATG16L1 within the atg8ylation E3 ligase (Fig. 1A). Normally, sphingomyelin in membranes is not exposed to the cytosol. It flips from the luminal to the cytosolic side of the stressed vacuoles harboring bacteria where TECPR1 recognizes it before overt membrane damage^92,93. The effector mechanisms downstream of STIL are ESCRT-dependent membrane repair⁷² and PI4P-dependent tubulation that removes the damage⁹⁴.

LAP (LC3-associated phagocytosis)⁹⁵, LAM (LC3-associated micropinocytosis)⁹⁶, and LANDO (LC3-associated endocytosis)⁹⁷ are related processes (Fig. 1B(x,xi)). In LAP, LC3 is conjugated to phagosomes¹³. LAP is induced in macrophages via toll-like receptors (TLR) 2⁹⁵ and 4⁹⁸. LAP is not directly connected to membrane damage and thus differs from VAIL but depends on the same E3 ligase ATG12—ATG5-ATG16L1. LAP employs RBCN (Rubicon) and NADPH oxidase (NOX2)⁹⁹ and is triggered by NOX2 products⁹⁹. NOX2 neutralizes phagosomal lumen by generating superoxide, which soaks up luminal H⁺ while dismutating into H₂O₂ thus deacidifying phagosomes and triggering LAP⁸⁰. Hence, while the signals are different, VAIL and LAP are nonetheless similar. LAP helps eliminate phagocytosed microbes⁹⁵ and stimulates type I IFN in plasmacytoid dendritic cells¹⁰⁰. LAP’s absence in myeloid cells triggers type I IFN signaling in tumor associated macrophages¹⁰¹ (Fig. 2B). LAP enhances removal of dying cells by phagocytes, which fails in the absence of LAP leading to inflammation¹⁰². LAP enhances MHC II presentation of phagocytosed antigens⁶⁹ (Fig. 2B). Thus, LAP affects immune outputs associated with phagocytosis and endocytosis.

Mobilization of neutral lipid stores from lipid droplets (LDs) affects neutrophil differentiation¹⁰⁵ (Fig. 2B) and occurs via canonical autophagy (lipophagy)¹⁰³ or noncanonical processes¹⁰⁴ including hemilayer atg8ylation of lipid droplets⁵³ (HALD; Fig. 1B(xii)). LDs also act as innate immune hubs and are enriched for innate immunity proteins¹⁰⁶. Among such proteins¹⁰⁶ is Irgm1, the murine ortholog of human IRGM, implicated in CD⁴⁷ and in control of infections^107,108. LDs are also sites for assembly of flaviviruses including hepatitis C virus, DENV, Zika virus, West Nile Virus, and Japanese encephalitis virus¹⁰⁹ (Fig. 1B(xii)). LD manipulation by SARS-CoV-2 aids viral replication and immune evasion¹¹⁰ (Fig. 1B(xii)).

In summary, diverse manifestations of membrane atg8ylation exceed the province of canonical autophagy and affect a variety of immune processes.

Xenophagy protects cells against microbes

The cell-autonomous defense roles of autophagic processes have been validated using viral^9,111 and bacterial¹⁰ pathogens in cellular and animal models. Several seminal works have contributed to the initial recognition of autophagy as an antimicrobial mechanism including the identification of the PI3KC3-C1 component BECN1 as an anti-viral factor²⁸ and studies showing that autophagy controls intracellular pathogens including Mtb¹⁰⁷, Shigella¹¹², Salmonella¹¹³ and invasive extracellular bacteria¹¹⁴.

In xenophagy, microbes are tagged by ubiquitin for capture by autophagosomes^10,18,19,115 (Fig. 1B(i, center)). This includes ubiquitylation of the host proteins^18,19, ubiquitylation of bacterial proteins inserted into host vacuolar membranes¹¹⁶, binding of ubiquitin to bacterial proteins¹¹⁷, and direct ubiquitylation of microbial surface components¹¹⁸. Ubiquitylated targets are then bound by SLRs^18,19: SQSTM1, NBR1, CALCOCO2, TAX1BP1, and OPTN. One end of SLRs binds ubiquitin-decorated microbes and the other binds atg8ylated phagophores^18,19. Ruptured or perturbed vacuoles are additionally decorated with galectins which bind to exposed luminal glycans and are targeted to the phagophores by CALCOCO2 and its cognate partner Galectin 8¹¹⁹, TAX1BP1 and Galectin 8¹²⁰, or TRIM16 and its partner Galectin 3¹²¹.

SLRs also restrict viruses such as HSV-1¹²² and poxviruses, specifically the Vaccinia virus¹²³. HSV-1 infection triggers TRIM23-enhanced cGAS-STING membrane atg8ylation¹²⁴, most likely VAIL^83,85, defending against HSV-1¹²⁴. The ATG12–ATG5-ATG16L1 E3 ligase cooperates with non-autophagic effectors to restrict Norovirus replication in a murine model¹²⁵. M2 protein of the influenza virus directly induces membrane atg8ylation⁸² to support proper viral biogenesis. Thus, membrane atg8ylation either helps the host or the pathogen.

Xenophagy closely resembles autophagy of mitochondria (mitophagy²¹) down to the use of the same SLRs¹⁸. This has led to a postulate that xenophagy and mitophagy have common evolutionary roots given that mitochondria have evolved from a Rickettsia-like endosymbiont^18,21,126.

Membrane atg8ylation processes are linked to innate immunity signaling

PAMPs and DAMPs induce membrane atg8ylation^{127 20}. DNA released from pathogens, such as Mtb, into the cytosol activates the cGAS-STING cascade^{128 129,130}. STING induces VAIL⁸⁵ in addition or instead of the canonical autophagy. It also activates type I IFN responses¹³¹. Mtb RNA is released into the infected macrophages and stimulates type I IFN production via RIG-I¹³² and MDA5 with concomitant membrane atg8ylation¹³³. These systems elevate type I IFN, which is pathogenic since Mavs^{−/− 132} and Irf3^{−/− 131} animals show improved survival compared to WT mice infected with Mtb and matches the type of interferon signature found in patients with active tuberculosis¹³⁴. Xenophagy and type I IFN responses are also connected via OPTN, an SLR important for intracellular control of Salmonella¹³⁵, which is required for optimal type I IFN response via IRF3 but not for NF-kB signaling¹³⁶. In summary, membrane atg8ylation, xenophagy and type I IFN activation are intertwined.

Encounters with viruses elicit innate immunity responses via PRR signaling including autophagic processes⁹. Viral PAMPs are recognized by cognate PRRs, e.g. dsRNA by TLR3, RLRs and PKR, ssRNA by TLR7/8, DNA by TLR9 and cGAS-STING, or viral capsids by TRIM5α¹³⁷. These systems activate membrane atg8ylation^85,95,127 concomitant with proinflammatory signaling.

In conclusion, xenophagy and other manifestations of membrane atg8ylation are interlaced with type I IFN and other inflammatory responses that can be protective or pathogenic.

In vivo studies reveal roles of autophagy and atg8ylation in infection and inflammation

Whereas the in cellulo studies support cell-autonomous defense functions of xenophagy, in vivo studies in mouse models of infection are more complex and xenophagy is often masked by inflammatory effects (Fig. 3(i)). For example, intestinal epithelial Atg5 is required to prevent dissemination of invasive bacteria to extraintestinal sites¹³⁸. However, whole body Optn KO mice¹³⁶, with Optn reflecting xenophagy¹³⁵, subjected to Salmonella infection show no changes in microbial loads although displaying increase in mortality¹³⁶ (Fig. 3(i)). Genetic ablation of the membrane atg8ylation apparatus (Atg5, Atg7, Atg16L1)^139,140 has been reported to increase Mtb bacterial burden but this has not been observed in studies by others^141–143. Nevertheless, membrane atg8ylation¹⁴⁰, LAP¹⁴⁴, ACVA⁹⁰, canonical autophagy¹⁴², secretory autophagy⁶⁰ and unique roles of Atg5^60,145, show clear impact in mouse survival studies of Mtb infection (Fig. 3(i)). In a model of Sindbis alphavirus CNS infection, no change in viral replication is observed in the absence of atg8ylation-competent Atg5 although mice display increased mortality ascribed to neuronal cell death¹⁴⁶. Similar findings have been observed using engineered Becn1 variants²⁸. In a mouse model of Toxoplasma gondii infection, loss of Atg5 in myeloid cells results in increased numbers of parasites in addition to increased lethality¹⁴⁷ (Fig. 3(i)). Thus, in vivo studies in mice disabled for membrane atg8ylation show inflammatory repercussions and tissue damaging consequences but do not always result in increased microbial burdens.

Another aspect of studies in murine models is that the early publications have mostly employed the genes of the membrane atg8ylation cycle. Atg5¹⁴⁷, Atg3, Atg7, as well as the E3 ligase Atg12-Atg5-Atg16L1¹⁴⁸ are necessary to control Toxoplasma gondii. The Atg16l1 alleles mimicking human polymorphisms associated with CD⁴⁷ have been shown to affect inflammation in mice¹⁴⁹. Even contemporary studies often rely on a single atg8ylation gene inactivated in mice or its polymorphisms in humans: a human ATG16L1^rs6861 polymorphism protects again HIV-1 in untreated patients¹⁵⁰; deletion of Atg5 in T cells is associated with liver fibrosis in mice and increased T helper-17 (Th17) cells¹⁵¹. The above genes are responsible for membrane atg8ylation in general and cannot be used to distinguish canonical autophagy from other immunological outputs of membrane atg8ylation (Fig. 1).

One incidental finding from in vivo modeling is that inactivation of canonical autophagy or membrane atg8ylation genes heightens baseline inflammatory processes paradoxically conferring protection against certain microbes tested in murine models. Compensatory mechanisms in myeloid cells lacking Atg16l1 enhance elimination of Shigella flexneri in mice¹⁵². Loss of canonical autophagy genes but not membrane atg8ylation genes creates a hyperimmune state against Listeria monocytogenes¹⁵³. Loss of membrane atg8ylation genes or canonical autophagy-specific genes suppresses herpes reactivation in mice by artificially generating heightened inflammatory response¹⁵⁴. Such studies suggest that the anti-inflammatory function is a key activity of autophagic processes in vivo².

Autophagic processes affect both acute and chronic inflammation

Canonical autophagy downregulates intracellular inflammatory hubs. ATG16L1 and IRGM associate with diverse PRRs¹⁵⁵, leading to a selective autophagy of NLRP3-ASC¹⁵⁶ and signaling complexes activating RIPK2 and NF-kB¹⁵⁷. Type I IFN regulatory systems including cGAS-STING and RIG-I-MAVS are subject to SQSTM1-dependent autophagic degradation¹⁵⁸. The TRIM class of PRRs control inflammation via selective autophagy of inflammasomes and TRIF¹⁵⁹. ATG9A-dependent processes can diminish STING activation and inflammation in the skin⁴¹.

In a recent study¹⁶⁰, an interesting dichotomy in virally induced inflammation in bystander vs. infected cells has been reported. SARS-CoV-2 suppresses proinflammatory output in infected cells via OPTN- and ATG9A-dependent mitophagy (MAVS resides mostly on mitochondria), protecting the virus¹⁶⁰. However, this does not occur in uninfected bystander cells which exuberantly produce inflammatory mediators¹⁶⁰. This may explain why fulminant viral infections can lead to patient death even when in clinical care.

A recent study has shown that autophagy can also promote chronic sterile inflammation⁸. The study shows by inactivating both membrane atg8ylation gene ATG7 and canonical autophagy genes ATG13 and RB1CC1 in human cells and their counterparts in model organisms that canonical autophagy can promote chronic inflammation. This occurs in certain states by degrading a subunit (WSTF) of chromatin remodeling complexes thus opening chromatin over the pro-inflammatory genes⁸ and increasing expression of inflammatory mediators in senescence, cancer, steatohepatitis, and osteoarthritis.

In summary, autophagy balances inflammation during infection providing sufficient duration of protective inflammation while limiting it by elimination of cytoplasmic inflammatory signaling hubs. In contrast, a drawn-out response associated with chronic diseases leads to autophagy-assisted chronic inflammation.

Autophagic processes affect immune cells and their functions

The effects of autophagy and the ATG genes in specialized immune cells (Fig. 2B) come in two flavors. The first one is the homeostatic effects in immune cells just like in any other cell^1,7,11. The other is more specialized for immune functions^2,161. In principle, canonical autophagy works as a metabolic modulator in immune cells whereas multiple autophagic processes affect immune cell effector functions, homeostasis, proinflammatory and inflammation-resolution phenotypes^2,11.

Autophagic processes regulate hematopoietic stem cell (HSC) development and differentiation, including emergence of HSCs from hemogenic endothelial cells in embryos¹⁶², pre-HSC T1 and T2 pre-HSC stages¹⁶³, adult HSCs^11,164, and aging HSCs affecting longevity of the HSC compartment¹⁶⁵. The hematopoietic lineage branches are differentially affected by the Atg genes. Early loss of Atg7 has a lopsided effect, suppressing the lymphoid lineage but favoring myeloproliferation. Myeloid cells (Fig. 2B; macrophages and neutrophils) in Atg mutants maintain numbers but show hyperinflammatory phenotype^60,143.

HSC-derived monocytes and fetal-derived tissue resident macrophages¹⁶⁶ are the principal amplifiers of early inflammatory responses via IL-6, TNF-α, and IL-1β and by recruiting other immune cells including neutrophils. During inflammation-resolution stages autophagic processes in macrophages influence clearance of phagocytosed microbes, antigen presentation, and noninflammatory clearance of apoptotic cells by efferocytosis¹⁰² (Fig. 2B). Modifications of mitochondrial proteins and mitochondrial function in macrophages promote resolution of inflammation¹⁶⁷ with mitochondria being a major target for autophagic homeostasis²¹.

The antimicrobial effects in macrophages¹⁶⁶ involve noncanonical membrane atg8ylation including LAP^{95 144,168}. LAP can slow down phagosome maturation to allow prolonged MHC class II antigen presentation¹⁶⁹ (Fig, 2B). AMLaS and ACVA (Figs. 1B(vii,viii) and 2B) are important for control of Mtb in macrophages^89,90. Downregulation of Atg5 in Kupffer cells is protective and leads to reduction of liver inflammation and bile duct damage¹⁷⁰. LANDO in brain microglia recycles TREM2 and other receptors for Aβ and prevents inflammation in a mouse model of age-induced Alzheimer’s disease^{97 171} (Figs. 1B(xi) and 2B). LANDO is defective in Atg16L^E230 (Atg16^ΔWD) mice, which display neuroinflammation with increased IL-1β, IL-6, and TNF-α in the hippocampus, along with β amyloid deposition, Tau phosphorylation, and memory decline¹⁷¹ (Fig. 1B(xi)). Incidentally, the mouse used in these studies is the same whole body Atg16l1^E230 mouse^171,172 that quickly succumbs to low dose Mtb infection and shows defective ACVA, LAP and other membrane atg8ylation processes which depend on ATG16L1’s interactions with V-ATPase⁹⁰ (Fig. 3(i)).

Bone marrow derived neutrophils from Atg5^fl/fl Lyz2-Cre mice show excessive activation and degranulation⁶⁰ (Fig. 2B). Hyperinflammatory myeloid cells in Atg5^fl/fl Lyz2-Cre mice have been validated in infectious models with bacteria^141,153,173 and viruses using a panel of Atg mutants^154,174. These activities have been ascribed to canonical autophagy whereas more recently are being linked to noncanonical functions. Defective lipophagy in Atg7-deficient neutrophil progenitors show differentiation defects associated with increased glycolytic activity and impaired fatty acid oxidation and accumulation of LDs¹⁰⁵ (Fig. 2B). Neutrophils can elaborate extracellular LC3⁺ structures to coat bacteria contributing to antimicrobial action¹⁷⁵. LAP can occur in neutrophils, but in the case of Staphylococcus aureus it paradoxically provides a survival niche for the bacterium¹⁷⁶ (Fig. 2B).

DCs (Fig. 2B) are specialized antigen-presenting cells instructing naïve T cells to provide protective immunity or establish immune tolerance. DCs can be classical DCs type I and type II derived from the common DC precursor in the bone marrow, plasmacytoid DCs known for rapid type I IFN production affected by Atg5 and autophagic processes¹⁷⁷ (Fig. 2B), myeloid DCs, and Langerhans cells, with multitude of functions including antigen uptake, migration to lymph nodes, maturation, and antigen presentation resulting in adaptive immunity, tolerance, allergies, or innate immune modulation¹⁷⁸. Atg5 in DCs (and lung macrophages) is needed for suppression of Th17 cell responses and neutrophil accumulation in a mouse asthma model¹⁷⁹ (Fig. 2B). Atg5 in DCs is required to stabilize regulatory T (Treg) cell expression. In Atg5^fl/fl CD11-Cre mice, the Treg cells are dysfunctional, unstable, and exacerbate pathology in a mouse model of rheumatoid arthritis (RA)¹⁸⁰ (Fig. 2B).

During T cell development and maturation (Fig. 2B) in the thymus, autophagy in medullary thymic epithelial cells plays a role in negative selection of self-peptide reactive T cells thus generating a self-tolerant T cell repertoire^181,182. In contrast to central tolerance, autophagic processes in peripheral tolerance are considered to feed autoantigens to processing compartments in DCs¹⁸³ (Fig. 2B). To counter this, Treg cells^184,185 inhibit autophagy-driven antigen presentation in DCs in a CTLA-4-dependent manner and prevent them from priming autoantigen-specific CD4 T cells to avoid autoimmunity^186,187 (Fig. 2B). Of further interest is the Treg cell-intrinsic autophagy, as it appears to be important for cell development and loss of autophagy in these cells diminishes T_reg suppressive action^188,189 (Fig. 2B).

Autophagy is necessary for survival upon T cell exit from the thymus once IL-7 becomes limited and reduction of mitochondrial content as thymocytes transition to mature T cells¹⁹⁰. Processes controlled by Atg5 are important, as Atg5^−/− CD8⁺ T lymphocytes display increased cell death. Atg5^−/− CD4⁺ and Atg5^−/− CD8⁺ T cells also have a proliferation defect after TCR stimulation¹⁹¹. Autophagy influences immunometabolic changes in CD8⁺ T cells. TCR engagement and stimulation with inflammatory cytokines inhibits autophagy in CD8⁺ T cells concomitantly with increased uptake of amino acids²³. When autophagy is active, naïve CD8⁺ T cells undergo mitochondrial pruning and enhanced survival, whereas in cytotoxic T cells, autophagy tends to degrade cytolytic effectors and nutrient transporters, thus rebalancing CD8 functions²³ (Fig. 2B). Furthermore, autophagy regulates mitochondrial inheritance in CD8⁺ T cells affecting memory potential²². Mice lacking the atg8ylation genes Atg5 or Atg7 in T cells fail to establish CD8⁺ T cell memory¹⁹² (Fig. 2B).

In humans, ATG16L1 polymorphism (rs6861; TT) is associated with stronger T cell effector responses and decreased T-cell exhaustion contributing to better control of HIV-infection¹⁵⁰ (Fig. 2A(i)). Another polymorphism in ATG16L1 (rs2241880; T300A, makes it sensitive to caspase cleavage at an adjacent site. Removing the WD domain (absent in yeast but present in mammalian ATG16L1) downstream of the caspase site increases Th1 and Th17 cell responses and decreases Treg cells¹⁹³. ATG16L1 WD domain loss has been interpreted as a defect in canonical autophagy, however recent studies^{90,171,172,194} in mice homozygous for Atg16l1^E230 truncation (losing an even longer C-terminal segment than the T300A allele) demonstrate that canonical autophagy is unperturbed when ATG16L1 lacks its C-terminal domain suggesting non-canonical effects of the ATG16L1 WD domain loss.

In CD4⁺ T cells, loss of Atg5 leads to pathology-promoting IL-17A⁺ IFN-γ⁺ Th17 T cells that drive liver fibrosis in mice via proinflammatory activities in macrophages¹⁵¹. Autophagy is important for invariant NKT (iNKT) cell development from thymic precursors and IL-15 activates autophagy for survival of iNKT cells¹⁹⁵.

Atg5 plays a role in development of B cells (Fig. 2B) although a significant portion of Atg5^−/− pre-B cells survive and populate the periphery¹⁹⁶. A B1 CD5⁺ lineage depends on Atg5 for development and maintenance¹⁹⁶. The B1 CD5⁺ cell is a self-renewing tissue-resident innate type of B cell with a capacity to rapidly respond in a T cell independent manner to a limited number of epitopes. B1 CD5⁺ cells acquire exogenous fatty acids stored in LDs and show high rates of both glycolysis and oxidative phosphorylation whereas Atg7 deletion leads to B1 cell loss, ascribed to defective mitochondrial functions¹⁹⁷. Autophagic processes, which control the ER content and quality, are required for immunoglobulin production by plasma cell¹⁹⁸, affect humoral responses¹⁹⁹ and are also of consequence in mucosal IgA responses²⁰⁰ (Fig. 2B).

In conclusion, autophagic processes contribute to the development, differentiation, polarization, function, immunometabolism, and homeostasis of diverse immune cell subsets (Figure 2B).

Autophagic processes affect antigen presentation

Classically, MHC class I molecules present peptides to CD8⁺ T cells from endogenous, i.e. cell’s own, proteins, whereas MHC class II molecules present peptides to CD4⁺ T cells from exogenous sources and cell’s own membrane antigens. When MHC class I molecules present exogenous peptides to CD8⁺ T cells it is referred to as cross-presentation. When MHC class II present peptides from endogenous nonmembrane sources to CD4⁺ T cells it is referred to as a nonclassical or endogenous MHC class II pathway. Autophagy has been implicated in all the above antigen presentation processes^201–203. The success of yellow fever vaccine has been attributed in part to autophagic enhancement of antigen presentation to both CD4⁺ and CD8⁺ T cells²⁰⁴.

Autophagy contributes to antigen presentation¹⁸³, with early studies relying primarily on downregulation of the atg8ylation genes such as Atg12 or Atg5. A recent study has confirmed that modulation of autophagic activity affects MHC-II presentation and that its dysregulation leads to higher production of proinflammatory cytokines from T cells²⁰⁵. Both HSV-1 and CMV encode factors that block transporter associated with antigen processing TAP to inhibit conventional MHC class I peptide loading in the ER, whereas autophagy has been reported to bypass that block and augment MHC class I presentation of HSV-1 antigens²⁰⁶. Similarly, an autophagic pathway promotes CMV antigen presentation to CD8⁺ T cells²⁰⁷.

In nonclassical (endogenous) MHC class II pathway, experiments employing pharmacological modulators of canonical autophagy affected presentation of cellular²⁰⁸ and viral proteins (e.g. Epstein-Barr virus EBNA)²⁰⁹. The selective autophagy receptor TAX1BP1 transports cytosolic antigens into the MHC class II loading compartments during endogenous MHC class II presentation²¹⁰. Autophagy’s role in endogenous MHC class II presentation extends to nuclear antigens during aging²¹¹ leading to antinuclear autoantibodies of import for age-related immunopathologies and autoimmune diseases.

Whereas early studies have argued against the role of autophagy in cross-presentation, atg8ylated structures delivering proteasomes into the endosomal lumen for cross-presentation have been recently described²¹², differing from cytosolic proteasome action.

In conclusion, autophagic processes contribute to the major modalities of antigen presentation (Figure 2B).

Autophagic processes modulate immunological aspects of disease

The investigations into canonical autophagy within a wide spectrum of diseases have been comprehensively reviewed⁶. Here, we present vignettes focusing exclusively on the intersections between immunity, autophagic processes, and disease.

The autophagic pathways affect autoimmune and chronic inflammatory diseases such as CD, systemic lupus erythematosus (SLE), MS, rheumatoid arthritis (RA), ulcerative colitis, asthma, celiac disease, and familial Mediterranean fever² (Fig. 2(ii)). Autophagy contributes to central tolerance¹⁸¹ whereas autophagic processes in peripheral tolerance act in the opposite direction¹⁸³.

CD is characterized by aberrant immune processes in the intestines. Autophagy-related proteins IRGM and ATG16L1 interact with NOD2¹⁵⁵, a PRR mutated in familial CD, with all three factors being genetically linked to CD⁴⁷ (Fig. 2A(ii)). The ATG16L1^T300A allele occurs at a surprisingly high frequency in human populations whereby T300A substitution renders ATG16L1 susceptible to cleavage by caspase at an adjacent site²¹³ removing ATG16L1’s V-ATPase-interacting C-terminal domain^49,50,90. Several manifestations of membrane atg8ylation depend on this C-terminal domain of ATG16L1 (Fig. 1B; VAIL/CASM, ACVA, LAP and EVAC) and may be incapacitated by elevated caspase activity in inflammatory sites. Recent mouse infection modeling using Atg16l1^T316A, the murine equivalent of the human T300A mutation, suggests that a heterozygous state Atg16l1^T316/WT leads to protection against bacterial gastroenteritis due to heightened inflammatory defenses whereas the Atg16l1^T316A/T316A homozygous mice are more susceptible to infection (Fig. 3(ii)) but not due to defective canonical autophagy²¹⁴. It is possible that gastrointestinal infections in human populations have driven selection for heterozygous Atg16l1^T300A/WT state providing elevated acute inflammation to fend off life threatening intestinal pathogens with a collateral cost of tissue damage in CD.

SLE is a type I interferonopathy characterized by autoantibodies against nucleic acids and chromatin components, immune complex deposition, and T cell dysregulation, with patients suffering from multi-organ complications²¹⁵. SLE risk polymorphisms include membrane atg8ylation genes ATG5^216,217, ATG7²¹⁶, as well as autophagic processes-related genes LRRK2²¹⁸ and IRGM²¹⁶ (Fig. 2A(ii)). Loss of Atg7 promotes B cell differentiation into plasma cells in an SLE model²¹⁹. In MS, autophagy has been variably reported as being protective or pathogenic²²⁰. Mice with DC Atg5 genetic ablation have diminished autoantigen-directed T cells in the experimental auto-immune encephalomyelitis, ascribed to defective LAP²²¹.

In cancer²²², the core autophagy gene BECN1 has been early on described as a haploinsufficient tumor suppressor deleted in human breast and ovarian cancers²²³ (Fig. 2A(iii)). Outcomes of anti-autophagy treatments with hydroxychloroquine (an inhibitor of canonical autophagy flux) as adjunct to the standard of care cancer chemotherapy (Fig. 2A(iii)), have shown mixed results²²⁴. Selective canonical autophagy of chromatin factors (WSTF; Fig. 2A(iii)) appears to promote chronic inflammation which may lead to precancerous states⁸. In cancer immunosurveillance, low MHC class I expression and lack of CD8⁺ T cell infiltration in pancreatic ductal adenocarcinoma (PDAC) tumors is linked to MHC class I downregulation via autophagic processes²²⁵. This can be inhibited by expressing mutant deatg8ylase Atg4B in a mouse PDAC model²²⁶ (Fig. 3(iii)). A similar mechanism acts in immunologically cold prostate cancers²²⁷. A paralog of IRGM in human cells, IRGQ, has been implicated in selective autophagy of MHC class I molecules with impact on CD8⁺ T cell immunity against hepatocellular carcinomas²²⁸ (Figs. 2A(iii) and 3(iii)). The phenomenon of hypoxia in tumors and its association with T cell exclusion or inactivity in hypoxic tumors has been recently linked²²⁹ to decreased MHC class I expression via hypoxia-induced autophagy, resulting in sluggish antigen presentation reversible by inhibition of autophagy to enhance antigen presentation. Thus, autophagic regulation of MHC class I levels is an important aspect of cancer immunology.

Autophagic processes have been implicated in the metabolic state of cachexia important in cancer and other conditions. In mice, tamoxifen-induced whole body Atg7 defect showed anorexia and cachexia due to the elevated CCL2 and neuroinflammation suppressing appetite by inhibiting neurons that produce orexigenic peptides²³⁰ (Fig. 3(iii)). Thus, the interplay of inflammation, immunity, and metabolic factors with autophagic processes is of significance in cancer.

Inflammation and autophagic processes are implicated in neurodegenerative disorders such as Alzheimer’s disease (AD)²³¹ (Fig. 3(iv)). Microglia-specific deletion of Atg7 in an AD mouse model impairs microglial ability to clear Aβ plaques causing neurotoxicity²³². In a mouse model of familial AD, conditional ablation of microglial Atg5, but not of Atg14 or the canonical autophagy gene Rb1cc1, affects neurogenesis in the hippocampus or subventricular zone²³³. It then follows that membrane atg8ylation other than canonical autophagy affects AD microglia. Aged mice homozygous for Atg16l1^E230, which can carry out canonical autophagy but not the noncanonical processes^15,172, accumulate extracellular plaques and hyperphosphorylated tau¹⁷¹ (Fig. 3(iv)). The defect in aged Atg16l1^E230 mice is based on an impaired recycling of Aβ receptors including TREM2 in the microglia¹⁷¹ (Fig. 1B(xi)). Inhibition of inflammasome can reduce tau pathology (albeit not Aβ deposition) and restore memory in aged Atg16l1^E230 mice¹⁷¹. A decline in ATG16L1 expression correlates with AD in human brains¹⁷¹ (Fig. 2A(iv)). These studies point to noncanonical autophagic processes and neuroinflammation in AD. Canonical autophagy, affected by the human late onset AD factor APOE4²³⁴, also contributes to protection against AD by its anti-inflammatory properties. Thus, a combination of canonical autophagy and noncanonical membrane atg8ylation processes protect against neuroinflammation in AD (Fig. 3(iv)).

Neuroinflammation has been implicated in Parkinson’s disease (PD) including activation of the microglia, T cells, and pro-inflammatory cytokines. PD incidence correlates with systemic immune inflammation prominently in female subjects²³⁵ whereas several risk genes are associated with autoimmunity including LRRK2, which is shared with CD^236,237. Mutations in the LRRK2 gene are the most common cause of familial PD²³⁶ (Fig. 2(iv)). Increased expression of LRRK2 is implicated in sporadic PD, along with the genes such as PRKN and PINK1^237,238 (Fig. 2(iv)). LRRK2 is expressed in immune cells and is responsive to IL-4 and inflammation²³⁷. DAMPs coming from damaged mitochondria not efficiently removed by the PINK-PARK mitophagy, activate cGAS-STING-Type I IFN pathway and inflammation, which has been linked to parkinsonism-like symptoms in mice²³⁸ (Fig. 3(iv)). Additionally, membrane atg8ylation via the STING-induced VAIL pathway recruits LRRK2 to lysosomes²³⁹. Mutant LRRK2 alters transport and maturation of autophagosomes in neurons in mice²⁴⁰ and triggers secretory autophagy, which preserves neurons but in the long run contributes to inflammatory pathology²⁴¹. Thus, neuroinflammation is important in PD, with relative contributions of specific autophagic processes remaining to be parsed.

Neuroinflammation has been implicated in other neurodegenerative diseases, such as amyotrophic lateral sclerosis (ALS) and frontotemporal dementia (FTD). A major proinflammatory signal transducer and type I IFN activator, TBK1, is a genetic risk factor in ALS and FTD²⁴². TBK1 cooperates with SLRs in removal of protein aggregates, damaged mitochondria, and bacteria^{18,135,243,244}. TBK1 and OPTN link selective autophagy and inflammation with ALS (Fig. 2A(iv)). Mutations in TBK1 causing ALS are far less frequent (1–4%) than C9orf72 mutations, which account for 40% of inherited and 10% of sporadic ALS. C9orf672 plays a role in myeloid cells including microglia and has been implicated in neuroinflammation²⁴⁵. It is found in close to 25% of (FTD) as well as in sporadic FTD cases, with FTD as a disease displaying significant neuroinflammation (microglia and astroglia) and markers of systemic inflammation²⁴⁶. The cause of inflammation is believed to be accumulation of toxic proteins, such as TDP-43, accounting for ca. 4% of familial ALS²⁴⁷, and others including mutant superoxide dismutase SOD1, as well as damaged mitochondria²⁴⁸, which are not properly cleared by processes including TBK1-directed selective autophagy thus triggering proinflammatory signaling. The role of OPTN-driven selective autophagy facilitated via phosphorylation by TBK1¹³⁵ has been affirmed as contributing to clearance of TDP-43 in a mouse model of combined Optn KO and TDP-43-transgene expression (Fig. 3(iv)) which also replicates motor dysfunctions²⁴⁹. In summary, multiple links exist between neuroinflammation and autophagic processes as contributors to neurodegenerative diseases.

An important point of convergence between metabolism, inflammation and autophagic processes is the metabolic syndrome (MetS)²⁵⁰ (Fig. 2A(v)). MetS has been first noted as a series of risk factors for cardiovascular disease and includes visceral-central obesity, hyperglycemia, dyslipidemia with high triglycerides and low high-density lipoprotein, cholesterol, hypertension, and systemic inflammation²⁵⁰. MetS is not equivalent to diabetes although it predisposes to type 2 diabetes (T2D) and steatotic liver disease. Cell-type specific inactivation of Atg genes has sometimes generated contradictory results in MetS studies depending upon the tissues examined. However, whole-body mice haploinsufficient for the atg8ylation gene Atg7 (Atg7^+/−) when fed high fat diet or crossed with the ob/ob mice (mutant for the satiety hormone Leptin), display MetS characteristics combined with inflammation²⁵¹ (Fig. 3(v)). The Atg7^+/− mice compared to Atg7^+/+ mice show increased insulin resistance, hepatic steatosis and serum markers of liver injury, adipose tissue inflammation with infiltrating macrophages and proinflammatory cytokines accompanying NLRP3 inflammasome activation²⁵¹. The recently described MetS obesity-induced airway disease (Fig. 3(v)) is mediated by an adipo-pulmonary axis via the secreted FABP4 (Adipokin-2) hormone²⁵², linked to the control of glucose production by hepatocytes and insulin secretion by pancreatic β-cells. Of note, FABP4 is unconventionally secreted from adipocytes (Fig. 3(v)). Another factor, acyl-coenzyme A binding protein, in humans increased in tissues and plasma during aging in association with obesity and inflammation, is secreted through a form of secretory autophagy²⁵³.

The natural hormone glucagon-like peptide 1 (GLP-1) and synthetic GLP-1 receptor (GLP-1R) agonists such as semaglutide promote release of insulin while blocking glucagon release by the pancreas, control glucose blood levels, slow down the rate of digestion, and increase feeling of satiety in the brain, with benefits beside weight loss. GLP-1R agonists have been reported to increase autophagy, restore lysosomal function, and prevent pancreatic β-cell death in the context of T2D²⁵⁴ (Fig. 2A(v)). Agonists of GLP-1R, as well as gastric inhibitory polypeptide GIP analog and GLP-1R agonist tirzepatide, promote autophagy and reduce inflammation thus acting in cardioprotection²⁵⁵, neuroprotection in AD and PD models²⁵⁶, and against liver disease²⁵⁷ (Fig. 2A(v)).

Inflammation is an acknowledged factor in T2D²⁵⁸. Atg7 haploinsufficiency in ob/ob mice aggravates insulin resistance along with inflammatory changes during progression to diabetes²⁵¹. Canonical selective autophagy protects against pancreatic β-cell loss by clearing cell death-promoting aggregates, including islet amyloid polypeptide (IAPP, amylin)²⁵⁹. IAPP is co-secreted with insulin by pancreatic β cells and is a beneficial hormone that helps regulate postprandial blood glucose levels, slows gastric emptying, promotes satiety synergizing with leptin in the ventromedial hypothalamus and arcuate nucleus, suppresses glucagon release by the liver, and regulates blood pressure²⁶⁰. These effects are remarkably similar to the effects of GLP-1 agonists, however human (but not the murine) IAPP has an unfortunate tendency to form harmful aggregates and promote β-cell death²⁶⁰. The protective role of autophagy, or perhaps other atg8ylation processes, has been demonstrated using Atg7^Δβcell h-IAPP mice expressing human variant of IAPP²⁵⁹. In more recent studies aimed at understanding type 1 diabetes (T1D), which involves autoimmune attacks on pancreatic β-cells, young Atg7^Δβ-cell mice develop diabetes and upregulate MHC-I antigen presentation resulting in elevated immune targeting of β-cell²⁶¹. Of note, similar Atg7 mice were used in the T2D and T1D model studies, so it remains to be understood whether the models fully distinguish between the two types of diabetes vis-à-vis immune/inflammatory components, and whether noncanonical pathways dependent on Atg7 and membrane atg8ylation are involved.

To conclude, the anti-inflammatory and immune manifestations of canonical autophagy and other forms of membrane atg8ylation play a role in a wide range of diseases, many of which are encountered in human populations as aging-associated health problems.

Conclusions and future directions

The research field of canonical and in particular noncanonical autophagic processes in immunity and inflammation continues to grow, both in its fundamental aspects and in the context of human disease. The intersections of autophagy with aging^262,263 in the context of immunity and inflammation represent a complex and evolving topic, touched upon in several sections. Future studies in this area should benefit from considering membrane atg8ylation processes beyond canonical autophagy. The studies of noncanonical processes, which have been appreciated as soon as the field of autophagy merged with the area of immunological studies and animal modeling, is entering a new phase. As a reminder, a conceptual roadblock has been that both canonical autophagy and other membrane atg8ylation processes share membrane atg8ylation factors ATG3, ATG5, ATG7, ATG12 and ATG16L1 (Fig. 1A), and thus differentiating canonical from noncanonical processes requires use of mutants in canonical-autophagy specific genes outside of the membrane atg8ylation system, e.g. ATG9A, RB1CC1, ATG13, etc. Such a workup is generally needed to rule in or out canonical autophagy vs. noncanonical processes especially in animal disease models with inflammatory and immune components. This should not be viewed as an obstacle but rather as a stimulating challenge offering new directions and discoveries especially in the noncanonical branch of the autophagic research. The author hopes that the simplicity of the membrane atg8ylation concept will be helpful and that the appreciation of the variety of processes that organically emanate from it, including canonical autophagy, will assist in fundamental studies and in the development of clinically useful drugs targeting specific autophagic processes.