Different routes of MHC-I delivery to phagosomes and their consequences to CD8 T cell immunity

Abstract

Dendritic cells (DCs) present internalized antigens to CD8 T cells through cross-presentation by major histocompatibility complex class I (MHC-I) molecules. While conventional cDC1 excel at cross-presentation, cDC2 can be licensed to cross-present during infection by signals from inflammatory receptors, most prominently Toll-like receptors (TLRs). At the core of the regulation of cross-presentation by TLRs is the control of subcellular MHC-I traffic. Within DCs, MHC-I are enriched within endosomal recycling compartments (ERC) and traffic to microbe-carrying phagosomes under the control of phagosome-compartmentalized TLR signals to favor CD8 T cell cross-priming to microbial antigens. Viral blockade of the transporter associated with antigen processing (TAP), known to inhibit the classic MHC-I presentation of cytoplasmic protein-derived peptides, depletes the ERC stores of MHC-I to simultaneously also block TLR-regulated cross-presentation. DCs counter this impairment in the two major pathways of MHC-I presentation to CD8 T cells by mobilizing noncanonical cross-presentation, which delivers MHC-I to phagosomes from a new location in the ER-Golgi intermediate compartment (ERGIC) where MHC-I abnormally accumulate upon TAP blockade. Noncanonical cross-presentation thus rescues MHC-I presentation and cross-primes TAP-independent CD8 T cells best-matched against target cells infected with immune evasive viruses. Because noncanonical cross-presentation relies on a phagosome delivery route of MHC-I that is not under TLR control, it risks potential cross-presentation of self-antigens during infection. Here I review these findings to illustrate how the subcellular route of MHC-I to phagosomes critically impacts the regulation of cross-presentation and the nature of the CD8 T cell response to infection and cancer. I highlight important and novel implications to CD8 T cell vaccines and immunotherapy.

Introduction

Major histocompatibility complex class I (MHC-I) molecules present peptides at the cell surface to CD8 T cells [1]. Classic MHC-I presentation relies on shuttling cytosolic peptides into the endoplasmic reticulum (ER) lumen by the transporter associated with antigen processing (TAP) [1]. The cytosolic peptides are generated via the proteasome and derived from either cellular proteins or viruses and bacteria that infect the cells. TAP is a heterodimer of TAP1 and TAP2 transmembrane ER proteins of the ABC transporter family, and it shuttles cytosolic proteasome-generated peptides across ER membranes [2]. Cryo-electron microscopy recently revealed the 3-dimensional organization of TAP within a macromolecular peptide loading complex comprised of the chaperones tapasin and calreticulin and the oxidoreductase ERp57, and tasked with loading and folding of MHC-I heavy-chain and β2-microglobulin heterodimers with peptides [3–7].

Contrasting with classic MHC-I presentation of endogenous cellular peptides, cross-presentation enables MHC-I presentation of exogenous antigens in the form of soluble proteins and viruses internalized through endocytosis or bacteria and dying cells internalized through phagocytosis [1,8–11]. Cross-presentation by dendritic cells (DCs) initiates CD8 T cell immunity against viruses, bacteria and tumors, and induces peripheral tolerance [1,8–13]. Its role is particularly important in the immune response against cancer or during infection with tissue-tropic viruses that do not infect hematopoietic antigen-presenting cells (APCs), such that priming CD8 T cell responses relies on cross-presentation by DCs that internalize dying cancer cells or infected cells [9,14]. A number of pathways and subcellular locations have been described where cross-presentation might occur [13,15]. Internalized proteins can be translocated to the cytosol and degraded by the constitutive proteasome and the immunoproteasome after which they are transported via TAP either back into phagosomes and endosomes for loading vacuolar MHC-I [13] or into the ER for loading nascent MHC-I (cytosolic pathway) [13]. Alternatively, internalized proteins are degraded by endosomal or phagosomal proteases and resultant peptides loaded onto MHC-I independently of immuno-proteasomal degradation and TAP function (vacuolar pathway) [13]. Delivery of the MHC-I peptide loading complex from a sub-compartment of the ER, the ER-Golgi intermediate compartment (ERGIC), to phagosomes and endosomes via the SNARE protein Sec22b, has provided a mechanism for MHC-I loading within phagosomes and endosomes [16].

Extensive investigations into the mechanisms of cross-presentation have been facilitated by a source of DCs that can be cultured at large numbers to enable detailed investigations at the cell biological, genetic, and biochemical level. Bone marrow derived DCs (BMDCs) have served this function well laying the foundation for understanding the biology, activation, and function of DCs. Original protocols entailed growing bone marrow progenitors in GM-CSF, but later improvements included growth in Flt3L, supplementation of Flt3L with GM-CSF, or coculture with feeder OP-9 cells expressing Delta-like 1 (DLL1) [17,18]. These improvements have enabled these cultures to faithfully represent cDC subsets found in vivo. BMDCs continue to provide a blueprint for understanding DC functions, including those of human DCs, under different conditions. Studies on BMDCs have also informed therapeutic harnessing of DCs for cancer immunotherapy [19–24].

In studying cross-presentation, the choice of ‘particulate’ versus ‘soluble’ antigens determines the internalization route into DCs [25–27]. The mode of antigen internalization is a large determinant of the regulatory mechanisms of cross-presentation. ‘Particulate’ antigens (typically >0.5 μm), including those from bacteria or dying cells, are optimally internalized by phagocytosis [28]. The contribution of phagocytosis steadily decreases with decreasing particle diameter such that endocytosis is considered to favor the internalization of particles <0.5 μm [29–34]. Viruses (20–400 nm), soluble proteins (4–6 nm) and subunit vaccines comprised of purified or recombinant proteins (e.g. Pertussis, influenza, hepatitis B, hepatitis A, meningococcal and pneumococcal vaccines) are internalized by endocytosis –also possibly internalized by micropinocytosis. Virus-like particle-based human papillomavirus vaccines, outer membrane vesicle- or liposome nanoparticle-based vaccines, which all range 25–100 nm, are also endocytosed and engage cross-presentation [35]. Other vaccines are based on targeting DC receptors which undergo endocytosis [36]. The route of internalization is an important factor to consider when studying the regulation of cross-presentation because it dictates the type of cellular receptors engaged, subcellular compartments antigens are delivered to, and the nature of subcellular vesicular trafficking mobilized to internalized antigens to mediate cross-presentation [25–27]. These differences necessitate tailored regulation. Therefore, mechanisms that regulate the cross-presentation of phagocytic antigens are anticipated to be different than those regulating the cross-presentation of endocytic antigens. A thorough dissection of these mechanisms has not been conducted to date. Careful deciphering of the regulation of cross-presentation taking the routes of internalization and receptors engaged into consideration is important because it directly impacts the design of direly needed CD8 T cell vaccines [37]. Understanding the mechanisms that regulate DC cross-presentation is critical for designing novel adjuvants that elicit strong CD8 T cell immunity. This review will discuss the current understanding of the regulation of cross-presentation focusing on the vesicular trafficking routes of MHC-I molecules to sites of internalized antigen.

1. The positive regulation of cross-presentation by Toll-like receptors

1.1. Cross-presentation can be licensed by inflammatory signals

Cross-presentation is a specialty of conventional DCs, classified into cDC1 and cDC2 [38]. cDC1, comprised of CD8α⁺ and CD103⁺ DC, are generally considered superior cross-presenting DC, largely based on their constitutive ability to do so [8–10,38–45]. As such, they are also thought to predominate cross-presentation during infection and inflammation. For example, in pulmonary influenza A virus infection, lymph node (LN) resident cDC1 cross-prime CD8 T cells to influenza virus after migratory tissue DCs carry infected dying cells from the lungs to the LNs and transfer the antigen to CD8α⁺ DCs for cross-presentation to CD8 T cells [46]. Migrant airway-derived or skin-derived CD103⁺ DCs in the inflamed draining LNs can also cross-present influenza or herpes simplex virus type 1 antigens, respectively, to CD8 T cells [47,48]. While cDC1s occupy center stage in cross-presentation, they are not the only cross-presenting DCs [49]. Voluminous work by many groups has shown that other murine and human DCs can acquire cross-presentation under inflammatory conditions, e.g. upon vaccination or upon ligation of inflammatory receptors on DCs [39,46,49–58]. These DCs include cDC2 and inflammatory DCs, the latter now validated to comprise CD14⁺ human DC3 or murine cDC2B, and which arise from monocytes recruited into inflamed tissues and comprise the most abundant DC population during inflammation [59–62]. In fact, the earliest studies by Michael Bevan’s group had shown that both the CD8α⁺ cDC1 and CD8α⁻ cDC2 subset of murine splenic DCs cross-present equally well, with the difference being that the CD8α⁺ cDC1s cross-present constitutively while the CD8α⁻ DCs cross-present upon activation, in this case when FcγR was engaged and activated during antigen-immune complex internalization [39]. Neonatal Fc receptor for IgG (FcRn) drives cross-presentation in CD8⁻CD11b⁺ inflammatory DCs and induces a proinflammatory T cell response in a model of chronic intestinal inflammation [54]. Work from the late Nobel Laureate Ralph Steinman has shown that CD14 is required alongside TLR4 in mobilizing inflammatory monocyte-derived DCs into LNs by LPS or Gram-negative bacteria, and these DCs are as active as cDCs in cross-presentation of proteins and live Gram-negative bacteria [62]. During a fungal infection, induction of a CD8 T cell response to fungal-derived antigen was restricted to CD8α⁻ DCs [50]. In influenza A virus infection, CD103⁻CD11b^hi cDC2 in the submucosa and lung parenchyma, can capture exogenous virus-derived antigen in the lung and migrate to the mediastinal LN to directly cross-prime CD8 T cells [46]. Antigens complexed to antibodies are cross-presented by both cDC1 and cDC2 [39]. Cross-presentation by inflammatory DCs mediates efficacy of anti-tumor chemotherapy or immunostimulatory agents [63]. Adjuvant-dependent tissue accumulation of inflammatory cDC2 is responsible for cross-priming CD8 T cells after immunization via the skin or mucosal routes [53]. Similarly, a subset of F4/80^hi cDC2 of monocytic origin in the small intestine, lung, spleen, mesenteric and mediastinal LNs, and strongly enriched in tumors, can cross-present efficiently [53]. During experimental autoimmune encephalomyelitis, inflammatory TNF and nitric oxide releasing ‘Tip’ DCs accumulate in the central nervous system and cross-present myelin basic protein to activate naive and effector CD8 T cells [59,60]. Inflammatory human CD14⁺ DC3 cDC2 acquire a multitude of pro-inflammatory functions including cross-presentation. All human BDCA3⁺CD141⁺ (cDC1), BDCA1⁺CD1c⁺ (cDC2), and plasmacytoid DCs, have the ability to cross-present extracellular antigens [10,49,64,65]. The mechanisms that license DC cross-presentation function during infection and inflammation are not well understood.

Recent comparative transcriptomic and functional studies on human and murine DCs revealed relatively homogenous cDC1, but extensive heterogeneity in cDC2 [66–69]. Within cDC2, a novel subset of inflammatory DCs was found that expressed CD14 and was named ‘DC3’ in humans and ‘cDC2B’ in mice [66–69]. The primary and consistent characteristic of CD14⁺ cDC2 in humans and mice, across multiple high-dimensional single cell approaches, was their pro-inflammatory profile, accumulation in inflamed tissues, and enrichment in transcripts associated with antigen presentation, compared to other cDC2 and cDC1 [66–69]. Importantly, among DCs, inflammatory human CD14⁺ DC3 expressed high levels of TLR2 and TLR4 [66]. Similarly, compared to cDC1 and other cDC2, murine CD14⁺ cDC2B expressed the highest levels of transcripts involved in antigen processing and presentation [69]. While cDC1 expressed TLRs 3, 11 and 12, cDC2 expressed TLRs 1, 2, 4, 5, 6, 7, 8 and 9 [69]. Inflammatory human DCs (comprised of CD14⁺ DC3 subset of cDC2 [66]), are efficient at cross-presentation and induction of cytotoxic CD8 T cells [70].

This collective body of work provides strong support to the concept that inflammatory signals have the potential to license cross-presentation by cDC2, including human CD14⁺ DC3 and murine cDC2B. Acquisition of cross-presentation capacity by cDC2 during infection and inflammation is consistent with regulation of cross-presentation by inflammatory signals such as those from TLRs which alert the immune system to a microbial infection [8,15,25–27,71–77].

1.2. Endocytic recycling of MHC-I molecules is intricately linked to cross-presentation

The study of cross-presentation has primarily been focused on the cytosolic and vacuolar pathways of liberating peptides for MHC-I presentation and has resulted in the classification of cross-presentation into either vacuolar or cytosolic [78–84]. The subcellular sources of MHC-I and the signals that control their trafficking to sites of internalized antigen have received less attention despite the central role of MHC-I molecules in cross-presentation [8,72,85]. Confocal imaging of murine BMDCs stained with a conformation dependent antibody, AF6-88.5, which detects fully assembled MHC-I molecules [86], shed important insight into the subcellular locations of MHC-I within DCs. Murine BMDCs, CD8α⁺ splenic DCs, and DC2.4 or JAWS-II DC lines, as well as human monocyte-derived DCs, all harbor rich stores of MHC-I within endosomal recycling compartments (ERC) [75,77,87,88]. Plasma membrane MHC-I undergo clathrin-independent endocytosis (CIE) and recycle through either the rapid route through early endosomes or the slow route through the ERC [8,72,85,89]. In DC2.4 cells, MHC-I were internalized into Rab3b/3c positive endosomes that were implicated in rapid recycling and colocalized with other Rabs specific to early endosomes and recycling endosomal tubules such as Rab5, Rab8, Rab10, and Rab35, and whose knockdown during a siRNA-mediated screen impaired cross-presentation [90]. Interestingly, a fraction of Rab3b/c colocalized near bacteria-carrying phagosomes with Rab27a [90], which contributes to phagosomal assembly of the NADPH oxidase NOX2 to mediate phagosome alkalinization and promote cross-presentation [91–94].

The CIE recycling pathway is regulated by the GTPase ARF6, which was originally described to be associated with tubular endosomes emanating from the ERC [89]. Although silencing ARF6 in murine BMDCs was associated with a 30% increase in surface MHC-I levels, it did not impact MHC-I entry into early endosomes or localization to the ERC [87]. These findings were taken to suggest an unidentified source of ERC-resident MHC-I other than the plasma membrane in DCs [87]. Alternatively, the lack of impact of ARF6 silencing on the ERC stores of MHC-I in DCs could have been due to the function of ARF6 specifically in the step of recycling CIE cargo back to the plasma membrane [89].

The ERC is maintained by the activity of the small GTPases Rab11a and is enriched for vesicle-associated membrane protein (VAMP)8/endobrevin and VAMP3/cellubrevin [8,75]. Concordantly, the subcellular ERC localization of MHC-I in DCs is dependent on the activity of Rab11a [75], and disrupting ERC formation by silencing Rab11a abrogated TLR-dependent cross-presentation of phagocytic antigen [75]. Silencing the small GTPase Rab22a, which is important for traffic from early endosomes to the ERC but also from the ERC back to the plasma membrane [89], led to the disappearance of a perinuclear pool of MHC-I (presumably in the ERC) and also impaired cross-presentation [88]. Silencing ARF6, on the other hand, which did not impact the ERC stores of MHC-I [87], had no effect on the cross-presentation of soluble, mannose-receptor targeted, or yeast-associated antigen [87,95], except notably the cross-presentation of Fc receptor-targeted immune complexes [87]. These studies collectively supported a critical role for ERC-resident MHC-I in cross-presentation, although importantly, there may be nuances to the role of the ERC depending on the internalization of cross-presented antigens by phagocytosis or endocytosis. Concordant with this possibility, cross-presentation of soluble antigens, which are internalized by endocytosis, remained intact when either Rab11a or Rab22a function was disrupted [95].

1.3. Control of MHC-I vesicular traffic is at the core of TLR regulation of cross-presentation

TLR engagement in phagosomes carrying microbial ligands transduces a signal through the TLR signaling adaptor MyD88 and the downstream kinase IKK2 to phosphorylate the synaptosomal associated protein 23 (SNAP-23), which flags those phagosomes for MHC-I recruitment from the ERC [75]. Phosphorylated SNAP23 stabilizes SNARE interactions between phagosomal syntaxins and ERC VAMP3 or VAMP8, which result in delivery of ERC-resident MHC-I specifically to phagosomes carrying microbial antigens [75]. Mobilization of ERC-to-phagosome vesicular traffic under the control of TLR signals ensures the availability of large numbers of MHC-I molecules to meet the increased demands of cross-presentation during infection. These findings were made mostly in the context of TLR4 engagement and MyD88 signaling during phagocytosis. TRIF signaling was not involved in the TLR regulation of phagocytic antigen cross-presentation [75]. Whether signals from other MyD88-dependent TLRs such as endosomal TLRs 7 and 9 also license vesicular traffic from the ERC to their respective ligand-carrying phagosomes remains to be tested directly.

Regulated ERC-to-phagosome delivery of MHC-I is a necessary but not sufficient step for completing the process of phagocytic antigen cross-presentation [75]. A parallel ERGIC-to-phagosome pathway of vesicular traffic must also be mobilized to deliver several components of the MHC-I peptide-loading complex and presumably facilitate MHC-I loading in phagosomes [96–98]. ERGIC-to-phagosome directed traffic relies on the pairing of ER SNARE Sec22b with phagosomal syntaxins [16,99], and is not subject to TLR regulation unlike ERC-to-phagosome traffic [75]. Thus, at least two pathways of vesicular traffic converge on the cross-presenting phagosome to deliver different components necessary for phagocytic antigen cross-presentation (Figure 1) [8]: the TLR-regulated ERC-to-phagosome pathway which delivers the centerpiece MHC-I molecules, and the TLR-independent ERGIC-to-phagosome pathway which delivers the accessory molecules potentially needed to both aid peptide generation and peptide loading of MHC-I. It is conceivable that other pathways of vesicular traffic aid cross-presentation by delivering additional components such as the replenishment of NOX2 components from lysosome related organelles mediated by Rab27a [93].

Figure 1. Subcellular localization and traffic of MHC-I molecules during TLR-regulated cross-presentation.

DC phagocytosis of microbes engages TLR-MyD88 signaling. A pathway of vesicular traffic is mobilized from MHC-I-rich ERC to TLR ligand (microbe)-carrying phagosomes. ERC-to-phagosome traffic is controlled by TLR-MyD88 signaling. Activation of IKK2 downstream of TLR-MyD88 signaling phosphorylates phagosomal SNAP-23 and stabilizes interactions of ERC R-SNAREs Endobrevin (VAMP8) and Cellubrevin (VAMP3) with phagosomal syntaxin 4 to mediate ERC-phagosome fusion. As a result, MHC-I molecules are delivered specifically to phagosomes carrying the microbial antigens which warrant cross-presentation during infection. This TLR-regulated pathway of cross-presentation favors the presentation of microbial antigens to CD8 T cells in the context of T cell costimulation during infection. Two other pathways of vesicular traffic emanating from the ERGIC and lysosome related organelles contribute additional components necessary for cross-presentation. See text for more details.

2. Emergency noncanonical cross-presentation during infection with immune evasive viruses

The proteasome-TAP pathway is considered as the conventional processing route for classic MHC-I presentation. Given the crucial role of TAP in translocating peptides to nascent MHC-I molecules within the ER, many clinically important human viruses such as Herpesviridae and Poxviridae have evolved strategies to block TAP and evade host CD8 T cell responses [7,100,101]. Virally encoded proteins that inhibit TAP mediated peptide transport include Infected Cell Protein 47 (ICP47) encoded by herpes simplex virus (HSV)-1 and HSV-2, US6 encoded by human cytomegalovirus (HCMV), and CPXV012 encoded by cowpox virus [7,100] The ability of cowpox to block TAP is particularly noteworthy as it was used by Edward Jenner in 1796 to effectively vaccinate against smallpox leading to the eventual global eradication of smallpox [102]. Pioneering work by Peter Cresswell and Victor Engelhard showed that TAP-deficient cells can present signal-sequence derived peptides by classical MHC-I [103,104]. Others extended these findings showing presentation of a selective peptide repertoire by TAP-deficient murine and human cells, varying in length, amino acid composition, and binding to both classical MHC-1a and nonclassical MHC-Ib [105–109]. In absence of the dominant proteasome-TAP pathway, peptides are liberated by alternate processing pathways which operate alongside the proteasome as evidenced by the presentation of TAP-independent peptides, albeit at low density, when TAP is functional [103,104,110,111]. Alternate processing of endogenous proteins can be mediated by furin, endopeptidases, aminopeptidases, ER signal peptidase, signal peptide peptidase [112,113] and autophagy [114,115].

2.1. Cross-presentation by bystander DC compensates for viral blockade of DC MHC-I presentation

Besides the alternate processing pathways of endogenous proteins that operate independently of TAP function, processing of exogenous proteins in DC during cross-presentation also compensates for TAP inhibition, and can take place in vacuolar compartments by cathepsins S and D [82,112,116], and by recently reported proteasomes within the phagosome lumen [117,118]. Thus, cross-presentation is an important pathway that overcomes impairment in classic MHC-I presentation [9]. When TAP function is compromised in infected tissues or when hematopoietic APCs lack the ability to present antigens through classic MHC-I presentation (for instance owing to compromised TAP function), cross-presentation by bystander or uninfected DCs can prime virus specific CD8 T cells [14,119–123]. Furthermore, hematopoietic deficiency in TAP compromises the ability to generate CD8 T cell responses to cancers [123] or to tissue tropic viral infections that do not infect DCs [121]. These reports have collectively supported the notion that only TAP-sufficient DCs can cross-present and prime CD8 T cells [9,121].

2.2. DC-intrinsic noncanonical cross-presentation in the face of viral interference of MHC-I presentation

The current paradigm suggests that CD8 T cell priming is impaired in the absence of TAP function in DCs [9,121]. Instead, as discussed above, virus-specific CD8 T cells are primed through cross-presentation by TAP-sufficient bystander DCs, which internalize virus-infected cells [8,9,13,14,119–124]. However, the dilemma is that CD8 T cells cross-primed by TAP-sufficient DCs would be unlikely to find their cognate epitopes on tissues infected with immune evasive viruses (Figure 2A). It has been suggested that TAP blockade, reduces but does not completely eliminate MHC-I presentation, which suffices for recognition by already primed effector TAP-dependent CD8 T cells [9]. While residual antigen presentation might allow for some control over infection, immune evasive viruses establish chronic infection and recrudescence causes significant health problems [125–127]. Furthermore, cells infected with immune evasive viruses would present peptides liberated independently of the TAP-proteasome processing pathway [111,112,118,128,129]. TAP blockade in infected tissues thus creates a diminished target or a mismatched target for TAP-dependent CD8 T cells. Can CD8 T cells be primed independently of TAP and if so, how? The answer may lie in noncanonical cross-presentation mobilized as discussed below, when TAP function is compromised (Figure 2B).

Figure 2. Model for cross-priming CD8 T cells to immune evasive viruses through noncanonical cross-presentation.

A. It is widely believed that priming a CD8 T cell response against immune evasive viruses such as Herpesviridae and Poxviridae that block TAP, can only be achieved by uninfected DCs which acquire viral peptides from internalized infected dying cells and present them to T cells. This solution accommodates the paradigm that CD8 T cell immunity is impaired without TAP. However, the presentation of TAP-dependent peptides is severely reduced on tissues infected with immune evasive viruses or in cancers that downmodulate TAP, and instead those tissues or cancers present peptides independently of TAP. Thus, the MHC-I peptidome on these tissues or cancers presents either a diminished or a mismatched target for TAP-dependent CD8 T cells primed by TAP-sufficient DCs leading to inefficient CD8 T cell targeting of infected or cancer cells. B. Noncanonical cross-presentation intrinsic to DCs that have compromised TAP function enables the cross-presentation of epitopes generated in the absence of TAP function to CD8 T cells. The resultant TAP-independent CD8 T cell response is best matched for targeting TAP-independent epitopes presented by tissues infected with immune evasive viruses or in cancers that downmodulate TAP. Mobilization of noncanonical cross-presentation can thus have critical impact on the design of CD8 T cell vaccines to immune evasive viruses and successful immunotherapy for immune evasive cancers.

The debated involvement of TAP in cross-presentation has ranged from necessary to dispensable, and likely dependent on the nature of the antigen, its route of delivery into DCs, and as highlighted in this review, the subcellular localization of MHC-I molecules [130]. Contrary to expectations of the current paradigm, TAP blockade in DCs does not abrogate their ability to prime CD8 T cells [77]. DC cross-presentation of phagocytic antigens can still occur in the absence of TAP function, but it has unique characteristics determined by the following set of events.

2.2.1. TAP blockade alters the location of MHC-I in DCs from the ERC to the ERGIC

Normally, nascent MHC-I molecules in the ER are newly loaded with peptides, exit the ER, and traffic via COPII-coated vesicles to the ERGIC where rigorous quality control ensures proper conformational flexibility and folding [8]. In resting murine DCs, fully assembled MHC-I molecules do not normally colocalize with calreticulin, calnexin, TAP2 or the ERGIC-resident lectin ERGIC-53, consistent with successful export to the plasma membrane once MHC-I molecules have passed ERGIC quality control [75,131,132]. Unstable, empty, and sub-optimally loaded MHC-I heavy-chain/β2-microglobulin dimers, which fail quality control, are prevented from reaching the plasma membrane [1]. This is in fact what happens in the absence of peptides with high-affinity binding to MHC-I when TAP is no longer functional [133–135]. TAP blockade depletes the ERC stores of MHC-I in BMDCs derived from TAP deficient mice, or upon TAP blockade in human monocyte-derived DCs upon infection with HCMV or treatment with ICP47 [77]. Instead of enrichment in the ERC, the majority of MHC-I in DCs are retained within the ERGIC as a consequence of their failure to bind high-affinity peptides and to pass quality control for export to the plasma membrane [77]. Thus, by blocking TAP function, viruses severely compromise not only the surface levels of MHC-I but also the intracellular pools of ERC-resident MHC-I. As such, these viruses can inhibit not only classic MHC-I presentation, but also ERC-dependent TLR-regulated cross-presentation by the DCs they infect.

2.2.2. Alternate MHC-I delivery from the ERGIC to phagosomes for noncanonical cross-presentation

Despite the MHC-I depletion of the ERC and the altered ERGIC localization of MHC-I in the absence of TAP function, the recruitment of MHC-I to phagosomes in DCs remains intact and is independent of Rab11a. Although ERC-to-phagosome vesicular traffic continues to be mobilized upon phagocytosis of TLR ligand-containing particulates, this traffic does not deliver MHC-I because MHC-I are no longer present within the ERC [77]. In the absence of TAP function, MHC-I delivery to phagosomes is sensitive to Sec22b inhibition, which disrupts vesicular traffic between the ERGIC and phagosomes [75,77,99]. Because of the altered MHC-I localization to the ERGIC instead of the ERC, Sec22b-dependent ERGIC-to-phagosome vesicular traffic becomes the dominant route of trafficking MHC-I molecules [77]. Thus, the subcellular source of MHC-I for cross-presentation switches from the ERC to the ERGIC in the absence of TAP function. These data show that when two major pathways of MHC-I presentation, classic MHC-I presentation and the Rab11a-dependent ERC pathway of TLR-regulated cross-presentation are impaired in the face of TAP blockade, DC mobilize a Sec22b-dependent ERGIC pathway of cross-presentation that rescues MHC-I presentation and restores CD8 T cell priming. We have termed this pathway as noncanonical cross-presentation (Figure 3A).

Figure 3. Subcellular localization and traffic of MHC-I molecules during noncanonical cross-presentation of microbial or self antigens.

A. The ERC, which normally harbors large numbers of MHC-I molecules in DCs, becomes depleted of MHC-I molecules upon blockade of TAP. While TLRs continue to mobilize ERC-to-phagosome traffic, this traffic can no longer deliver MHC-I to phagosomes negatively impacting TLR-regulated cross-presentation. Under these conditions, DCs traffic MHC-I to phagosomes from the ERGIC instead of the ERC, in a manner dependent on the ER-SNARE protein Sec22b, to restore the MHC-I presentation of exogenously derived microbial peptides and rescue CD8 T cell priming. Because of the reliance on abnormal traffic of MHC-I, this pathway of antigen presentation is named as noncanonical cross-presentation. B. Noncanonical cross-presentation escapes regulation by TLR signals because it relies on trafficking MHC-I to phagosomes from the ERGIC, where they abnormally accumulate upon TAP blockade. Because TLRs do not control Sec22b-mediated traffic from the ERGIC to phagosomes, noncanonical cross-presentation carries the risk of cross-presenting peptides derived from phagocytosed apoptotic cells to self-reactive CD8 T cells in the context of infection-induced T cell costimulation.

By generating bone marrow chimeric mice where all hematopoietic cells are deficient for TAP, insight could be gained into the physiological relevance of a CD8 T cell response primed by DCs lacking TAP function. TAP-deficient mice have greatly reduced numbers of CD8 T cells because of the central role of MHC-I in thymic CD8 T cell selection [136], but this is not the case in TAP deficient chimeric mice that retain TAP-sufficient radioresistant and non-hematopoietic thymic epithelial cells. Infection of TAP deficient chimeric mice with influenza A virus was associated with an increase in activated CD8 T cells, and revealed that the mice were not particularly susceptible to infection or lethal challenge over a month later [77]. TAP deficient chimeric mice recovered from infection similarly to wild-type chimeric mice, and despite their inability to mount a TAP-dependent CD8 T cell response. These findings were surprising in light of the fact that all DC within these mice lack TAP, including DC that are infected directly with influenza virus as well as the bystander uninfected DC reported to conduct cross-presentation. Virally infected non-hematopoietic cells that retain TAP expression in either wild-type or TAP deficient chimeric mice are unable to directly prime CD8 T cells [121]. Presumably, parallel cross-presentation by uninfected DC and direct presentation by infected DC primes the TAP-dependent CD8 T cell response in wild type chimeric mice [9,14]. In contrast, TAP-dependent CD8 T cell responses to two immunodominant influenza A virus epitopes which require proteasome degradation, were absent in TAP deficient chimeric mice, consistent with the global TAP deficiency in all hematopoietic cells and the absence of residual TAP-sufficient APC persisting post lethal irradiation [77]. Importantly, despite the absence of TAP-dependent CD8 T cells in TAP deficient chimeric mice, their protection against a lethal challenge of influenza virus was nevertheless dependent on CD8 T cells. Indeed, systemic depletion of CD8 T cells significantly compromised the ability of TAP deficient chimeric mice to survive the infection [77], and in effect phenocopied the previously reported susceptibility of TAP deficient mice, which lack CD8 T cells [137], to infection with the intracellular pathogen Mycobacterium tuberculosis [138].

2.2.3. The ERGIC-to-phagosome route of MHC-I traffic is not regulated by TLR signals

Because TLRs control ERC-to-phagosome but not ERGIC-to-phagosome traffic, TLR signals can no longer control TAP-independent phagocytic antigen cross-presentation [75]. DCs derived from TAP deficient mice were shown to efficiently cross-present phagocytic antigen derived from apoptotic cells or antigen-conjugated microspheres in the absence of TLR signals and despite the provision of T cell costimulation. This was unlike the requirement of TLR signaling for cross-presentation of the same antigens by DCs derived from wild-type mice. Cross-presentation by TAP deficient DCs was independent of IKK2 activity [77], which under normal conditions of functional TAP would be necessary to license the traffic of MHC-I molecules to phagosomes from the ERC [75]. This finding was also consistent with ERC depletion of MHC-I in the absence of TAP function in both human and murine DCs [77].

Despite the advantages to anti-viral immunity, noncanonical cross-presentation comes at the cost of bypassing the control that TLRs exert over cross-presentation [77,130] (Figure 3B). It is tempting to speculate that the symptoms of chronic inflammation presented in individuals with genetic TAP1 and TAP2 deficiency [139,140], and perhaps cases of autoimmunity reported following HCMV infections [141] might be triggered by unregulated MHC-I presentation of apoptotic cell antigens as a result of noncanonical cross-presentation. Several single nucleotide polymorphisms (SNPs) in TAP1 and TAP2 genes exist within the human population with no defined effect on peptide transport [142]. Although TAP1 and TAP2 genes are located within the genomic region harboring MHC class II genes that confer high risk for autoimmune diseases, SNPs in TAP2 have been significantly associated with systemic lupus erythematosus (SLE) and autonomously from the HLA allele disease associations [143]. SNPs in TAP1 have shown positive and negative associations with Grave’s disease not explained by linkage disequilibrium between TAP and the HLA-DQ alleles known to confer predisposition to this autoimmune disorder [144]. Interestingly, downregulation of TAP transcripts has been observed in several MHC-I-linked autoimmune diseases including type I diabetes mellitus, Sjögren’s syndrome, Graves’ disease, and Hashimoto’s disease [144], and whether this disrupts the regulatory control of self versus non-self antigen cross-presentation is an important question for future investigations.

The observations discussed above collectively paint a picture of a DC cell-intrinsic pathway of cross-presentation that overcomes the block in both classical MHC-I presentation and conventional cross-presentation in the face of TAP inhibition by immune evasive viruses. This emergency cross-presentation is termed noncanonical cross-presentation because it traffics MHC-I to antigen-containing phagosomes from a new location and via an unconventional route. In this manner, noncanonical cross-presentation results in continued MHC-I presentation of viral antigens by DCs where TAP function is compromised to ensure the priming of an anti-viral CD8 T cell response. The trade-off, however, is the indiscriminate cross-presentation of self and non-self antigens because the alternate route of trafficking MHC-I from the ERGIC to phagosomes is not controlled by TLR signals [77] (Figure 3B).

3. Implications of noncanonical cross-presentation to tumor immunity

Like viruses, many tumors evade cytotoxic CD8 T cells by lowering surface MHC-I expression. This can occur through reduced expression or activity of NLRC5, a master transcriptional coactivator of the genes involved in MHC-I presentation including TAP, immunoproteasome subunits, and β2-microglobulin [130,145]. TAP genes are frequently lost or epigenetically silenced in cancer [146]. Resistance to checkpoint blockade therapy has also been associated with loss of TAP1/2 and MHC-I expression as a functional consequence to mutations in JAK1/JAK2 and lack of responsiveness to IFN-γ in patients with metastatic melanoma [147,148]. CD8 T cells mobilized by vaccines based on TAP-dependent peptides are unlikely to recognize the peptides presented on the surface of TAP-deficient tumor cells [149]. It has been shown that the peptides presented by MHC-I on the surface of TAP-deficient tumor cells are not expressed by normal cells and as such comprise neoantigens [110,150,151]. Designing strategies that mobilize non-canonical cross-presentation in DC should in theory enable the TAP-independent presentation of peptides derived from phagocytosed tumor cells and the priming of TAP-independent CD8 T cells that effectively target immune escape tumor variants. CD8 T cells specific to alternate peptides generated upon the downregulation of TAP have been identified and shown to target a broad range of murine and human tumor cells [106,152].

Conclusions

Studying the vesicular traffic of MHC-I molecules within DCs under different functional states has revealed the existence of two routes of MHC-I delivery to antigen carrying phagosomes: the ERC route and the ERGIC route. During infection, MHC-I molecules are recruited from rich intracellular stores of MHC-I within the ERC to phagosomes carrying microbial antigen. During conditions of TAP dysfunction, modeled by genetic TAP deficiency in murine DCs or infection of human DCs with viruses that encode proteins which block TAP function to escape CD8 T cell immunity, DCs mobilize a cell-intrinsic noncanonical pathway of cross-presentation that relies on an alternate route of MHC-I traffic from the ERGIC –where MHC-I abnormally accumulate upon TAP dysfunction– to antigen carrying phagosomes. Noncanonical cross-presentation operates only when the two major pathways of MHC-I presentation in DCs, classic MHC-I presentation and cross-presentation, are blocked. MHC-I traffic from the ERC to phagosomes is regulated by TLRs, and serves to augment cross-presentation and ensure robust priming of a microbe-specific CD8 T cell response. The ERGIC route of MHC-I traffic operates upon infection with viruses that actively impair the major pathways of MHC-I presentation in DCs; classical MHC-I presentation and cross-presentation. To ensure delivery of MHC-I molecules to phagosomes for cross-presentation, DCs bypass TLR control by mobilizing MHC-I traffic from the ERGIC to phagosomes to restore MHC-I presentation in DCs through noncanonical cross-presentation which rescues CD8 T cell priming. Noncanonical cross-presentation has important implications to immunity and immunotherapy against immune evasive viruses and tumors.

Highlights.

Control of MHC-I traffic is at the heart of TLR regulation of cross-presentation

ERC-to-phagosome MHC-I delivery is critical for phagocytic antigen cross-presentation

TAP blockade inhibits classic MHC-presentation and TLR-regulated cross-presentation

ERGIC-to-phagosome MHC-I delivery upon TAP blockade rescues MHC-I presentation

DC-intrinsic noncanonical cross-presentation cross-primes TAP-independent CD8 T cells

Abbreviations:

MHC-I

major histocompatibility complex class I

TAP

transporter associated with antigen processing

DCs

dendritic cells

APCs

antigen-presenting cells

BMDCs

bone marrow-derived dendritic cells

ER

endoplasmic reticulum

ERGIC

ER-Golgi intermediate compartment

ERC

endosomal recycling compartment

DLL1

Delta-like

ICP47

Infected Cell Protein 47

HSV

herpes simplex virus

HCMV

human cytomegalovirus

LN

lymph node

TLR

Toll-like receptor

IKK2

inhibitor of nuclear factor-κB kinase 2

SNAP-23

synaptosomal associated protein-23

VAMP

vesicle-associated membrane protein