Pathways of MHC I cross-presentation of exogenous antigens

Abstract

Phagocytes, particularly dendritic cells (DCs), generate peptide-major histocompatibility complex (MHC) I complexes from antigens they have collected from cells in tissues and report this information to CD8 T cells in a process called cross-presentation. This process allows CD8 T cells to detect, respond and eliminate abnormal cells, such as cancers or cells infected with viruses or intracellular microbes. In some settings, cross-presentation can help tolerize CD8 T cells to self-antigens. One of the principal ways that DCs acquire tissue antigens is by ingesting this material through phagocytosis. The resulting phagosomes are key hubs in the cross-presentation (XPT) process and in fact experimentally conferring the ability to phagocytize antigens can be sufficient to allow non-professional antigen presenting cells (APCs) to cross-present. Once in phagosomes, exogenous antigens can be cross-presented (XPTed) through three distinct pathways. There is a vacuolar pathway in which peptides are generated and then bind to MHC I molecules within the confines of the vacuole. Ingested exogenous antigens can also be exported from phagosomes to the cytosol upon vesicular rupture and/or possibly transport. Once in the cytosol, the antigen is degraded by the proteasome and the resulting oligopeptides can be transported to MHC I molecule in the endoplasmic reticulum (ER) (a phagosome-to-cytosol (P2C) pathway) or in phagosomes (a phagosome-to-cytosol-to-phagosome (P2C2P) pathway). Here we review how phagosomes acquire the necessary molecular components that support these three mechanisms and the contribution of these pathways. We describe what is known as well as the gaps in our understanding of these processes.

1. Introduction

All cells use their MHC I molecules to bind, and then display on their surface, peptides that are derived from their expressed genes [1,2]. Effector CD8 T cells then survey these MHC I complexes, searching for ones with immunogenic peptides, which typically are ones whose sequences are different from normal host proteins. This allows the immune system to identify and eliminate abnormal cells, which in nature are ones that are virally infected or harboring mutations (e.g. cancers), and in an iatrogenic setting, cells in allogenic transplants (whose polymorphic protein sequences are recognized as minor-histocompatability antigens).

Most cells are unable to acquire and generate MHC I-presented peptides from exogenous proteins [3]. Biologically this makes sense because if this were to occur, then otherwise healthy cells that were simply exposed to foreign proteins, e.g. from a vaccine or a neighboring cancer, would potentially “sign” their death warrant by displaying immunogenic peptide MHC I-complexes. Given such a lethal consequence, it is important that an individual cell’s MHC I molecules report with high fidelity only what the cell is synthesizing and not what it is exposed to.

A possible exception to this rule was discovered in 1976 during studies of CD8 T cell responses to allogeneic transplants [4,5]. It was found that minor histocompatibility antigens from transplanted cells could somehow elicit CD8 T cells that recognized these transplant antigens presented on the transplant recipient’s MHC I molecules. This was referred to as “cross priming”. At this time, the mechanism for cross-priming was a mystery and in fact, it was not even known yet that MHC I molecules functioned by binding and displaying peptides. Several years later, peptides were found to be presented on MHC I molecules [6, 7]. At this time, the only known mechanism as to how exogenous antigens could be presented on the MHC I molecules of cells (a phenomenon called XPT) was by exogenous peptides binding to surface MHC I molecules, and it was considered that peptides might somehow be released from one cell (e.g, by regurgitation) and then bind to surface MHC I molecules on another cell [8,9]. Although this peptide release and binding may occur in vitro [8,9], it is probably not physiologically relevant. Subsequently, in 1990, it was discovered that when spleno-cytes were incubated with intact extracellular antigen, there were cells that could generate and XPT peptides from the extracellular antigen [10]. These cells turned out to be dendritic cells (DCs) and macrophages [8,11–17].

Initially, cross-priming and XPT were obscure phenomena and their biological significance, if any, was unclear. However, we now appreciate that XPT plays a key role in immune surveillance of tissues. Virally infected parenchymal cells and cancers do not directly stimulate responses from naïve CD8 T cells [18,19]. This is because naïve T cells do not enter the peripheral tissues and, even if they did, the parenchymal cells lack the requisite costimulatory molecules needed to stimulate the naïve T cells. Instead, DCs gather antigens in the peripheral tissues, and then migrate to secondary lymphoid tissues, where they report their findings as peptides bound to MHC I molecules to naïve CD8 T cells. During their sojourn in tissues, DCs might get infected with some viruses, synthesize their antigens, and then report their presence via the classical MHC I pathway (see below). However, the same process cannot occur for antigens that are exclusively made by other cells (e.g. antigens from cancers, minor histocompatibility polymorphisms or tissue-tropic viruses that do not infect DCs). In all of these latter situations, it has been shown that XPT is essential for immune surveillance and the elicitation of CD8 T cell responses [18,19]. In addition, there are intracellular microbial pathogens, such as mycobacterium, fungi, and protozoa, that have evolved to survive and grow in the phagosomes of macrophages, and these infected cells pose a continued threat to the host. Such pathogens are exogenous to a cell (and not in its cytosol) and therefore XPT of the microbial antigens is needed to enlist the help of CD8 T cells, which play a role in controlling and clearing these reservoirs of infection [20–23].

On the opposite side of the coin, there is evidence that XPT of tissue antigens plays a role in tolerance to self-antigens. XPT of exogenous antigens by thymic DCs both in vitro and in vivo was first reported in 1992 [11]. Subsequent studies found that thymic resident and migratory DCs XPT self-antigens to developing T cells [24]. Antigen presentation by these APCs can induce negative selection (deletion) of T cells and thereby contribute to the establishment of central tolerance. In addition, antigen presentation by DCs can induce T regulatory cells both within and outside of the thymus, and these T cells help enforce peripheral tolerance [24]. In addition, DCs can induce peripheral tolerance when they XPT tissue antigens but are in a state where they lack co-stimulatory signals or are suppressive [24].

Given the importance of these various roles of XPT in immunity, as well as its interesting cell biology, this form of antigen presentation has been the subject of considerable investigation over the last 30 years. We now appreciate that there are multiple mechanisms that can lead to XPT. Herein we will focus on what is known about these mechanisms as well as highlighting gaps in our understanding of these processes.

2. The classical MHC I and MHC II pathways of antigen presentation

Before further discussing XPT, it is useful to review in more depth the classical MHC I and MHC II pathways because this will introduce some of the components that participate in XPT and offers some useful comparisons for later discussion.

In the classical MHC I pathway (Fig. 1), all cells use their MHC I molecules to bind and then display on their surface peptides that are derived from a cell’s expressed genes [25,26]. In this process, MHC I heavy and light (beta2-microglobulin) chains are transported by the Sec61 translocon into the ER, wherein they fold and form MHC I heterodimers. Prior to binding peptides, the “empty” MHC I molecules are unstable at physiological temperature [27] and the heterodimer is prevented from dissociating through associations with ER chaperones and other molecules, and localizes into a peptide-loading complex [28,29]. This complex, centered around the ATP-binding cassette (ABC) peptide transporter TAP [30], contains two molecules of MHC I, each bound to the chaperone calreticulin, the oxidoreductase ERp57, and another MHC-encoded protein, Tapasin [29]. Tapasin binding to peptide-empty MHC I molecules helps to stabilize and retain them. In fact, in cells that lack peptides in the ER most MHC I molecules are retained intracellularly in the ER [31,32] or an endosomal recycling compartment [33]. In parallel to this process, cellular proteins are degraded in the cytosol and nucleus into oligopeptides by the ubiquitin-proteasome pathway [34] and a fraction of these peptides are then transported via the TAP peptide transporter into the ER [35]. Once in the ER, the peptides may be further trimmed on their N-termini by the aminopeptidase ERAP1 [36–39]. ERAP1 processes longer peptides with a “molecular ruler” such that it trims its substrates down to a final size of 8–9 residues [38], which is precisely the size range that is needed to bind to MHC I molecules [40]. Peptides that are the right size and sequence can then bind to MHC I molecules and in this process Tapasin can help to select for peptides that bind with high affinity [41] by distorting the MHC I peptide binding groove in ways that widen one of the peptide binding pockets, and also may alter the interactions between the heavy chain and ß2M [42,43]. Once MHC I molecules have bound peptides, they lose their interactions with Tapasin and the peptide-loading complex, then egress from the ER. In the ERGIC and Golgi, TAPBPR, a molecule that functions similarly to Tapasin, can also edit MHC I peptide-binding [44]. Based on these mechanisms, the classical MHC I pathway presents peptides from the majority of proteins that are in a cell’s cytosol and nucleus. These processes do not discriminate between normal (self) or abnormal (microbial or mutant) proteins. In the following sections, we will return to many of these antigen presentation components and processes in our subsequent discussion of the mechanisms of XPT.

Fig. 1.

Overview of the Classical MHC I and MHC II Antigen Presentation Pathways. MHC I Presentation. 1) Endogenous proteins are conjugated with ubiquitin. 2) Poly-ubiquitinated proteins are targeted for degradation by proteasomes. 3) The proteasome hydrolyzes the endogenous proteins into oligopeptide fragments. 4) Some of the peptides are transported into the ER by TAP. 5) In the ER, the aminopeptidase, ERAP1 further trims N-extended peptides to mature epitopes (typically 8–9 amino acids long). Peptides of the right length and sequence bind to the empty MHC I molecules that have been retained and stabilized in the ER by the peptide-loading complex (PLC). The PLC is made up of the proteins: calreticulin, tapasin and ERp57. 6–8) The resulting peptide:MHC I complexes dissociate from the PLC and are transported to the cell surface for presentation. MHC II Presentation. 1) In the ER, MHC II molecules assemble and associate with the invariant chain/CD74, which blocks peptide-binding. 2) Sorting sequences on the invariant chain directs the transport of MHC II molecules into endosomes/phagosomes. 3) In these vacuoles, the invariant chain is proteolytically cleaved from the MHC II molecules leaving only the CLIP region bound in the peptide-binding groove. 4) Exogenous antigens are internalized into endosomes and phagosomes. 5) Cathepsins and other proteases hydrolyze antigens in the endosomes/phagosomes into peptides. 6) In order for these peptides to bind to the MHC II molecule, HLA-DM catalyzes the removal of CLIP from the peptide binding groove and its replacement with high affinity peptides. 7,8) The resulting MHC II-peptide complexes are then transported to the cell surface for presentation. (Created with BioRender.com)

In contrast, the MHC II pathway of antigen presentation (Fig. 1) monitors the proteins present in endocytic compartments (including phagosomes), which originate from molecules ingested from the extracellular milieu or cellular proteins that are imported into these vesicles [45,46]. After forming, endocytic vesicles, including phagosomes, mature and fuse with lysosomes and in these processes acquire cathepsins and other proteases [45,46]. Proton pumping vacuolar ATPases in the vesicular membrane causes the lumen to acidify, and the resulting low pH increases the proteolytic activity of many of the cathepsins. As a result, the vesicles become catabolic compartments and digest their intravesicular proteins into oligopeptides. Through this process, endocytic vesicles generate a library of peptides, which are derived from their ingested or resident proteins, that are then available for antigen presentation. In parallel, and similar to MHC I molecules, MHC II molecules are born and assemble in the ER. However, and in contrast to MHC I molecules, the MHC II complexes quickly associate with another transmembrane protein, the invariant chain (CD74) [47,48]. The invariant chain has a region called CLIP that rapidly occupies the MHC II molecules’ peptide-binding grooves and thereby prevents these complexes from binding peptides in the ER [47]. The invariant chain also contains sorting sequences that directs the transport of these complexes to the endocytic compartments. Once in these vesicles, the invariant chain is proteolytically cleaved, leaving just its CLIP region bound in the peptide-binding groove. HLA-DM molecule binds these MHC II complexes and distorts their peptide-binding groove in ways that catalyze the release of CLIP and binding of high affinity peptides. In contrast to MHC I molecules, peptide-empty MHC II molecules are stable. Consequently, MHC II molecules also can recycle between the cell surface and endosomes, and in so doing acquire new peptides for presentation [49, 50]. As we will discuss, this pathway has some similarities and some differences with one of the XPT pathways.

3. The first step in XPT: antigen internalization

XPT was first demonstrated when APCs were incubated with a soluble extracellular antigen in vitro and were subsequently able to stimulate CD8 T cell response to sequences in the exogenous antigen [10]. This process required energy and protein synthesis [10], so it was clearly distinct from the way investigators had previously generated peptide-MHC I complexes on the surface of cells by incubating them with exogenous peptides, which is something that can occur even on dead cells [51]. While XPT of exogenous soluble antigen was clear cut, its detection required very high concentrations of the input antigen. The necessary concentrations were generally out of the range of what would be encountered under most physiological conditions, and this is presumably why injection of small doses of soluble antigens in vivo, e.g in vaccines, generally failed to elicit CD8 T cell responses. The inefficiency of this process is related at least in part to the mechanism of antigen internalization. When APCs are exposed to soluble antigen, they internalize only small amounts of the protein by fluid phase pinocytosis. Such uptake and subsequent XPT can be enhanced if the soluble antigen binds to cell surface receptors and triggers receptor-mediated endocytosis (e.g. when mannosylated antigens bind to a cellular mannose receptor) [52]. Similarly, if antibodies or chemokine-fusions are used to deliver antigens to cell surface receptors (e.g. XCR1, Dec205), XPT is also enhanced [53–56]. Whether this enhancement is due to an increase in the amount of antigen internalized and/or its localization to a particular subcellular vesicle is not clear.

In contrast, when APCs acquire exogenous particulate antigen by phagocytosis, XPT is detected at 1000–10,000-fold lower amounts of antigen than needed for soluble protein [8,9,12]. This was first observed when APCs were fed a protein conjugated to micron-size particles or bacteria [8,9,12]. Immature DCs and macrophages avidly ingest exogenous particles, internalizing them into phagosomes. While artificial particles are clearly not physiological substrates, they have been very useful to the study of XPT both in vitro and in vivo and are mimics of more biologically relevant phagocytic substrates. Thus in nature, DCs and macrophages avidly ingest microbes, which as noted may inhabit phagosomes, and also dying cells and debris. The cancerous or virally infected cells that are dying, and intracellular bacteria that end up in phagosomes, are exactly the potentially pathogenic events that need to be detected and their antigens reported via XPT. For injured cells, this process is promoted when dying cells display “find me and eat me” signals that facilitate the ingestion of such cells by phagocytes [57–59]. Interestingly, one of these signals is the exposure of F-actin and this is recognized by the Clec9a receptor on conventional type I DCs (cDC1) [60], which are key APCs for the immunosurveillance of dying cells. Upon recognition of F-actin, Clec9a promotes the acquisition and subsequent XPT of antigens from dead cells [60–62].

Antigens from viable cells injected in vivo or expressed as transgenes in what are thought to be long-lived and healthy cells (e.g. beta cells of the Islets of Langerhans) are XPTed in vivo [63–66]. While it is difficult to rule out that these exogenous (to the DC) antigens are released from small numbers of dying antigen-donor cells, there is evidence that DCs may acquire and XPT antigens from live cells [63–66]. One way in which this may occur is when DCs ingest extracellular vesicles, such as exosomes and ectosomes, that are released from living antigen donor cells [67]. Exosomes are vesicles that form in the endocytic compartment and ectosomes are vesicles that bud from the plasma membrane [68]. Both of these extracellular vesicles can contain cellular proteins. Exosomes are internalized by phagocytes through endocytic mechanisms including phagocytosis and macropinocytosis [69]. Antigens in exosomes are XPTed by DCs [67,70,71]. Interestingly, migratory and plasmacytoid DCs can transfer acquired exogenous antigen to other DCs for XPT by releasing vesicles and this may occur during intercellular contact [71]. Another way in which antigens may be acquired from cells is via trogocytosis [72–77]. In this process, DCs and macrophages can take a “bite” out of living cells and in so doing acquire membrane proteins and, at least in some cases, cytosolic components. Trogocytosis may involve phagocytotic mechanisms, but these two processes may not be identical [72,78]. Interestingly, live cancer and infected cells may also release find and eat me signals that lead to phagocytosis [79].

Non-professional APCs that are phagocytic (e.g. neutrophils) can XPT ingested antigens [80], although the biological role of this, if any, is unclear. Interestingly, even non-hematopoietic cells that are unable to naturally internalize particles and XPT exogenous antigens, can be converted into XPTing cells upon transfection of a receptor that confers the ability to phagocytize bound antigenic particles [81,82]. Thus, phagocytosis may be sufficient to permit non-professional APCs to XPT, although how well they do this relatively to DCs has not been compared and is likely less efficient. In any case, phagosomes, and probably to a lesser degree other endocytic compartments, are the entry points to XPT in APCs. The phagosome then serves as a nexus for several different XPT pathways: The phagosome-to-cytosol pathway (P2C) (Fig. 2), phagosome-to-cytosol-to-phagosome pathway (P2C2P) (Fig. 3) and the vacuolar pathway (Fig. 4).

Fig. 2.

The Phagosome to Cytosol (P2C) Pathway of Cross Presentation. 1) Exogenous antigen is taken up by an antigen presenting cell through phagocytosis. 2) Nox2 in the phagosome produces ROS, and in DCs, the resulting elevation in pH reduces destruction of antigen by limiting its hydrolysis. 3) The largely intact antigen exits the phagosome into the cytosol potentially through an ERAD-like mechanism, possibly involving Sec61 (black dashed arrow), or through phagosomal rupture, the latter of which may be enhanced by ROS damage to the vacuole’s membrane. Subsequent steps 4–9 are similar to Fig. 1 A. 4) In the cytosol, the antigen is degraded by the proteasome into peptide fragments. 5) Some of the peptide fragments enters the ER through the TAP transporter. 6) Long peptides are trimmed by ERAP1, and then peptides of the right size and sequence bind to MHC I molecules. 7–9) The peptide-MHC I complexes are released from the ER and transported to the cell surface for presentation. Dashed lines indicate steps that are not absolutely certain. (Created with BioRender.com)

Fig. 3.

The Phagosome to Cytosol to Phagosome (P2C2P) Pathway of Cross Presentation. 1) Phagocytosis of an exogenous antigen by an antigen presenting cell. 2) Similar to Fig. 2, the antigen exits the phagosome into the cytosol potentially through phagosomal rupture and/or an ERAD-like mechanism. 3) The antigen is hydrolyzed by the proteasome into peptide fragments. 4) Some of the peptide fragments go back into the phagosome through the TAP transporter. 5) In the phagosome, long peptides are trimmed by the aminopeptidase IRAP and possibly ERAP1. 6) (A) MHC I molecules are delivered to the phagosome from the cell surface, potentially through an endosomal recycling compartment (ERC) increased by TLR signaling, or (B) from the ER, (C) possibly associated with the PLC. Rab39a helps deliver empty MHC I complexes from the ER. Peptides of the right size and sequence then bind to MHC I molecules. It is not clear whether imported peptide-occupied MHC I molecules can exchange peptides in phagosomes. 7) Once the peptide:MHC I complex is formed in the phagosome, it exits and is transported to the cell surface via mechanisms potentially involving the GTPases, Arf6, Rab11 and Rab22. Dashed lines indicate steps that are not absolutely certain. (Created with BioRender.com)

Fig. 4.

The Vacuolar Pathway of Cross Presentation. 1) Exogenous antigen is internalized through phagocytosis. 2) The antigen is cleaved by intraphagosomal proteases, such as Cathepsin S, into peptides and these are potentially further trimmed by the aminopeptidases, IRAP and ERAP1. 3A-C). The peptides of the right size and sequence are bound by MHC I molecules in the phagosome (the potential sources of intraphagosomal MHC I molecules (dashed red arrows) are the same as discussed in Fig. 3). 4,5) Once peptides are loaded onto MHC molecules, the peptide-MHC I complexes will exit the phagosome and are transported for display on the cell surface of the antigen presenting cell as described in Fig. 3. Dashed lines indicate steps that are not absolutely certain. (Created with BioRender.com)

4. The phagosome-to-cytosol (P2C) pathway of XPT

As noted above, most cells are unable to XPT exogenous antigens on MHC I molecules. However, if an exogenous antigen is artificially forced into the cytosol of such cells, e.g. by osmotic lysis of pinosomes or electroporation, the antigen is processed and presented through the classical MHC I pathway [15,83,84]. Thus, exogenous antigens are not XPTed in most cells because they cannot gain access into the cytosol. However, the situation is different in DCs and macrophages. Multiple lines of evidence have shown that these APCs can transfer antigen from phagosomes into the cytosol. The first indication of this was the finding that XPT by these cells was dependent on proteasomes and TAP, results which implied that the antigen was being hydrolyzed in the cytosol and then accessing MHC I via TAP [85]. In this study and subsequent ones, transfer of antigen from phagosomes into the cytosol was directly demonstrated by detecting the transfer of exogenous enzymes (a ribosomal-inactivating protein, peroxidase, β-lactamase, luciferase etc.), tracers (dextran), proteins (cytochrome C, ovalbumin) and substrates of cytosolic enzymes [85–90]. After this P2C transfer, the antigen is thought to be processed and presented through the classical MHC I pathway like any endogenous antigen (see above), although some of this process may follow an alternate route back into phagosomes (discussed below).

How antigens are transferred from phagosomes to the cytosol is still incompletely understood and more than one mechanism may participate. One proposed mechanism is transfer of proteins through a channel. Protein transport of luminal proteins across a membrane and into the cytosol is well established to occur in the ER during ER-associated degradation (ERAD) [91,92]. In this process, abnormal and/or mis-folded luminal or membrane proteins in the ER are transferred to the cytosol where they are degraded by proteasomes. This process involves ubiquitination of proteins, ER transmembrane proteins such as Derlin family members, and a cytosolic ATPase (p97) that powers the extraction of proteins across the ER membrane. The channel through which ER luminal proteins are translocated into the cytosol is still not entirely clear.

Sec61, the channel involved in transport of secretory and membrane proteins into the ER, has been proposed to also allow retrotranslocation to the cytosol [93–95]. Furthermore, Sec61 binds the 19 S regulatory particle of the proteasome, which may allow for efficient degradation of ERAD substrates [96,97]. Through a genome-wide CRISPR screen to assay for ERAD proteins in mammalian cells, the protein Hrd1 gave the strongest phenotype [98]. In that same screen however, the role of Sec61 was found to be ambiguous – as knocking out certain subunits (Sec61γ, Sec61α1) decreased translocation, while knocking out others (Sec61α2, Sec61β) increased it [98]. More recently, the cryo-EM structure of Derlin-1 has been characterized. Unlike in yeast (where Der1 complexes with other proteins such as Hrd family members) [99], mammalian Derlin-1 has been shown to form an ER-channel through homote-tramerization [100]. Whether these proposed channels play redundant roles in cells or whether they exhibit substrate specificity remains to be explored.

The possibility that ERAD might also be involved in P2C XPT was raised when analysis of the proteome of phagosomes revealed the presence of ER components, including Sec61 [101,102]. How do phagosomes acquire proteins that are normally resident in the ER? Transfer of membrane proteins between subcellular compartments within cells most often occurs through vesicular trafficking and fusion events. Such fusion events are controlled by SNARE proteins, and it was found that loss of the Sec22b SNARE blocked the delivery of ER components to phagosomes and inhibited both XPT and P2C transfer of antigen in vitro and cross-priming in vivo in some [103,104] but not all studies [105, 106]. One implication of these findings was that ER components in phagosomes might participate in P2C antigen transfer; other implications will be discussed below.

Some studies have presented evidence for a P2C ERAD-like mechanism and a role for Sec61 in this process [107–109]. One difficulty with studying the effects on XPT of losing Sec61 is that this channel is needed for the import of many key components of the XPT pathway (TAP, MHC I, etc.), and loss of these other components might account for the inhibition of XPT. However, experiments using an intracellular antibody fragment to retain functional Sec61 in the ER (and which did not reduce protein import into the ER) while depleting it from endocytic compartments, still blocked XPT [109]. On the other hand, other studies have not found evidence for a role in XPT of Sec61 or other ERAD components (Hrd1, gp78, HERP and Derlin-1) except for a role for the p97 ATPase [108,110–113]. Release of a luminal protein (luciferase) from purified phagosomes was promoted by p97 [108]. Thus, at present, whether and how much a phagosomal ERAD-like or other translocation mechanisms contribute to XPT is not clear.

Another proposed mechanism for P2C antigen transfer is rupture of the phagosomal membrane. Such rupture has been documented to occur in phagosomes containing particles [82,114]. There are a number of factors that can destabilize phagosomes, including mechanical stress and membrane damage [115]. Such damage can occur when reactive oxygen species (ROS) are generated in phagocytes by NADPH-oxidase or other mechanisms that generate or transport ROS (e.g. mitochondrial ROS) [116,117], and ROS in phagosomes can then cause peroxidation of lipids [118]. NADPH-oxidase is recruited to phagosomes by VAMP8 and loss of this SNARE protein decreases P2C antigen transfer and XPT [119]. Interestingly, when Clec9a on DCs binds and internalizes particles, it triggers the kinase Syk, which in turn causes local activation of NADPH oxidase. Loss of NADPH oxidase has been found to partially inhibit XPT in several [118,120–122] but not all studies [116]. How much the loss of NADPH oxidase reduces XPT by affecting the pH of phagosomes versus the integrity of the vacuoles is unclear. In any case, while in cells lacking NADPH-oxidase or VAMP8 P2C antigen transfer and XPT are reduced, they are not eliminated [118,120–122], and therefore there are other mechanisms that contribute to these processes. Cells possess mechanisms, such as ESCRT proteins, that repair internal membrane damage. Loss of such repair proteins increases the transfer of antigens from endocytic compartment to the cytosol and enhances their XPT [123], which is consistent with a mechanism wherein membrane rupture contributes to XPT.

Rupture of phagosomes can affect cells in other ways. It will deliver into the cytosol not just intravesicular antigen but also all other phagosomal contents, including cathepsins. The released cathepsins can trigger the NLRP3 inflammasome, if present, which subsequently causes the maturation and release of proinflammatory cytokines (e.g. bioactive IL-1) and can also cause phagocytes to die by pyroptosis [114,124]. In theory, the death of the XPTing cell could limit the duration of antigen presentation and the production of pro-inflammatory cytokines could affect cross-priming, e.g. by inducing DC maturation [125] or other effects [126]. However, the impact of these events on XPT and cross-priming has not been well-studied although there is some evidence that inflammasome activation in parenchymal cells and macrophages can promote cross-priming by adjacent DCs [127,128].

After P2C transfer, exogenous antigens are degraded into oligopeptides by the ubiquitin-proteasome pathway. Proteasome inhibitors block XPT through this pathway, and for epitopes that are dependent on immunoproteasomes, knock out of immunoproteasome subunits inhibits their XPT and ability to cross-prime in vivo [129]. Similarly, depletion of ubiquitin or mutant ubiquitin reduces XPT [130]. Proteasomes frequently generate peptides that are too long to stably bind to MHC I molecules but can be presented after extra N-terminal residues are trimmed by ERAP1 (see above), and loss of ERAP1 reduces XPT [131–134]. Thus, after P2C transfer, exogenous antigens are thought to be processed and presented through the classical MHC I pathway (see above), although some of this process may follow an alternate route back into phagosomes (see below).

5. The phagosome-to-cytosol-to-phagosome (P2C2P) pathway of XPT

Analysis of the proteome of phagosomes also revealed the presence of many MHC I pathway components, including TAP and other components of the peptide-loading complex (TAP1, TAP2, Tapasin, ERp57, and Calreticulin) [135]. These components are normally resident in the ER and their delivery to phagosomes is dependent on Sec22b [103] (see above). Subsequent studies revealed that TAP in phagosomes could transport peptides from the cytosol into the vesicular lumen [136,137]. Since MHC I molecules are also present in phagosomes, this raised the possibility that after P2C transfer of antigen, some of the proteasome-generated peptides could be imported back into phagosomes by TAP. These imported peptides could then potentially be bound by intraluminal MHC I molecules for subsequent display on the cell surface (in other words, by a phagosome-to-cytosol-to-phagosome (P2C2P pathway)). It has been suggested that phagosomes might have a second peptide-import mechanism that could contribute to this process [137,138].

A P2C2P pathway of XPT would require that MHC I molecules traffic to phagosomes, and they are indeed found in these vacuoles [136, 139–141]. There are multiple pathways that have been implicated in the delivery of MHC I to endocytic vesicles and phagosomes. During the formation of phagosomes, MHC I molecules can be internalized from the plasma membrane [142–144]. MHC I molecules can also be internalized by other endocytic mechanisms [144,145], although in many studies this has been assayed only in non-professional APCs and by following the fate of pre-bound anti-MHC I antibodies (a caveat of this approach is that antibody crosslinking can affect internalization of cell surface molecules). Some of the internalized complexes are destined for degradation but others can recycle back to the cell surface [144] (see below) and yet others might be delivered to phagosomes [143]. The short cytosolic tail of MHC I molecules has a non-classical sorting motif that helps traffic MHC I to endosomes in DCs and when mutated reduces XPT to soluble exogenous antigen [146]. Loss of the deubiquitinase UCH-L1 reduces both endosomal MHC I and XPT in LPS-treated cells and it was suggested that this was due to ubiquitination somehow controlling the recycling of MHC I [142].

MHC I molecules can also be exported from the ER to phagosomes in several different ways. In one mechanism, Rab39a, a GTPase involved in vesicular trafficking, aids in the delivery of MHC I molecules to phagosomes and loss of Rab39a reduces XPT [139]. Interestingly, the loss of Rab39a inhibits XPT by cDC2 cells but not cDC1 [139]. Whether this is because cDC1 cells deliver MHC I molecules to phagosomes via distinct and/or redundant pathways is unclear. In another mechanism, the invariant chain (CD74), which is involved in transporting MHC II to endosomes, binds MHC I molecules and also transports them to endocytic compartments [147,148]. However, loss of the invariant chain does not reduce phagosomal XPT [149], although it reduced XPT of soluble protein, presumably in endosomes [150]. The E3 ligase MARCH9 can help direct MHC I trafficking to endosomes by ubiquitinating cytosolic tail of MHC I molecules and loss of this ligase reduced XPT [151]. In yet another mechanism, Toll-like receptor (TLR) agonists that are associated with particulate antigens, stimulate the delivery of MHC I molecules from an endosomal recycling compartment (ERC) to phagosomes via a SNAP23-dependent mechanism [141]. In this latter mechanism, these intravesicular MHC I molecules are not trafficking from the cell surface [106]. While TLR agonists and TLR signaling are not required for XPT, their presence could augment XPT and might be of particular importance in the setting of phagocytized microbes.

In a P2C2P pathway, the MHC I molecules that traffic into phagosomes would bind peptides within this vacuole. There is evidence that this can occur. MHC I molecules in phagosomes that have bound exogenous peptides have been visualized with antibodies specific for SIIN-FEKL peptide-H-2Kb complexes [137,142], although the levels of XPT complexes are generally below the limit of detection of such antibodies (see also section on the vacuolar pathway of XPT).

The findings that in phagosomes, TAP can import peptides, and MHC I molecules are present and can bind peptides, raised the potential of a P2C2P pathway of XPT, but did not indicate whether and how much such a pathway contributes to XPT. One line of evidence that this pathway does indeed contribute to XPT comes from analyses of APCs that are impaired in their ability to target MHC I and other ER components to phagosomes. As noted above, DCs that lack Sec22b have a reduced ability to XPT [103,104]. While this could be simply due to a reduction in P2C antigen transfer, a partial defect in XPT is also seen in cells that have a defect in the delivery of MHC I to phagosomes due to loss of Rab39a [139]. Importantly, loss of Sec22b or Rab39a does not affect classical MHC I or MHC II antigen presentation [103,104,139], which demonstrates a selective impairment of XPT. Moreover, XPT is partially inhibited in Sec22b and Rab39a-deficient mice [103,104,139], demonstrating that these components contribute to CD8 T cell priming in vivo. The reduction but not elimination of XPT/cross-priming in Sec22b and Rab39a-deficient mice is consistent with contributions to such antigen presentation both from MHC I molecules binding XPTed peptides in the ER (P2C) and within the phagosome (P2C2P). Another line of evidence supporting a P2C2P pathway comes from the analysis of cells deficient in IRAP. IRAP is an M1 aminopeptidase that is a relative of ERAP1 [152]. It is principally found in endocytic vesicles including phagosomes and can trim peptides [152]. Loss of IRAP partially inhibits XPT and this appears to be due, at least in part, to the loss of peptide-trimming activity and possibly also an acceleration of phagosomal maturation [153,154]. Together, these findings suggest that a P2C2P pathway contributes to XPT, although a caveat is that it is possible that some of this XPT is occurring through a vacuolar pathway (see below).

There are several gaps in our knowledge as to how MHC I molecules are able to bind peptides in phagosomes and then traffic to the cell surface. One issue is that in order to bind peptides MHC I molecules need to be peptide receptive (i.e., empty). In fact, in phagosomes, peptide empty MHC I molecules are present and can be delivered to this location by a Rab39a-dependent mechanism [139]. These complexes are presumably trafficking from the ER as this process is inhibited by brefeldin A [139]. What is unclear is how empty MHC I molecules, which are highly unstable at physiological temperature, survive the journey to and their sojourn in phagosomes. A related issue is that phagosomes acidify, and the low pH should further destabilize empty MHC I molecules; this may not be an issue in DCs because their phagosomes are at more neutral pH [120]. In the ER, empty MHC I molecules are stabilized by chaperones e.g., within the peptide-loading complex [155]. Perhaps empty MHC I molecules during their transit and residence in phagosomes remain associated with the same ER chaperones (which are present in phagosomes) or other stabilizing mechanisms. A further related issue is what would help the empty MHC I molecules load with peptides in phagosomes? Tapasin fulfills this role in the ER [31,41,156,157] and since Tapasin is present in phagosomes, it might subserve this same function in these vacuoles. Loss of Tapasin inhibits XPT [158], but whether this effect is occurring in phagosomes (P2C2P XPT) versus in the ER (P2C pathway) is unknown. The opposite conceptual problem arises if the XPTing MHC I molecules originate from the cell surface. The issue here is that since these molecules are not starting empty, what then makes them peptide-receptive in phagosomes? There is limited evidence suggesting that MHC I complexes, which contain moderate but not very high affinity peptides, can exchange peptides in endocytic compartments including phagosomes) [159,160], at least in a vacuolar pathway of XPT (see below). Perhaps such peptide-exchange could be accomplished by Tapasin, internalization of MHC I molecules that have lost their original peptide, low pH, and/or some unknown mechanism.

To be XPTed, MHC I complexes that form in phagosomes, ultimately have to traffic to the cell surface. How XPTed MHC I molecules traffic from phagosomes to the cell surface is incompletely understood. There are multiple routes by which molecules can traffic from endosomal vesicles to the cell surface [161]. One of these mechanisms involves the small GTPase ARF6 and this molecule has been implicated in MHC I recycling to the cell surface [162,163]. The GTPases Rab22 and Rab11 have also been linked to MHC I recycling [164].

A final unresolved question is whether P2C antigen transfer due to phagosomal rupture would prevent P2C2P XPT from the disrupted vacuole. In other words, would rupture inactivate the P2C2P pathway? While this may be the case, ruptured phagosome can be repaired by ESCRT or other mechanisms [115,123] and it is also possible that all the necessary events for XPT might still occur from the disrupted structure before it is cleared by autophagy.

6. The vacuolar pathway of XPT

The vacuolar pathway of XPT is analogous to the MHC II antigen presentation in so far as MHC I molecules acquire and XPT peptides generated in phagosomes. Where examined, vacuolar XPT, but not the other XPT pathways, is blocked by cathepsin inhibitors or genetic deficiency of Cathepsin S [33,149]. In contrast, peptide generation in the cytosol does not contribute to vacuolar XPT, since proteasome inhibitors and TAP-deficiency (which can transport peptides into phagosomes as described above) do not block this pathway [149]. Moreover, XPTed epitopes are generated by purified cathepsin S [149] and by purified phagosomes, the latter of which is blocked by cathepsin inhibitors [165]. Cathepsin S is different from many other cathepsins in that it is active at neutral pH [166–168]. This property may be important for this protease to generate XPTed peptide in DCs, whose phagosomes have a less acidic pH [120]. Although proteasomes have been reported to be in phagosomes [169], they do not seem to contribute peptides for vacuolar XPT since proteasome inhibitors do not block this XPT pathway [149]. Thus, in the vacuolar pathway, XPTed peptides are generated in phagosomes in a similar way to MHC II-presented peptides. Whether MHC II presentation, P2C XPT and P2C2P XPT can occur in the same phagosome or only in different ones that have matured in distinct ways is unclear.

In the vacuolar XPT pathway, MHC I molecules acquire peptides generated in the phagosome. The issues around how MHC I molecules traffic to, become peptide-receptive in, and then load with peptides in phagosomes are the same as were discussed for the P2C2P XPT pathway. However, in addition, loss of the TAP peptide transporter has been reported to cause MHC I molecules to traffic to phagosomes from the ERGIC via a Sec22b-dependent mechanism [33]. In this situation, the XPT is dependent on Cathepsin S and therefore via the vacuolar pathway [33]. This altered trafficking of MHC I molecules might be important in nature, e.g., when DCs are infected with pathogens, such as certain Herpes family members, that encode immune evasion molecules which block TAP [170]. However, whether blocking of TAP (as opposed to TAP deficiency) triggers MHC I trafficking to phagosomes was not studied. The vacuolar pathway is operative in TAP-sufficient APCs, as inhibiting cathepsins in these cells reduces XPT [149].

The differential in the requirements for Cathepsin S (vacuolar XPT) and TAP (P2C/P2C2P XPT) has allowed the contribution of these XPT pathways to be evaluated in vivo [149]. Where examined, the loss of just TAP in APCs substantially inhibits cross-priming, the loss of just cathepsin S also inhibits this process, but to a lesser extent. The loss of both TAP and Cathepsin S have additive effects and almost completely inhibit this cross-priming. Thus, the vacuolar and P2C pathways both are operative and contribute to CD8 T cell responses in vivo.

It is interesting that the nature of a phagocytic particle can change the pathways by which an antigen is XPTed. Ovalbumin on iron oxide or latex particles is XPTed more through the P2C pathways, but when it is in poly-lactide-poly-glycolide particles it is presented by both the P2C and vacuolar pathways [149,171]. Similarly, antigens in bacteria can undergo substantial vacuolar XPT [8,33,80,172,173]. While the basis for this is not known, it is likely that these different forms of antigen are affecting the phagosomal environment in ways that affect the XPT pathways. For example, bacteria will stimulate TLRs and this can cause MHC I to traffic from the ERC to phagosomes (see above) and different particles can affect endosomal composition [174].

For intracellular bacteria, the peptides that are generated and presented by the vacuolar pathway in the DCs that prime CD8 T cells are expected to be the same as those produced by the vacuolar XPT in infected phagocytes in tissues. Therefore, the effector CD8 T cells that are cross-primed in such infections should be able to recognize the same presented peptides produced by their target cells. However, because the vacuolar XPT and the classical MHC I pathway may not produce all the same peptides, the specificities of CD8 T cells primed to tumor or viral antigens XPTed via the vacuolar pathway in DCs, may not entirely match the peptides presented by their targets (cancers or virally infected cells), which are presenting their antigens via the classical MHC I pathway. However, CD8 T cells generated against the vacuolar “unique” peptides might nevertheless be useful because, although proteasomes have sequence preferences for cleavages, they are less precise than many proteases (e.g. trypsin) and can cleave after most amino acids (albeit at different rates). Thus in theory, the vacuolar “unique” peptides may not be truly unique and still be made at some level by proteasomes. If so, CD8 T cells primed to the vacuolar pathway “unique” peptides might still recognize their targets. The effect of the differences in peptides stimulating cross-priming versus those being recognized on target cells has not been studied and could potentially change immunodominance, reduce the number of useful effector T cells or even expand the repertoire of responding CD8 T cells and their effectiveness.

7. Cross-dressing

Antigen presenting cells can acquire peptide-MHC I complexes from other cells and use them to stimulate CD8 T cells. This process has been termed cross-dressing and is distinct from XPT. APCs acquire MHC I molecules from donor cells by trogocytosis or transfer from exosomes [175], which results in the transfer of the complexes into the membrane of the APC. Cross-dressing can stimulate CD8 T cell responses in some settings [75,76,176–178]. While this occurs, cross dressing is not required for cross-priming because MHC I-negative or MHC I-mis-matched antigen donor cells (which cannot donate appropriate MHC I complexes) stimulate strong CD8 immunity. In addition, cross-dressing is not sufficient to prime responses in several systems. For example, antigen-donor cells that expressed high levels of a peptide-MHC I complex but little to no pools of intracellular antigen, failed to cross-prime [179]. Similarly, cross-priming to cancer or virally infected cells was defective in mice whose antigen presenting cells had defects in XPT due to a lack of TAP [18,180]. In both of these situations, the transfer of peptide-MHC I complexes to APCs is not occurring at a level that is sufficient to stimulate responses.

The contribution of cross-dressing and XPT differ in various scenarios, and what tips the balance for their importance to the expansion of antigen specific CD8 T cells remains to be explored. Perhaps cross-dressing plays a larger role when target cells (tumors) express a relatively high amount of MHC-I [176]. In a previous work, rejection of C1498.SIY tumors had a greater dependence on cross-dressing than XPT. On the other hand, efficient rejection of B16.OVA tumor (which expressed very low amounts of basal MHC-I), both cross-dressing and XPT played important roles [176]. XPT might also play a larger role in priming CD8 T cells when the classical MHC-I pathway in target cells have been compromised via tumor immunoselection [181] or via viral evasion mechanisms [182]. Another possibility is that certain APC subsets utilize cross-dressing to compensate for their relatively poor XPT ability and that certain conditions in the tumor microenvironment enhance the participation of these cells (eg. interferon stimulation) [76].

8. Capacity of different APCs to use different phagosomal pathways for presenting exogenous antigens

We have reviewed above four different mechanisms by which APCs can present exogenous antigens in phagosomes: P2C, P2C2P, vacuolar, and MHC II pathways. Can all of these pathways operate in the same individual APC? Immortalized cloned DC can present exogenous antigens on MHC II and also carry out XPT that is partially dependent on: TAP and proteasomes (P2C and P2C2P XPT) [15,143], Rab39a (P2C2P XPT) [139], and cathepsins (vacuolar XPT) [15,149,183]. This suggests that all pathways can operate in a single cell type and potentially within a single individual cell, although it is conceivable that there could be heterogeneity within the cloned cell population, e.g. cells in different states of differentiation or activation, resulting in different cells presenting by different mechanisms. On the other hand, some of the processes needed for each of the phagosomal pathways can antagonize one another. For example, the MHC II pathway needs low pH and active cathepsins to degrade peptides in phagosomes and these conditions can reduce XPT by destroying antigens and/or intraphagosomal MHC I molecules. In fact, cDC1 APCs are excellent XPTing cells but do not present well via the MHC II pathway, perhaps because they have phagosomes with more neutral pH (see above) [53,120]. In addition, there is some evidence that different components or pathways may operate in different APCs. For example, Clec9a, which plays an important role in P2C XPT of dead cells, is selectively expressed in cDC1s [60]. Rab39a, which likely participates in the P2C2P pathway (see above), plays a non-redundant role in XPT by cDC2 but not cDC1 APCs [139]. On the other hand, another Rab protein, Rab43, has been shown to specifically enhance XPT by cDC1 and not cDC2 and moDC [184]. Similarly, WDFY4 was required for XPT by cDC1 but not moDC [185]. Macrophages, which play an important role in killing bacterial pathogens, have more acidic and catabolic phagosomes than DCs [120] and this might influence which pathways are operative. Additional study is needed to understand the contribution of the different antigen presentation mechanisms in different APCs.

A related issue, for which there is little known, is how different stages of phagosomal maturation influence the activity of the various antigen presentation pathways. A further related issue is whether in an individual cell all the phagosomes use the same XPT mechanism(s) or are regulated autonomously. There is evidence that phagosomes can function autonomously when some of their receptors engage luminal cargo and that this can influence MHC II antigen presentation. Phagosomal TLRs sensing ingested microbial agonists have been reported to trigger maturation of individual vacuoles in some [186] but not all studies [187, 188]. In addition, in individual phagosomes such sensing triggers autonomous tubulation, leading to interactions between phagosomes, and enhances MHC II antigen presentation [189,190]. Phagosomal Fc receptors sensing antibody cargo, can trigger vacuole maturation and enhance MHC II antigen presentation [191]. Similarly, phagosomal sensing of microbial cargo stimulates the delivery of MHC I molecules to phagosomes, which enhances XPT, presumably through one of the intravacuolar (P2C2P or vacuolar) XPT pathways; whether this was phagosome autonomous was not examined [141].

9. Future research and opportunities for translation of XPT

As discussed above, there are currently many gaps in our understanding of XPT. These include: In the vacuolar and P2C2P pathways, how are empty MHC I molecules stabilized and get to or are generated in phagosomes and then loaded with peptides? In the P2C pathways, are proteins transported out of phagosomes, if so what are the molecules involved in this process, and how much does this contribute to XPT? Do ruptured vesicles contribute to XPT after delivering antigen to the cytosol? What are the contributions to XPT of the P2C vs P2C2P pathways and under what circumstances? Do the three XPT pathways work simultaneously in the same vesicle, in the same cell and how is this regulated? In addition, there are genes and molecules that have been shown to influence XPT but do so through unknown mechanisms. For example, STIM1, a calcium sensor, is needed for optimal XPT through an unknown mechanism [192,193]. In addition, genetic screens have identified other possible and unexpected participants in this process [117,139,143,185,194]. For example, the WDFY4 protein was discovered in a CRISPRcas9 screen and found to participate in XPT both in vitro and in vivo, however its mechanism of action is currently unknown [185]. It is also unclear to what extent the various XPT mechanisms are qualitatively or quantitatively the same in different APC subsets, and across species, such as between mice and humans.

Beyond the basic science, there is potential for translation of XPT into the clinical arena. In particular, there is a medical need for better immunization regimens that can elicit CD8 T cell immunity, particularly for protein-based vaccines. Such approaches could be useful for eliciting protective immunity to viruses and also for cancer immunotherapy. With the advent of checkpoint blockade immunotherapy, which removes the “brakes” on T cell responses by blocking inhibitory check-points, there is interest in potential combination therapy that adds an immunogen to stimulate specific T cells. Protein-based subunit vaccines generally do not elicit CD8 T cell responses, and this is almost certainly because they are poorly XPTed on MHC I molecules. However, when soluble proteins are bound to phagocytic substrates, which target antigens into phagosomal XPT pathways, and then injected in vivo, they can elicit strong CD8 T cell immunity [195]. This approach is attractive as it is targeting antigens into the key APCs (DCs) that are needed to initiate CD8 T cell responses and therefore could be useful for vaccines and immunotherapies. To this end, antigen formulations that promote XPT have been the subject of many publications in recent years [196,197]. On the other side of the coin, targeting antigens into DCs under the right conditions can induce tolerance in peripheral CD8 T cells [198]. Therefore, there is potential for developing antigen formulations that target antigens for XPT in ways that are tolerogenic and this might be useful for the treatment of autoimmune diseases. In addition to targeting antigens into DCs, it may be possible to enhance the activity of XPT pathway to enhance the elicitation of CD8 immunity. For example, small molecule drugs that promote P2C antigen transfer were shown to enhance anti-tumor responses to checkpoint blockade to a mouse tumor [194].

Nucleic acid-based vaccines can elicit CD8 T cell responses. When injected in vivo, these vaccines transfect cells locally (e.g. keratinocytes or muscle cells), resulting in the synthesis of the vaccine antigen. DCs can then acquire these antigens from the transfected cells and XPT them. Although such vaccine can potentially transfect DCs, the ability of various DNA-based vaccines to elicit CD8 T cell responses in vivo is largely dependent on cross-priming [199–202]. It is likely that similar mechanisms apply for RNA-based vaccines [203]. Therefore, a better understanding of the underlying cell biology and regulation of such XPT might identify approaches to enhance the efficacy of such vaccines.

10. Conclusions

The starting point for XPT is the ingestion by DCs or macrophages of antigens in the environment into phagosomes or other endocytic compartments. After this initial event, the phagosome is not simply a passive portal that antigens transit, but rather serves as an active nexus for multiple pathways of XPT. In this process, these vacuoles acquire MHC I molecules; TAP, IRAP, plus other antigen presenting machinery; NADPH oxidase; and proteases, and thereby transform the vesicles into antigen presenting organelles. This maturation of vacuoles supports at least three distinct XPT mechanisms: Vacuolar, P2C and P2C2P pathways. There has been substantial progress in understanding the mechanisms that lead to this phagosomal transformation and in turn the contribution of the various pathways to XPT and cross-priming. Nevertheless, there is still much that is not understood, and it is likely that there are additional pieces of the puzzle that are yet to be discovered.

Abbreviations:

APC

Antigen Presenting Cells

cDC1

Conventional type I DCs

DC

Dendritic cell

ER

endoplasmic reticulum

ERAD

endoplasmic reticulum-associated degradation

ERC

endosomal recycling compartment

MHC

major histocompatibility complex

P2C

phagosome-to-cytosol

P2C2P

phagosome-to-cytosol-to-phagosome

ROS

reactive oxygen species

TLR

Toll-like receptor

XPT

Cross-present(ation)

XPTed

Cross-presented

XPTing

Cross-presenting