Autophagy and antigen presentation


  • Christian Münz

    1. Laboratory of Viral Immunobiology and Christopher H. Browne Center for Immunology and Immune Diseases, The Rockefeller University, 1230 York Avenue, New York, NY 10021, USA.
    Search for more papers by this author

E-mail; Tel. (+1) 212 327 7611; Fax (+1) 212 327 7887.


CD4+ T cells co-ordinate adaptive immunity and are required for immunological memory establishment and maintenance. They are thought to primarily recognize extracellular antigens, which are endocytosed, processed by lysosomal proteases and then presented on major histocompatibility complex (MHC) class II. However, recent studies have demonstrated that viral, tumour and autoantigens can gain access to this antigen presentation pathway from within cells by autophagy. This review will discuss the autophagic pathways that contribute to endogenous MHC class II antigen processing. Furthermore, potential characteristics of autophagy substrates, qualifying them to access these pathways, and regulation of autophagy will be considered. Finally, I will suggest how antigen presentation after autophagy might contribute to immune surveillance of infected and transformed cells.


T cell of the adaptive immune system monitor the protein degradation products of cells in order to detect pathogen-derived peptides. CD8+ and CD4+ T cells, the two main T cell compartments, thereby survey the output of the two main proteolytic machineries within cells, the proteasome and the lysosome. CD8+ T cell epitopes are usually eight to nine amino acids in length, are presented on major histocompatibility complex (MHC) class I and originate in their majority from cytosolic and nuclear proteins after proteasomal degradation (Fig. 1). On the contrary, CD4+ T cells are stimulated by longer peptides, which are presented by MHC class II molecules and primarily the products of lysosomal proteolysis (Trombetta and Mellman, 2005). Because phagocytes, like dendritic cells and macrophages, very efficiently load extracellular antigens onto MHC class II products and these cell types are instrumental for the initiation of many immune responses, MHC class II antigens are thought to primarily originate from extracellular sources. However, peptide ligand isolation after affinity purification of MHC class II molecules from cell types with less endocytic capacity, like B cells, has revealed that the majority of peptides originate from intracellular sources of these cell types (Rammensee et al., 1999). This suggests that apart from endocytosis other intracellular pathways must exist that deliver antigens to lysosomal degradation and MHC class II presentation. Therefore, rather than the topographical localization of the antigen, the degradation behaviour of a protein might determine, if it is recognized by CD8+ or CD4+ T cells. In other words, substrates of proteasomes and lysosomes will differ in their ability to be presented on MHC class I and II. Consistent with these considerations, recent studies implicate autophagy, a group of degradation pathways that deliver cytoplasmic constituents to lysosomes, in MHC class II antigen processing of intracellular antigens (Brazil et al., 1997; Nimmerjahn et al., 2003; Dengjel et al., 2005; Dörfel et al., 2005; Paludan et al., 2005; Zhou et al., 2005). This review will outline the different autophagy pathways, which have been shown to contribute to endogenous MHC class II antigen processing. Furthermore, it will discuss features of autophagy substrates, which have been revealed in recent studies mainly on neurodegenerative diseases, and will point out how innate and adaptive immunity might regulate autophagy. Finally, I will suggest ways in which MHC class II antigen processing after autophagy might contribute to immune responses.

Figure 1.

The degradation behaviour of three distinct cytosolic/nuclear protein pools might determine preferential processing onto MHC class I or II. 1. Soluble short-lived proteins are primarily degraded by the proteasome and peptide products of this proteolysis gain access to the endoplasmic reticulum (ER) via the transporter associated with antigen processing (TAP). In the ER, newly synthesized MHC class I molecules load some of these ligands and migrate to the cell surface via the secretory pathway to be recognized by CD8+ T cells. 2. Soluble proteins or peptides that contain a lysosomal targeting signal peptide get imported into the MHC class II loading compartment (MIIC) via the LAMP-2a transporter under the assistance of cytosolic and lysosomal Hsc70 members. In MIICs, they get degraded by lysosomal proteases and their fragments get loaded onto MHC class II molecules, which then migrate to the cell surface for CD4+ T cell stimulation. 3. Aggregate-prone long-lived proteins get incorporated into autophagosomes, which then fuse with MIICs. Their content is also degraded by lysosmal hydrolases and peptides derived thereof are loaded onto MHC class II. Stable MHC class II/ligand complexes then move to the cell surface and display their cargo to CD4+ T cells.

Pathways of autophagy

In higher eukaryotes three main autophagic pathways exist that deliver substances from the cytoplasm to lysosomes: macroautophagy, microautophagy and chaperone-mediated autophagy (Klionsky and Ohsumi, 1999). Only macroautophagy and chaperone-mediated autophagy have been characterized in mammalian cells to some degree, and implicated in processing of intracellular proteins onto MHC class II so far (Paludan et al., 2005; Zhou et al., 2005). Therefore, I will focus on these two pathways in this review.

Macroautophagy has been described to contribute to MHC class II antigen presentation of viral, self- and tumour antigens (Brazil et al., 1997; Nimmerjahn et al., 2003; Dörfel et al., 2005; Paludan et al., 2005). During this process a cup-shaped isolation membrane forms from membranes of unknown origin (Yorimitsu and Klionsky, 2005). Two conjugation systems, which involve ubiquitin like coupling of either Atg8/LC3 to the autophagosomal membrane or Atg12 coupling to Atg5, which then localizes via association with Atg16L to the isolation membrane, are essential for the formation of the isolation membrane (Yorimitsu and Klionsky, 2005). This isolation membrane engulfs cytoplasmic content and upon closure forms autophagosomes with one or multiple intravesicular membranes and a fairly large diameter of 0.5–1.5 µm. These autophagosomes fuse with lysosomes and late endosomes (Liou et al., 1997; Berg et al., 1998) (Fig. 1). The ultrastructural morphology of the fusion vesicles between autophagosomes and late endosomes, so-called amphisomes, has been described to be multilamellar and multivesicular. Lysosomal hydrolases in the fusion vesicles then degrade the intravesicular membranes and autophagosome content. While Atg5/Atg12/Atg16L trimeric complexes recycle from the isolation membrane after autophagosome formation, Atg8/LC3 stays coupled to phosphatidyl-ethanolamine on the out- and inside of the autophagosome and is partially degraded with its content (Kabeya et al., 2000).

Interestingly, the vesicular compartment in which MHC class II molecules acquire their peptide cargo resemble in many features amphisomes. First, MHC class II loading compartments have been described to be late endosomes, in which endocytosed antigens meets MHC class II molecules and peptide loading occurs with the assistance of the non-classical MHC class II molecule HLA-DM after lysosomal degradation of the invariant chain chaperoning MHC class II to these vesicles (Tulp et al., 1994). These compartment display, similar to amphisomes, a multivesicular or multilamellar morphology (Zwart et al., 2005). Second, when peptides were eluted from affinity-purified HLA-DR molecules of Epstein Barr virus (EBV) transformed B cell lines and 404 MHC class II ligand sequences were identified, two peptides turned out to be derived from Atg8/LC3 (MAP1LC3B93−109 and MAP1LC3B93−110) (Dengjel et al., 2005), indicating that autophagosome associated proteins can gain access to MHC class II loading. Taken together these lines of evidence suggest that autophagosomes might fuse with MHC class II loading compartments.

Chaperone-mediated autophagy delivers cytosolic proteins directly into lysosomes and has been described to enhance autoantigen presentation by MHC class II (Zhou et al., 2005). This pathway translocates cytosolic proteins, like RNAse A, via the protein transporter LAMP-2a assisted by cytosolic and lysosomal Hsc70 chaperone proteins from the cytoplasm across the lysosomal membrane (Agarraberes and Dice, 2001) (Fig. 1). Protein sorting into this pathway is guided by a mostly pentameric signal peptide sequence (KFERQ in RNAse A). Because antigens for MHC class II presentation are primarily degraded by lysosomal proteases, substrates of chaperone-mediated autophagy might readily gain access to the MHC class II loading compartment. It remains to be determined if the signal peptide requirement for transport significantly limits the number of substrates that can gain access to MHC class II presentation via this pathway.

Substrates of autophagy

The above-listed studies suggest that autophagy delivers antigens to the MHC class II loading compartment. Understanding the characteristics of substrates for macroautophagy and chaperone mediated autophagy should allow us to predict which pathogen derived proteins can be detected by CD4+ T cells via this pathway and which signals should be used to target antigens into this pathway for enhanced CD4+ T cell stimulation.

Macroautophagy substrates have been described to be cell organelles, intracellular pathogens, long-lived proteins and protein aggregates (Mizushima, 2005). Mitochondria and endoplasmic reticulum fragments have been found in ultrastructurally defined autophagosomes, and due to their large vesicle size autophagosomes are assumed to deliver aged cell organelles to lysosomes for degradation (Klionsky and Ohsumi, 1999). In addition, macroautophagy has been described to mediate innate resistance against invading pathogens. Both bacteria and viruses can be targeted by autophagic degradation (Kirkegaard et al., 2004; Levine, 2005). The best characterized examples for pathogen autophagy, also termed xenophagy, are engulfment of Mycobacterium tuberculosis containing phagosomes (Gutierrez et al., 2004), trapping of cytosolic group A Streptococci in autophagosomes (Nakagawa et al., 2004) and immune escape from autophagy by Shigella (Ogawa et al., 2005). Furthermore, macroautophagy was found to mediate protection against Herpes simplex virus infection of mouse embryonic fibroblasts as part of the type I Interferon induced antiviral response (Talloczy et al., 2002). Moreover, beclin-1, the mammalian homologue of the essential autophagy gene product Atg6, was found to limit Sindbis virus mediated encephalitis in mice (Liang et al., 1998). These studies implicated macroautophagy in the clearance of aged organelles and intracellular pathogens, suggesting specific recognition of these structures by the forming autophagosome.

Specificity of macroautophagy for distinct protein substrates is much less clear. Early studies found a dichotomy of protein substrates for proteasomal versus lysosomal degradation, guided by the protein half-life (Henell et al., 1987). Long-lived proteins were preferentially degraded in lysosomes after autophagy, while short-lived proteins were targeted for proteasomal proteolysis. In good agreement with these initial findings, glyceraldehyde-3-phosphate dehydrogenase (GAPDH), a protein with an extraordinarily long half-life of 130 h (Dice and Goldberg, 1975), was reported to be a substrate for chaperone-mediated autophagy (Aniento et al., 1993) and could be isolated from autophagosomes (Fengsrud et al., 2000). Interestingly, GAPDH derived peptides could be isolated from four different HLA-DR and -DQ alleles, while it was never found as a source of natural MHC class I ligands (Rammensee et al., 1999). Vice versa, short-lived cyclins are among the most frequent sources of natural MHC class I ligands, but not a single cyclin-derived peptide has so far been eluted from MHC class II molecules (Rammensee et al., 1999). These findings suggest that long-lived proteins are preferentially targeted by autophagy and frequently presented on MHC class II.

However, if autophagy is selective for long-lived proteins, how are these detected by the forming autophagosome. This questions still remains largely unanswered. However, recent studies primarily on proteins involved in neurodegenerative diseases such as Alzheimer’s, Huntinton’s and Parkinson’s disease speculate, that protein aggregates might be suitable substrates for macroautophagy. It could be demonstrated that aggregates of proteins with polyglutamine domains like Huntingtin accumulate upon inhibition of autophagy and their levels can be reduced upon stimulation of autophagy with rapamycin (Iwata et al., 2005a,b; Ravikumar et al., 2002; 2004; 2005). Interestingly, treatment with the rapamycin analog CCI-779, which upregulates autophagy, ameliorated neurodegeneration in animal models of Huntington’s disease (Ravikumar et al., 2004). Two of these studies implicated transport on microtubules in the delivery of aggregate-prone proteins to autophagosomes (Iwata et al., 2005b; Ravikumar et al., 2005). Protein aggregates could even in a broader sense serve as macroautophagy substrates, because conditional knockout of the essential autophagy gene product Atg7 in the liver (Komatsu et al., 2005) and lack of Atg5 in neonates (Mizushima, 2005) led to the accumulation of ubiquitinated protein aggregates. Two cytosolic proteins have been suggested to assist in the recruitment of autophagosomal membranes to ubiquitinated protein aggregates. One of them, Alfy can bind posphatidylinositol-3-phosphate, the product of class III PI3Kinases, which induce autophagy, and localizes to protein aggregates and autophagic membranes (Simonsen et al., 2004). The other, p62/SQSTM1 contains an Ubiquitin binding domain (UBA) and its polymerization is required for efficient recruitment of the Atg8/LC3 to these structures (Bjorkoy et al., 2005). However, none of these two protein aggregate associated proteins has been shown to directly interact with essential autophagy gene products and therefore initiate the nucleation of isolation membranes around protein aggregates. Nevertheless, protein aggregates have been described to evade proteasomal degradation (Bence et al., 2001) and might therefore be preferentially targeted for lysosomal proteolysis after macroautophagy. From these studies it is tempting to speculate that long-lived proteins might aggregate more easily and then become substrates for macroautophagy.

Regulation of autophagy

Traditionally, upregulation of autophagy has been reported to overcome starvation periods or growth hormone depletion (Mizushima, 2005). Nutrient or growth hormone depletion is thought to inhibit the mTOR kinase, which prevents the assembly of the autophagy-inducing Atg1 kinase complex. In addition, lack of amino acids triggers the assembly of the type III PI3 kinase/beclin-1 signalling complex for autophagy upregulation. While some of the players of starvation mediated autophagy induction have been identified it is much less clear how and to which extent autophagy is regulated upon cellular stress like bacterial and viral infections.

Both type I and II Interferons have been reported to influence autophagy levels. Herpes simplex virus infection of mouse embryonic fibroblasts was reported to upregulate autophagy via the IFNα/β inducible kinase PKR, which phosphorylates the eukaryotic initiation factor-2-α for translation arrest during viral infections (Talloczy et al., 2002). This finding suggested that innate immune responses to viral infection might trigger macroautophagy, but it is unknown which components of the autophagy machinery participate in this regulation. In addition, macroautophagy might be increased during immune responses due to the action of IFNγ. Autophagic clearance of M. tuberculosis infection of mouse macrophages was enhanced upon IFNγ treatment (Gutierrez et al., 2004). Therefore, autophagy regulation might be used to clear pathogens during infection and it might lead to enhanced trafficking of antigens to MHC class II loading compartments for increased CD4+ T cell stimulation.

MHC class II antigen presentation after autophagy

The above-listed evidence suggested that autophagic pathways exist and deliver proteins from the cytosol to late endosomes, the vesicular compartment, in which MHC class II loading occurs, and that these pathways might even be upregulated upon infection. In addition, when peptides were eluted from affinity-purified MHC class II molecules, up to 20% were found to originate from cytosolic sources, while only few ligands from extracellular proteins were identified (Chicz et al., 1993; Rammensee et al., 1999; Dongre et al., 2001; Dengjel et al., 2005). In addition, especially HLA-DR ligands, originating from cytosolic and nuclear proteins, were found upregulated on starved EBV transformed B cells, while peptides from membrane and secreted proteins were unaffected by amino acid depletion (Dengjel et al., 2005). Interestingly, four MHC class II ligands stood out in this analysis and their presentation was upregulated between 1.8- and 2.9-fold upon starvation for 24 h. These were peptides from elongation factor 1-alpha 1, cathepsin D, heat shock 70 kDa protein 1 and RAD23 homologue B. Interestingly, RAD23 homologue A and B proteins are long-lived and contain a C-terminal UBA2 domain that protects them from proteasomal degradation (Heessen et al., 2005). The UBA2 domains transfers half-life extension and proteasomal proteolysis protection to other proteins. These RAD23 characteristics fit the criteria for autophagy substrates and RAD23’s MHC class II presentation was upregulated upon starvation induced autophagy.

For a small number of proteins, involvement of autophagy in MHC class II presentation and CD4+ T cell recognition was also demonstrated directly. In the earliest of these studies, Stockinger and colleagues have shown that complement C5 is endogenously processed for CD4+ T cell recognition in mouse macrophage and B cell lines (Brazil et al., 1997). MHC class II presentation of C5 was dependent on lysosomal degradation and could be blocked by NH4Cl, and was sensitive to the class III PI3 kinase inhibitor 3-Methyladenine (3-MA), which potently inhibits macroautophagy (Seglen and Gordon, 1982), but to some extent also compromises other membrane trafficking events. This study suggested that ectopic overexpression leads to endogenous MHC class II antigen processing of complement C5 after macroautophagy in mouse macrophage and B cell lines. In a second study, Mautner and coworkers transfected EBV transformed human B cells and a renal cell carcinoma cell line with neomycin phophotransferase II (NeoR) and documented that NeoR is endogenously processed for MHC class II presentation to CD4+ T cells (Nimmerjahn et al., 2003). They could show that NeoR is degraded in lysosomes in these cell types and its targeting to lysosomes is 3-MA sensitive. Implicating macroautophagy in endogenous MHC class II processing of NeoR, the authors could block CD4+ T cell recognition of their NeoR transfectants with 3-MA and the somewhat less specific PI3 kinase inhibitor wortmannin. These data suggested that overexpression of a cytosolic protein can lead to macroautophagosomal delivery of this protein for MHC class II presentation. Thirdly, our studies documented that the nuclear antigen 1 of EBV (EBNA1) gains access to endogenous MHC class II antigen processing in EBV transformed B cells (Münz et al., 2000). EBNA1 could be visualized in autophagosomes after inhibition of lysosomal proteolysis, and inhibition of macroautophagy by either 3-MA or Atg12 specific RNA interference significantly decreased recognition of EBV transformed B cells by EBNA1 specific CD4+ T cell clones (Paludan et al., 2005). EBNA1 was the first pathogen-derived antigen that was demonstrated to gain access to MHC class II presentation after macroautophagy at physiological expression levels. Interestingly, EBNA1 is a long-lived protein, carrying a Glycine-Alanine repeat domain that downregulates proteasomal degradation and MHC class I presentation (Levitskaya et al., 1995; 1997). When its Glycine-Alanine repeat domain is deleted EBNA1’s half-life shortens and its recognition by CD8+ T cells is enhanced (Münz, 2004). In addition, Glycine-Alanine repeat domains of different length transfer protection from proteasomal proteolysis to other proteins (Dantuma et al., 2002). These characteristics are reminiscent of the above discussed degradation behaviour of RAD23, suggesting that antigens with long half-life and evasion from proteasomal degradation are well presented on MHC class II after macroautophagy. A third example of a protein falling into this category is the Influenza matrix protein 1 (MP1). Long and colleagues have demonstrated that MP1 is intracellularly processed onto MHC class II (Jaraquemada et al., 1990). However, when the half-life of MP1 is shortened by N-end rule modification, which makes proteins more susceptible for proteasomal degradation, endogenous MHC class II processing is lost and CD4+ T cells fail to detect mutant MP1 expression in EBV transformed B cells (Gueguen and Long, 1996). Therefore, long protein half-life, evasion from proteasomal degradation, macroautophagy and endogenous MHC class II processing might be linked. In a fourth study on MHC class II loading via an intracellular route, Brossart and coworkers demonstrated that MHC class II presentation of the tumour antigen Mucin 1 (MUC1) after transfection into dendritic cells by RNA electroporation was sensitive to 3-MA and wortmannin treatment (Dörfel et al., 2005). Because, however, the authors also could decrease CD4+ T cell recognition of MUC1 RNA-electroporated dendritic cells by proteasome inhibition, the pathway leading to the observed MHC class II presentation is not entirely clear. Finally, in a fifth study Blum and colleagues described that MHC class II presentation of the autoantigens glutamate decarboxylase 65 (GAD65) and SMA, a mutant human immunoglobulin κ chain, could be increased by overexpression of Lamp-2a, the transporter associated with chaperone-mediated autophagy (Zhou et al., 2005). CD4+ T cell recognition of these autoantigens was also found to be dependent on proteasomal degradation and the authors therefore suggested that GAD65 and SMA are processed to peptides in the cytosol by proteasomes and the resulting peptides are then transported into MIICs by chaperone mediated autophagy. The above list of antigens suggest that autophagic pathways, previously primarily considered to generate nutrients during starvation conditions, might deliver a substantial amount of cytosolic and nuclear antigens for MHC class II presentation to CD4+ T cells.

Role of autophagy during immune responses

Recent studies have implicated autophagy in MHC class II presentation of intracellular antigens to CD4+ T cells. However, the importance of this pathway for the immune system and its responses against pathogens and tumours remains unclear. Several lines of evidence, however, suggest that this pathway might contribute to immune surveillance of intracellular antigens and might be involved in T cell selection in the thymus. In a GFP-Atg8/LC3 transgenic mouse constitutive macroautophagy was observed in many tissues (Mizushima et al., 2004), but was especially elevated in cortical thymic, lens and kidney epithelial cells as well as some exocrine gland cells such as gastric chief cells and seminal vesicle cells. Of these, the cortical thymic epithelial cells are of special importance to the immune system, because they have been implicated in positive selection of T cells, including MHC class II restricted CD4+ T cells (Starr et al., 2003). Because this tissue, however, lacks robust endocytic capacity, cortical epithelial cells might employ intracellular pathways, including autophagy, to load their MHC class II molecules for positive selection of CD4+ T cells.

In addition to thymic epithelium, autophagy might contribute to MHC class II loading in other epithelial tissues that upregulate MHC class II molecules upon exposure to inflammatory cytokines, predominantly IFNγ (Reith and Mach, 2001), and CD4+ T cells might contribute to immune surveillance of these tissues. This notion is supported by the fact that cytolytic CD4+ T cells can be found in close proximity to MHC class II positive epithelial cells in inflamed tissues (Wu et al., 1994; Muller et al., 1998; Yawalkar et al., 2001). These cytolytic CD4+ T cells seem to be much more frequent than previously appreciated and have now been identified ex vivo in both mouse and man (Appay et al., 2002; Amyes et al., 2005; Jellison et al., 2005). In addition, cytolytic CD4+ T cells have been found to target tumour and virus infected cells in vitro (Münz et al., 2000; Bickham et al., 2001; 2003; Nikiforow et al., 2001; 2003; Paludan et al., 2002; Hegde et al., 2005) and in vivo (Stevenson et al., 1999; Robertson et al., 2001; Fu et al., 2004; Sparks-Thissen et al., 2004). These data suggest that immune surveillance by CD4+ T cells might target infected and transformed cells after intracellular antigen processing onto MHC class II molecules via autophagy.

Apart from direct immune effector functions by CD4+ T cells in response to display of intracellular antigens on MHC class II molecules, stimulation of these cells is also crucial to establish and maintain immunological memory in B and T cell compartments (Bevan, 2004; McHeyzer-Williams and McHeyzer-Williams, 2005). Therefore, broadening the repertoire of MHC class II presented antigens by displaying peptides from intracellular antigens for the stimulation of CD4+ T cell help should be in the interest of the immune system. In addition, cytosolic and nuclear antigens displayed by epithelial cells on MHC class II at sites of inflammation could boost CD4+ T cell responses, which then mature dendritic cells for more efficient priming of CD8+ T cell responses and provide support factors for CD8+ T cell maintenance. This MHC class II presentation at sites of infection might eventually even lead to the development of tertiary lymphoid organs at sites of pathogen invasion and tumorigenesis.

Altogether, immune surveillance of intracellular antigens through MHC class II presentation after autophagy might complement CD8+ T cell responses restricted by MHC class I quite well, and selectively target long-lived aggregate-prone proteins, which are inefficiently degraded by proteasomes and therefore difficult to present on MHC class I.


Recent studies by a number of laboratories have implicated autophagy in the delivery of cytosolic and nuclear antigens for MHC class II loading. These studies suggest that CD4+ T cells are able to immune survey intracellular antigens in addition to extracellular antigens, which are taken up and processed onto MHC class II by phagocytes. Two autophagic pathways, macroautophagy and chaperone-mediated autophagy, have been implicated in delivering endogenous antigens for MHC class II loading. Interestingly, these pathways might tab into a different cytosolic and nuclear antigen pool than endogenous MHC class I antigen processing. Primarily long-lived and aggregate-prone proteins, which manage to escape degradation by the proteasome, might be channelled into macroautophagy and displayed on MHC class II for CD4+ T cell stimulation. Endogenous MHC class II antigen processing after autophagy might lead to immune surveillance by CD4+ T cells in MHC class II positive inflamed tissues and enhance CD4+ T cell help for the initiation and maintenance of adaptive immune responses.


I thank the National Cancer Institute (R01CA108609), the Arnold and Mabel Beckman Foundation, the Alexandrine and Alexander Sinsheimer Foundation, and the Foundation for the National Institutes of Health, Grand Challenges in Global Health, for supporting my research.