The evolving biology of cross-presentation

Abstract

Cross-priming was first recognized in the context of in vivo cytotoxic T lymphocyte (CTL) responses generated against minor histocompatibility antigens induced by immunization with lymphoid cells. Even though the basis for T cell antigen recognition was still largely unclear at that time, these early studies recognized the implication that such minor histocompatibility antigens were derived from the immunizing cells and were obtained exogenously by the host’s antigen presenting cells (APCs) that directly prime the CTL response. As antigen recognition by the T cell receptor became understood to involve peptides derived from antigens processed by the APCs and presented by major histocompatibility molecules, the “cross-priming” phenomenon was subsequently recast as “cross-presentation” and the scope considered for examining this process gradually broadened to include many different forms of antigens, including soluble proteins, and different types of APCs that may not be involved in in vivo CTL priming. Many studies of cross-presentation have relied on in vitro cell models that were recently found to differ from in vivo APCs in particular mechanistic details. A recent trend has focused on the APCs and pathways of cross-presentation used in vivo, especially the type 1 dendritic cells. Current efforts are also being directed towards validating the in vivo role of various putative pathways and gene candidates in cross-presentation garnered from various in vitro studies and to determine the relative contributions they make to CTL responses across various forms of antigens and immunologic settings. Thus, cross-presentation appears to be carried by different pathways in various types of cells for different forms under different physiologic settings, which remain to be evaluated in an in vivo physiologic setting.

1. Introduction

Here we will present our view on the current status of the field of “cross-presentation.” This phenomenon has a long history [1], with periods of both controversy and consensus [2]. In the remote past, even its existence was controversial [3,4], as it initially appeared to challenge whether H-2 restriction held for in priming of cytotoxic T lymphocytes (CTLs) [5]. Such controversy predated a clear appreciation for the distinct types of antigen presenting cells (APCs), but gave way to a consensus of acceptance in the 1990s [2], particularly as specific types of dendritic cells (DCs) were recognized as playing a dominant role in this form of CTL priming [6]. The last two decades gave rise to extensive investigation into the cellular mechanism of cross-presentation [7–10], based largely on an in vitro model of DCs derived from GM-CSF-treated cultures of bone marrow (BM) or monocytes [11,12]. Currently, a trend in the field is the study of in vivo cross-presentation performed by DCs [13–15]. As we will see, the complexity of both the process itself and the technical methods required to study cross-presentation by DCs in vivo present substantial challenges in this effort.

DCs comprise several transcriptionally distinct lineages that promote appropriate types of adaptive immune responses to widely differing pathogens [16,17]. DCs are classified into subsets based on their various developmental requirements for transcription factors as well as on the differential engagement of immune modules tailored to combat the specific type of stimuli that evokes the response [18,19]. For instance, viral infections provoke plasmacytoid DCs (pDCs) to produce type I interferons [20–23]. Under inflammatory conditions, monocytes differentiate and acquire DC-like features [24–27]. Classical (conventional) DCs (cDCs) can be further classified into cDC1s and cDC2s, distinguished by differential surface marker expressions and developmental dependences on transcription factors [28]. The cDC2 subset expresses Irf4 [29,30] and includes the Notch2-dependent cDC2s required for IL-23 production in response to Citrobacter rodentium infection [31,32]. cDC2s also show a Klf4-dependent functional arm that acts in promoting type 2 immune responses [33]. By contrast, the cDC1 subset requires Batf3 and Irf8 for development and are required for immunity against tumors and viral infections [34–37]. cDC1s are required for IL-12 production in promoting innate immune protection against Toxoplasma gondii (T. gondii) infection as well [38,39].

Presentation of processed antigens in the form of peptides bound to the major histocompatibility complex (MHC) molecules is the basis for initiation of adaptive immunity by cDCs [40–42]. Endogenous antigens present within the cytosol are presented by MHC class I (MHC-I) molecules through a “classical” pathway involving cytosolic proteasomes and a peptide loading complex (PLC) present in the endoplasmic reticulum (ER) [40]. This pathway operates in all somatic cells and is responsible, for example, of the display of virus-derived antigens on the surface of infected cells for recognition by CTLs, leading to their elimination. Exogenous antigens are subjected to endolysomal processing for presentation onto MHC class II (MHC-II) molecules, which operates in MHC-II expressing cells, primarily B cells, macrophages and DCs [43]. However, priming cytotoxic CD8 T cells against tumors or virally-infected cells requires that exogenous antigens be loaded on MHC-I molecules, a process largely operating in cDC1, at least in vivo [6, 36,44]. Recent studies of cross-presentation have understandably focused on the molecular and cell biology of vesicular trafficking and antigen processing [45]. However, since multiple cell types besides cDC1s are able to cross-present various forms of antigens both in vitro and in vivo, it may be important to keep in mind that there may be diverse molecular pathways of “cross-presentation” and to consider each in appropriate context.

We first will recount the initial discovery and early studies of cross-presentation, noting differences between in vitro and in vivo models and the physiologic setting of the original observations [1]. Moving toward more technical aspects, we next discuss how various forms of antigen are cross-presented differently by distinct cell types. Lastly, we present the current state of understanding for the molecular basis of cross-presentation, emphasizing where evidence for genes has been shown to operate in vivo.

2. Origins and experimental model systems of cross-presentation

We recently reviewed the discoveries of DCs and of cross-presentation [2]. The original identification of DCs dates to 1973 with Steinman’s report of stellate cells [46], followed by several studies that established these cells as a novel immune lineage with distinct APC functions. These cells were found to be distinct from previously known lymphocytes or phagocytes [47], derived from BM and to have a short life span [48], present in spleen [49], and to potently trigger an allogeneic mixed lymphocyte reaction (MLR) [50]. Furthermore, these cells expressed high levels of MHC-II [51] and induced a strong syngeneic MLR [52].

Bevan reported cross-priming in mice in 1976 [1], followed by a series of studies showing that a specialized type of APCs exhibits this ability to process cell-associated antigens for recognition by CD8 T cells [4–6,44]. Molecular studies of cross-presentation were accelerated by the development of an in vitro cell culture system using monocytes or BM cells treated with GM-CSF and, in some studies, IL-4 (GM-DCs) [11, 53–55]. In large part, these early molecular studies of cross-presentation were not concerned with the APC’s identity or with the specific form of antigen, but extended consideration to include presentation of any exogenous antigen, including soluble proteins and microbial antigens, presented by MHC-I molecules on any cell type.

2.1. Discovery of CD8α⁺ DCs as the major cross-presenting cell subset

14 years after the original report of in vivo cross-priming against minor histocompatibility antigens [1], Bevan demonstrated similar in vivo processing of a specific exogenous cell-associated antigen for CTL priming, renamed now as cross-presentation [44]. The identity of the responsible APC remained a mystery. For pathogens that do not productively infect DCs, induction of CTL responses would need to occur through the direct antigen presentation by infected cells [4]. Since somatic cells are unable to drive CTL expansion, this suggested that a specialized APC capable of cross-priming should exist to circumvent a viruses’ ability to bypass adaptive immune responses simply by avoiding the infection of APCs [4].

In agreement, subsequent studies showed that in vivo cross-presentation of tumor or viral antigens was not mediated directly by tumor cells or infected cells, but by BM-derived cells acquiring these antigen exogenously [56,57]. MHC-II-expressing cells in spleen had been shown to cross-present soluble antigens in vitro [58,59], but the identity of these cells had not been determined. Several years later, other studies identified subsets within splenic DCs, distinguished by the expression of CD8α among others [60,61]. Later, CD8α⁺ DCs, but not CD8α⁻ DCs, were identified as being able to selectively cross-present cell-associated antigens in vivo [6,62,63].

2.2. Early studies of cross-presentation and evolving model systems

Much of the work toward understanding the mechanism of cross-presentation has been based on the use of GM-DCs and various cell lines [45,64]. Recent studies have shown that, while they efficiently cross-present many forms of antigens in vitro, GM-DCs are a heterogeneous population of macrophage-like and DC-like cells [65]. Further, GM-DCs are transcriptionally distinct from cDC1s in terms of both development and function [66,67]. In Batf3^−/− mice [36] and Irf8 +32^−/− mice [37], which both lack cDC1s, CD8 T cell responses against viruses and tumors were absent and yet GM-DCs from these mice still develop and are able to cross-present in vitro [15]. These results indicate that at least in vivo, GM-DCs do not compensate to provide cross-presentation for CTL priming in the absence of cDC1s. Furthermore, GM-DCs are unlikely to participate in priming of CTLs against tumors, since the adoptive transfer of GM-DCs into Irf8 +32^−/− mice [37] failed to induce anti-tumor immunity [68]. These results confirm the previous suggestion that GM-DCs rely on antigen transfer to endogenous host cDC1s in the setting of DC vaccine-mediated T cell responses [69,70]. Importantly, in vitro cross-presentation by GM-DCs is not sensitive to the loss of Rab43 or Wdfy4, unlike the dependence on these factors by cDC1s [13,15]. These results indicate that GM-DCs carry out cross-presentation by a distinct molecular pathway from that required by cDC1s in vivo.

GM-DCs have been a mainstay in studies of cross-presentation, but a recent trend has turned attention to primary cDC1s. FLT3L-treated cultures of BM cells (FLT3L-DCs) generate distinct cDC1 and cDC2 subsets that express phenotypic markers and function similarly to their in vivo counterparts [71–74]. For example, the identification of Wdfy4 as an in vivo requirement for cross-presentation by cDC1s was based on the use of FLT3L-cDCs in a CRISPR/Cas9 screen [15]. Notably, Wdfy4 would not have been identified in a screen based on GM-DC cross-presentation, since GM-DCs do not require Wdfy4 for their activity [15]. Alternatively, biochemical assays require large numbers of cells, and many investigators have begun to utilize the mutuDC cell line, which is derived from the transformation of splenic CD8α⁺ DCs in a mouse expressing the SV40 large T antigen as a transgene under the control of the CD11c promoter [75]. Importantly, mutuDCs are capable of cross-presenting in vitro and harbor a similar proteomic signature as cDC1s in vivo [75]. Further, this cross-presentation, specifically of cell-associated antigens, by mutuDCs is dependent on Wdfy4 (Ohara and Murphy, unpublished). Thus, such approaches may allow genetic and biochemical analysis of the cross-presentation mechanisms that operate for cDC1s in vivo.

3. Impact of the form of antigen on the characteristics of cross-presentation

Cross-priming was initially recognized in CTL responses against minor histocompatibility antigens and not against a specific viral or model antigen [1]. Later studies showed in vivo cross-priming operated for ovalbumin (OVA) as well, offered as exogenous soluble peptide, protein or cell-associated forms, although with different efficiencies [44, 76,77]. Since then, the different forms of antigen were recognized as having an impact on various characteristics of cross-presentation by different cell types, for example for soluble versus particulate antigens [78]. The form of antigen may harbor varying physiological relevance depending on the immune context, with genetic and molecular requirements of each varying across different APCs. There are gaps in our understanding of how different types of APCs process various forms of antigen. Below, we review the characteristics of cross-presentation that have been described for different forms of antigen.

3.1. Soluble antigens

While initially described for minor histocompatibility antigens in cell-associated form [1,44], much of the analysis of cross-presentation has examined processing of soluble proteins as a model antigen [58,59]. Indeed, early studies demonstrated that splenic APCs, in the presence of soluble antigens, could induce a robust antigen-specific T cell proliferation in vitro [58,59]. This cross-presentation was sensitive to chloroquine treatment, suggesting that vesicular protein degradation is required for cross-presentation of endocytosed soluble antigens [79]. Subsequently, GM-DCs were also shown to efficiently cross-presentation soluble antigens [11,80]. This cross-presentation was dependent on transporter-associated with antigen processing (TAP) and was sensitive to inhibitors of both the proteasome and the coatomer protein complex I (COP-I) [81]. Furthermore, endotoxin-induced activation was required for optimal cross-presentation of soluble antigens by GM-DCs [82]. Indeed, the increase in cross-presentation of endotoxin-containing soluble OVA was reversed in the absence of TLR4, MYD88, or TRIF. The increased cross-presentation of soluble OVA by TLR signaling suggested that active remodeling of the vesicular compartments may facilitate both processing and loading of the antigens [82,83].

How soluble antigens internalized through macropinocytosis are processed and loaded onto MHC-I molecules remains unclear. The dependence on TAP and proteasomes by GM-DCs might suggest that cytosolic entry and transport to the ER in a classical processing pathway is involved [81]. Indeed, exogenous β₂-microglobulin (β₂m), offered as soluble antigens to β₂m-deficient GM-DCs, was shown to functionally restore surface MHC-I expression and endogenous antigen presentation, suggesting that intact soluble antigens could enter the ER upon endocytosis [78]. Conceivably, an alternate trafficking route could include a retrograde transport from the endosomes, through the Golgi, to the ER. From the ER, soluble antigens could enter the cytosol by ER-associated protein degradation (ERAD) [84], processed by proteasomes and transported back into the ER by TAP for presentation on to MHC-I molecules in a classical route. Whether this mechanism operates in cDC1 cross-presentation of soluble antigens remains to be investigated.

Both cDC1s and cDC2s from the spleen and from FLT3L-treated BM cultures were also capable of cross-presenting soluble antigens in vitro, though splenic cDC2s are less efficient [13,15,73,85]. Notably, in vivo cross-presentation of soluble antigens appears to be primarily mediated by cDC1s [86]. A recent study demonstrated that in vitro cross-presentation of soluble antigens by Tap1-deficient splenic cDC1s and cDC2s was largely intact [15]. Thus, TAP appears to be required for cross-presentation of soluble antigen by GM-DCs, but not by cDCs, suggesting some basic differences in the pathways used by these different cell types.

3.2. Particulate antigens

Here, by particulate antigens, we mean antigens associated with beads of various materials such as iron oxide, latex, or other polymers. Particulate antigens were used to study cross-presentation of antigens internalized through phagocytosis [87,88]. In peritoneal macrophages, uptake of particulate antigens was shown to be more efficient than for soluble antigens and consequently led to better in vitro cross-presentation. Similarly, antigen-specific CD8 T cell responses were more robust in mice immunized with particulate antigens compared with soluble antigen [87,88]. Subsequent studies showed that both CD8α⁺ and CD8α⁻ DC subsets could phagocytose particulate antigens in vivo and in vitro [62,89,90], but did not examine cross-presentation. Importantly, both cDC1s and cDC2s were later found to be capable of cross-presenting particulate antigens in vitro, though cDC2s are less efficient [73].

Similarly, GM-DCs could also cross-present particulate antigens more efficiently than soluble antigens [91]. This cross-presentation was initially TAP-dependent, although higher antigen concentration could bypass this TAP dependency and induce similar levels of antigen-specific CD8 T cell proliferation [91]. Furthermore, studies of NADPH oxidase 2 (NOX2)-deficient GM-DCs reported a reduction in particulate antigen cross-presentation, suggesting the role of phagosomal alkalinization via reactive oxygen species (ROS) for the prevention of antigen degradation [92]. In apparent contrast, some studies reported that GM-DCs are less effective at particulate antigen cross-presentation [73] and that LPS-induced TLR signaling is required [83,93]. The discrepancy between these studies for particulate antigen cross-presentation by GM-DCs may be due to differences in bead composition or antigen coupling method (e.g. passive adsorption, covalent binding, or protein-ligand interaction), which conceivably may affect the characteristics of particulate antigens. Also, as is the case for soluble antigens, endotoxin signaling may positively affect the cross-presentation signal in some experimental settings.

3.3. Antigen-antibody conjugates and immune complexes

Cross-presentation has also been examined for antigens provided in the form of antigen-antibody conjugates or immune complexes. For example, some studies examined responses to antigen conjugated to the monoclonal antibody (mAb) NLDC-145, which recognizes DEC-205, an endocytic receptor selectively expressed by cDC1s [94–96]. Immunization with such antigen-conjugated NLDC-145 induced the proliferation of antigen-specific CD8 T cells and, to a lesser extent, CD4 T cells at a higher efficiency than the soluble form of the same antigen [94–96]. Notably, TAP was required for antigen presentation to CD8 T cells, but not to CD4 T cells. DEC-205-expressing GM-DCs failed to cross-present OVA in response to NLDC-145-OVA conjugate in vitro [95]. The mAb 33D1 recognizes the cDC2-specific DCIR2 surface receptor. Interestingly, administration of 33D1-OVA conjugate also induced proliferation of both CD4 and CD8 T cells, although CD8 T cell proliferation was observed only at the highest dose [96]. In mice expressing human DEC-205, both CD8α⁻ and CD8α⁺ DCs could cross-present antibody-conjugated antigens when expressing the same endocytic receptor [73]. However, cross-presentation was reduced if the target was not expressed by the DC subset (e.g. NLDC-145-OVA cross-presentation is weaker with CD8α⁻ DCs compared to CD8α ⁺ DCs, and 33D1-OVA cross-presentation is weaker with CD8α⁺ DCs) [73].

Immune complexes (IC) were used to examine the role of Fc γ receptors (FcγRs) in cross-presentation and DC maturation [97,98]. Notably, IC are cross-presented more efficiently by several orders of magnitude compared with soluble antigens [99]. Both CD8α⁻ DCs and CD8α⁺ DCs could cross-present antigens provided in ICs [100]. However, CD8α⁻ DCs, but not CD8α⁺ DCs, required the FcγR common γ-chain for in vivo cross-presentation of antigens delivered by ICs [100]. A recent study confirmed that FcγRs are dispensable for in vivo cross-presentation of ICs by CD8α⁺ DCs [101]. Instead, it appears that C1qa is required for both uptake and cross-presentation of ICs by CD8α⁺ DCs [101]. In contrast to in vivo results, in vitro (where complement has been inactivated) both the uptake and cross-presentation of ICs required FcγRs [101]. This might suggest that C1Q functions in vivo in concert with ICs, which may not operate in vitro. Indeed, the addition of recombinant C1Q was sufficient to induce the uptake of ICs by FcγR-deficient DCs in vitro [101]. It remains unclear how cDC1s recognize C1Q-marked ICs, or why cDC2s require FcγRs for IC cross-presentation. Possibly, the signaling domains of FcγRs and the putative C1Q receptor may contribute to cDC maturation which may be required for cross-presentation.

3.4. Microbe-associated antigens

An early study of cross-presentation asked how antigens from microbes entering the cytosol are processed to generate CD8 T cell responses [102]. Escherichia coli (E. coli) and Salmonella typhimurium (S. typhimurium) were efficiently taken up by peritoneal macrophages and cross-presented in vitro, which was interpreted as processing within a strictly vesicular compartment. However, this was based largely on their presumed vesicular life cycle but did not exclude potential cytosolic entry of antigens or bacteria for classical MHC-I presentation. Also, being based on cross-presentation by peritoneal macrophages, not in vivo DC subsets, the relevance to in vivo CD8 T cell priming is unclear. Subsequent reports examined cross-presentation of microbe-associated antigens by splenic DC subsets. Mice infected with live Leishmania major (L. major) or T. gondii displayed strong antigen-specific T cell proliferation in vivo [91]. Interestingly, while T cell activation induced by T. gondii was strictly TAP- and proteasome-dependent, that induced by L. major was completely TAP-independent [91]. Another study examined S. typhimurium, where both cDC1s and cDC2s could phagocytose and cross-present antigens in vivo [103]. In the case of Saccharomyces cerevisiae, a fungal model system, both cDC1s and cDC2s phagocytosed the heat-killed yeast in a Dectin-1-dependent manner and presented antigens to both antigen-specific CD8 and CD4 T cells [104]. Similarly, both cDC1s and cDC2s were shown to phagocytose and cross-present paraformaldehyde-fixed E. coli in a proteasome-dependent manner in vitro [90,105]. In contrast, heat-killed Listeria monocytogenes (L. monocytogenes) was shown to be cross-presented in vitro by splenic cDC1s (and to a lesser extent by GM-DCs), but not by cDC2s [13,15], in a TAP-dependent manner [15]. Notably, in GM-DCs, heat-killed E. coli were efficiently cross-presented in vitro, but that was abrogated in mice doubly deficient in TRIF/MYD88 or TLR2/TLR4 [83]. These results document the enhancement of cross-presentation by TLR signaling. Interestingly, this cross-presentation was intact in GM-DCs deficient in TAP or the protease cathepsin S (CTSS) and only abrogated when doubly deficient in TAP/CTSS [93]. TAP-independence of microbe-associated antigens was also reported for heat-killed Sendai virus, which induced antigen-specific CD8 T cell responses through splenic APCs [106]. Thus, there is a diverse range of findings regarding TAP-dependence or independence for cross-presentation of microbial antigens from various organisms by GM-DCs and cDC2s. Conceivably, microbial factors may influence the way the associated antigens might be processed by various APCs.

3.5. Cell-associated antigens

Cross-priming was discovered as a CTL responses against minor histocompatibility antigens delivered as cell-associated antigens in the immunizing lymphocytes [1]. Thus, the phenomenon has always been closely linked with cell-associated antigens. Later, Bevan developed a technique to avoid direct presentation by the immunizing splenocytes, based on osmotic loading of β₂m-deficient splenocytes [6]. That study was also first to report that in vivo cross-presentation of cell-associated antigens by splenic DCs was TAP-dependent and mediated specifically by the CD8α⁺ subset of DCs. Notably, this selective capability of CD8α⁺ DCs to cross-present cell-associated antigens was not due to selective antigen capture, however, since in this study, both CD8α⁺ and CD8α⁻ DCs were equally capable of antigen uptake in vivo, as tracked by the use of fluorescent beads loaded into the splenocytes. Later studies agreed that CD8α⁺ DCs, but not CD8α⁻ DCs, could cross-present cell-associated antigens in vivo and in vitro [62,90]. However, these studies observed that CD8α⁺ DCs, but not CD8α⁻ DCs, were capable of antigen uptake, as tracked by the chemical dye used to label the splenocytes. The disagreement regarding antigen capture between CD8α⁺ DCs, but not CD8α⁻ DCs, may be due to the use of different forms of cell labeling.

A subsequent study reported that CD8α⁻ DCs could phagocytose cell-associated antigens, although less efficiently than CD8α⁺ DCs [105]. This was accompanied by CD8α⁻ DCs presenting cell-associated antigens on MHC-II molecules, implying that a difference in phagocytosis does not explain why cDC2s do not cross-present cell-associated antigens. While cDC1s may be superior in phagocytosis of cell-associated antigens compared with cDC2s, cDC1s likely possess a specialized capacity for cross-presentation. In agreement, subsequent studies showed that cDC1s express a receptor, C-type lectin domain family 9 member A (CLEC9A), that recognizes necrotic cells operating in cross-presentation [107] and other proteins, SEC22B and WDFY4, that appear to be regulators of vesicular trafficking that contribute to anti-tumor and anti-viral CD8 T cell responses [14,15].

In summary, the physiological relevance of the many different antigenic forms used to study cross-presentation across various cell types to the induction of in vivo CD8 T cell responses remains unclear. For anti-tumor and anti-viral immunity, cDC1-dependent CD8 T cell responses are presumably generated against tumor or viral antigens captured in association with necrotic tumor cells or virally infected cells. It is unclear whether tumor or virally infect cells provide antigens to DCs in a soluble form. What is clear is that cDC1s are required for most tumor rejections and CTL responses to viruses [15,36,37]. Understanding the relevance of each cell type and form of antigens during in vivo cross-presentation will be key in delineating the diverse molecular pathways reported using in vitro assays of cross-presentation.

4. Current molecular pathways of cross-presentation

Models of cross-presentation for antigens are divided largely between those that emphasize a vesicular pathway and those that invoke a cytosolic pathway [45] (Fig. 1). In a vesicular pathway, internalized antigens are processed and loaded onto MHC-I molecules strictly within the endocytic compartments [102]. By contrast, models invoking a cytosolic pathway propose that antigens enter the cell through an endocytic route but, at some point, translocate into the cytosol, degraded by the proteasome, and load onto MHC-I molecules through the classical pathway of TAP-dependent ER transport or through a transport back into the vesicular compartments capable of MHC-I peptide loading [40,108].

Fig. 1.

Model antigens and pathways of cDC1 cross-presentation. cDC1s cross-present many different forms of antigen, including soluble, cell-associated, microbe-associated, particulate, and antibody/immune complex, via several distinct mechanisms. In the ① Vacuolar Pathway, antigens in the phagosomes are processed into peptides by vesicular proteolytic enzymes such as CTSS. In the ② Cytosolic Pathway, antigens in the phagosomes are transported to the cytosol by various mechanisms such as membrane rupture via ROS-mediated lipid peroxidation triggered by CLEC9A signaling and activation of SYK and NADPH oxidase. Cytosolic antigens are then processed by proteasomes and either transported to the ER for classical MHC-I presentation (not depicted) or back to the phagosome by transporters such as TAP. In either pathway, MHC-I molecules stored in the RAB11A⁺ ERC traffic to the phagosomes in a TLR-dependent manner. TAP transporters and PLC traffic to the phagosomes from the ERGIC in a SEC22B-dependent manner. Together, peptides are loaded in the phagosomes and are transported to the cell surface for cross-presentation and cross-priming of CD8 T cells.

4.1. Vesicular pathway in cross-presentation

The vesicular pathway model emerged from the observations that, in peritoneal macrophages, some antigens can be cross-presented independently of cytosolic proteasomes or TAP proteins [91,102,109,110]. First, cross-presentation of microbe-associated antigens was unaffected by cycloheximide or Brefeldin A, compounds that block endogenous MHC-I antigen processing [102]. In agreement, TAP-deficient macrophages could cross-present soluble, particulate, and microbe-associated antigens [109,110]. Additionally, in vitro cross-presentation of live L. major was reported to not be impaired in TAP-deficient GM-DCs [91]. These results collectively suggested that exogenously derived antigens might be processed by a route involving proteases/peptidases in phagosomes, which are then loaded onto MHC-I molecules without any transport to the cytosol.

By contrast, several studies have also used proteasome inhibitors to support the cytosolic pathway model of antigen processing [80,81,111]. However, a caveat to this interpretation arose from a recent study that presented evidence that proteasomes may not always be exclusively located in the cytosol [112]. That study reported that functional proteasomes are present within phagosomes of GM-DCs [112]. Purified phagosomes containing proteasomes were capable of generating peptides for MHC-I molecules competent for cross-presentation. This intriguing finding complicates the interpretation of studies that rely on chemical proteasome inhibition. The mechanism of proteasome delivery into the vesicular compartments remains unknown [112].

Specific molecules have been identified that may contribute to vesicular antigen processing for cross-presentation. Insulin-regulated aminopeptidase (IRAP) was initially demonstrated to play a role in cross-presentation through peptide trimming of antigens internalized in the phagosomes [113]. IRAP-deficient mice exhibited defects in cross-presentation of cell-associated antigens in vivo and IRAP-deficient splenic cDCs also exhibited defects in vitro for particulate antigen cross-presentation. These results suggest that IRAP may be required by cDC1s for efficient MHC-I peptide generation. However, a second study disagreed, reporting that IRAP was dispensable for splenic cDC1 cross-presentation of both soluble and cell-associated antigens in vitro and in vivo [114]. However, a third study found IRAP to be necessary for splenic cDC1 and cDC2 cross-presentation of soluble and microbe-associated antigens in vitro [85]. Notably, this study did not again confirm the requirement for IRAP in vivo, so that its role in cross-presentation of viral or tumor antigens remains to be investigated.

As mentioned above, CTSS, found in phagosomes and lysosomes, has also been implicated in the vesicular cross-presentation pathway, since CTSS-deficient mice showed a partial defect in CTL response against inoculation with virally infected cells [115]. That same study showed that dual deficiency in CTSS and TAP completely impaired the response, implying a combined vesicular and cytosolic/proteasome pathway.

4.2. Phagosome-to-cytosol (P2C) pathway in cross-presentation

“Phagosome-to-Cytosol Pathway” of cross-presentation was postulated by Kovacsovics-Bankowski and Rock in 1995 as a mechanism in peritoneal macrophages for which phagocytosed particulate antigens are transferred into the cytosol for TAP-dependent loading onto MHC-I molecules [111]. Similarly, cytosolic transfer of phagocytosed soluble antigens in peritoneal macrophages was proposed by Reis e Sousa and Germain as a loss of membrane integrity caused by phagocytic overload [116]. These fundamental studies showed that cross-presentation was sensitive to inhibitors of proteasome and COP-I (to prevent ER-golgi transport) and was dependent on TAP [111,116]. Moreover, TAP has also been shown to be required for cross-presentation of cell-associated antigens in vivo [117]. Further, an assay for cytosolic transfer of antigen was developed based on ribosome inhibitors conjugated to beads, with inhibition of new protein synthesis as an indicator [111,116]. All together, the results indicated that cross-presented antigens are processed by the same pathway of MHC-I loading as endogenously synthesized proteins. Indeed, the cytosolic pathway was indirectly documented in vivo by the use of sex-mismatched cell transfer experiments, where CD8 T cell response against a cell-associated H-Y antigen was impaired in mice deficient for LMP7, a subunit of immunoproteasomes [118]. Furthermore, in vivo administration of soluble cytochrome c, an initiator of the intrinsic apoptosis pathway, resulted in a selective depletion of CD8α⁺ DCs [119], suggesting that cDC1s are capable of transferring endocytosed antigens into the cytosol in vivo.

Since the proposal of this pathway, key questions of many subsequent studies include how antigens are transported from the vesicle to the cytosol and whether the peptides produced in the cytosol are loaded onto MHC-I molecules by transport into the ER or by transport into some distinct TAP-containing vesicle. One study proposed that peptide loading occurs in a specialized ER-phagosome fusion compartment [120]. Phagosomes of an immature splenic DC cell line (D1) contained several ER resident proteins in the phagosomes, including TAP and other members of the PLC. This was interpreted as fusion of early phagosomes with the ER immediately after phagocytosis [120], similar to observations in macrophages [121]. Further, purified phagosomes from GM-DCs fed with particulate antigens were self-competent for peptide import via TAP and were capable of inducing CD8 T cell activation in a proteasome- and TAP-dependent manner [120]. Taken together, loading of peptides onto MHC-I molecules was thus proposed to occur in this specialized ER-phagosome fusion compartment, though how antigens were transferred to the cytosol for proteasomal processing and transported into this specialized organelle remained unknown. However, a study challenging this model performed microscopic analyses of GM-DCs and a macrophage cell line (RAW 264.7) but failed to detect any membrane continuity thought to occur during ER-phagosome fusion nor ER markers on the phagosomes, including SEC61 [122]. Thus, the contribution of the ER membrane during phagocytosis remains to be clarified.

Other studies of phagosomes of various cell types suggested that DC phagosomes contain proteins necessary for loading of peptides onto MHC-I molecules. One study using a macrophage cell line (J774) showed that proteasomes and SEC61 are in association with the phagosomes, suggesting that phagocytosed antigens could be retrotranslocated to the cytosol by the SEC61 translocon complex and processed locally by the associating proteasomes [123]. Similarly, phagosomes of a human macrophage cell line (KG-1) also contained MHC-I molecules and all components of the PLC, suggesting that phagosomes in DCs are a sufficient compartment for MHC-I peptide loading [124]. Furthermore, a role for SEC22B in GM-DCs for delivery of PLC components from the ER-Golgi intermediate compartment (ERGIC) to phagosomes was reported [108]. These results support a model in which cytosolically processed antigens are transported back into phagosomes containing MHC-I PLC.

A role of ERAD in cross-presentation has been suggested previously [125,126]. Various ERAD proteins, including p97, BiP, PDI, calreticulin, CHIP, and SEC61, have been shown to associate with internalized soluble OVA. Moreover, siRNA knockdown of SEC61, p97, and CHIP partially reduced cross-presentation of soluble OVA by DC2.4 cell line [125]. ATPase p97, which is involved in retrotranslocation of misfolded proteins during ERAD, was sufficient for antigen from purified phagosomes. A dominant-negative mutant of p97 that inhibits the retrotranslocation complex also inhibited cross-presentation of soluble OVA in human KG1 cells expressing K^b molecules [127]. Derlin-1, another ERAD component, and SEC61β are present in phagosomes of KG-1 cells and co-immunoprecipitate with recombinant p97. Phagosomes purified from DC2.4 cells also contained SEC61β along with internalized exogenous soluble OVA [128]. While, Derlin-1 was not required for cross-presentation of soluble OVA by GM-DCs [129], knockdown of SEC61A1 partially reduced cross-presentation of soluble OVA and OVA-conjugated latex beads by GM-DCs.

SEC61 is a core component of the ER transporter for secretory or transmembrane proteins, and has been implicated in the retrograde transport of misfolded proteins from the ER to the cytosol for degradation [130–132]. This pathway is suggested as a mechanism for how antigens within vesicles gain access to the canonical MHC-I antigen processing machinery in the cytosol. Along with other ER proteins, SEC61 was identified in phagosomes [120,123]. The specific role of SEC61 as a retrograde transporter in cross-presentation has yet to be tested in vivo.

In support of such a vesicular loading model, another study identified a major source of MHC-I molecules in GM-DCs as vesicular reservoirs within the endosomal recycling compartment (ERC) marked by RAB11A [83]. The TLR engagement upon phagocytosis of heat-killed E. coli or LPS-conjugated beads induced a vesicular fusion event, allowing phagosomal antigens access to MHC-I molecules in the ERC [83]. Recruitment of MHC-I molecules to the ERC required RAB11A, and trafficking of the ERC to phagosomes required IκB kinase-mediated phosphorylation of SNAP23 downstream of TLR signaling [83]. GM-DCs deficient in TRIF and MYD88 showed significantly reduced cross-presentation of various antigens containing TLR ligands [83]. Moreover, shRNA-mediated silencing of Rab11a in GM-DCs reduced MHC-I expression in the ERC and cross-presentation of these antigens [83]. Importantly, fluorescent imaging analyses of splenic CD8α⁺ DCs showed similar co-localization of MHC-I molecules with RAB11A. In summary, SEC22B- and RAB11A-mediated delivery of PLC and MHC-I molecules, respectively, supports the notion that peptides could be transported back into phagosomes via TAP for MHC-I loading.

Notably, one caveat of interpretations made regarding the TAP requirement for cross-presentation is the use of cells deficient in TAP. Tap1^−/− mice have reduced MHC-I surface expression and very low numbers of CD8 T cells in the thymus, spleen, blood, and LNs [133]. Decreased numbers of MHC-I molecules alone might explain impaired cross-presentation seen with Tap1^−/− APCs. Interestingly, however, reduction in cross-presentation of microbe-associated antigens by TAP-deficient, M-CSF-treated BM-derived macrophages could be rescued by MHC-I molecule stabilization through low temperature incubation [134]. Similarly in GM-DCs, TAP is required for efficient cross-presentation of particulate and microbe-associated antigens, but that was also rescued by low temperature stabilization of MHC-I molecules [135]. These studies suggest that TAP may be required in cross-presentation to generate a pool of stable MHC-I molecules at the cell surface, such that loading of peptides could also occur within the vesicular compartments onto recycling MHC-I molecules. Table 1.

Table 1.

Factors involved in cross-presentation. Experimental details of the factors covered in this review are listed. Conditions that generated no changes to cross-presentation are not listed.

Factor	Assay	Cell/Mouse	Antigen	Reference
CTSS	In vitro	Ctss−/− GM-DCs	OVA-loaded PLGA microspheres	[115]
	In vitro	Ctss−/− Tap1−/− GM-DCs	Escherichia coli-OVA (heat-killed)	[93]
	In vivo	Ctss−/− Tap1−/− mice	OVA-loaded PLGA microspheres OVA-DAP cells	[115]
CLEC9A	In vitro	Clec9a−/− FLT3L-CD8α+ DCs	OVA-MEF cells (irradiated)	[107,141]
	In vitro	Clec9a gRNA MutuDC1940	OVA-MEF cells (irradiated) F-actin-myosin II-OVA-conjugated latex beads	[141]
	In vivo	Clec9a−/− mice	OVA-MEF cells (irradiated)	[107]
	In vivo	Clec9a−/− mice	F-actin:myosin II: OVA-conjugated latex beads	[141]
	In vivo	Clec9a−/−:Batf3−/− mixed BM chimera	OVA:poly(I:C)-pulsed 5555 BRAFV600E cells (irradiated)	[141]
	In vivo	Clec9a−/− mice	FLT3L gene therapy: B16F10/MC38 tumor models	[158]
IRAP	In vitro	Lnpep−/− GM-DCs	OVA-SF9 cells (necrotic)	[113]
	In vitro	Lnpep−/− FLT3L-CD8α+/CD8α− DCs	Soluble OVA Ab-opsonized Saccharomyces cerevisiae-OVA	[85]
	Ex vivo	Lnpep−/− CD11c+ DCs	OVA-coated latex beads	[113]
	Ex vivo	Lnpep−/− CD8α+/CD8α− DCs	Soluble OVA Antibody-opsonized Saccharomyces cerevisiae-OVA	[85]
	Ex vivo	Lnpep−/− Ly6C+ CD11b+ CD11c+ cells (inflammatory moDCs)	Soluble OVA	[114]
	In vivo	Lnpep−/− mice	OVA-loaded splenocytes	[114]
MYD88	In vitro	Myd88−/− GM-DCs	Escherichia coli-OVA (heat-killed)	[83]
NOX2	In vitro	Cybb−/− GM-DCs	OVA-coated latex beads	[92]
	In vitro	Cybb−/− FLT3L-CD8α+ DCs	OVA-MEF cells (irradiated)	[141]
	Ex vivo	Cybb−/− CD8α+ DCs	Soluble OVA	[145]
	In vivo	Cybb−/− mice	αDEC205-OVA	[92]
	In vivo	Cybb−/− mice	F-actin-myosin II-OVA-conjugated latex beads	[141]
	In vivo	Cybb−/−:Batf3−/− mixed BM chimera	OVA:poly(I:C)-pulsed 5555 BRAFV600E cells (irradiated)	[141]
PPT1	In vitro	Ppt1−/−:Batf3−/− GM-DCs	Soluble OVA OVA-conjugated latex beads OVA-loaded splenocytes (irradiated)	[151]
	In vitro	Ppt1−/−:Batf3−/− OP9-DL1:FLT3L-CD8α+ DCs	Soluble OVA	[151]
	In vivo	Ppt1−/−:Batf3−/− mixed BM chimera	OVA-loaded splenocytes (irradiated) MC38 colon adenocarcinoma tumor model B16F10-OVA melanoma tumor model	[151]
RAB11A	In vitro	Rab11a shRNA GM-DCs	OVA:LPS-conjugated beads Escherichia coli-OVA (heat-killed)	[83,93]
RAB27A	In vitro	Rab27a−/− GM-DCs	OVA-coated latex beads	[144]
RAB43	Ex vivo	Rab43−/− CD8α+ DCs	Listeria monocytogenes-OVA (heat-killed)	[13]
	In vivo	Rab43−/− mice	OVA-loaded splenocytes (irradiated)	[13]
RAC2	In vitro	Rac2 shRNA GM-DCs	OVA-coated latex beads	[145]
SEC22B	In vitro	Sec22b shRNA JAWSII	Soluble OVA OVA-coated latex beads Toxoplasma gondii-OVA (irradiated) Escherichia coli-OVA	[108]
	In vitro	Sec22b shRNA GM-DCs	Escherichia coli-OVA (heat-killed)	[83,93]
	In vitro	Sec22b−/− GM-DCs	Soluble OVA OVA-beads Vaccinia virus-OVA-infected RAW cells (irradiated)	[14]
	Ex vivo	Sec22b−/− CD11c+ DCs	Soluble OVA	[14]
	In vivo	Sec22b−/− mice	3T3-RIPK3-OVA cells (necroptotic) EG-7-OVA tumor model	[14]
SEC61	In vitro	Sec61a1 siRNA GM-DCs	Soluble OVA OVA-coated latex beads	[129]
SYK	In vitro	Syk−/− FLT3L-CD8α+ DCs	OVA-loaded splenocytes (irradiated)	[107]
TRIF	In vitro	Ticam1−/− GM-DCs	Soluble OVA	[82,129]
WDFY4	In vitro	Wdfy4−/− FLT3L-CD24+ DCs	Listeria monocytogenes-OVA (heat-killed)	[15]
	Ex vivo	Wdfy4−/− CD24+ DCs	Soluble OVA Listeria monocytogenes-OVA (heat-killed) OVA-loaded splenocytes (irradiated) OVA-MEF cells (irradiated)	[15]
	In vivo	Wdfy4−/− mice	OVA-loaded splenocytes (irradiated) Cowpox virus infection model West Nile virus infection model 1969 fibrosarcoma tumor model	[15]

4.3. Molecular basis of antigen translocation from phagosomes to cytosol

One model of cytosolic transfer of antigens relies on ERAD, the process for transporting misfolded proteins from the ER to the cytosol. The ER translocon protein SEC61, which is localized to endosomes, was reported to mediate antigen transport across the membrane towards the cytosol [129]. Targeting of SEC61 was carried out by siRNA and by an intracellular antibody that sequesters SEC61 in the ER. This was found to inhibit the transport of soluble OVA into the cytoplasm in GM-DCs, resulting in reduced cross-presentation [129]. By contrast, recent evidence using a specific SEC61 inhibitor, mycolactone, suggested that acute SEC61 blockade had no effect on antigen export to the cytosol in the mutuDC cell line [136]. Instead, prolonged SEC61 blockade in mutuDCs affected cross-presentation indirectly by reducing expression of MHC-I molecules at the cell surface [136].

Another model proposes that continuous leakage of phagosomal contents occur by lipid peroxidation induced by ROS from NADPH oxidase recruitment [137,138]. In addition to the alkalinization of phagosomes induced by NADPH oxidase assembly and ROS production, these studies propose that ROS also oxidizes vesicular lipids in GM-DCs causing phagosome membrane disruption. Additionally, another study reported that aquaporin-3 (AQP3) channel present on the phagosomes directly imports H₂O₂ into the phagosomes to induce lipid peroxidation and membrane damage [139]. Splenic cDC1s from Aqp3^−/− mice showed reduced cytosolic transfer and GM-DCs overexpressing AQP3 showed increased efficiency of cross-presentation for soluble and antibody-conjugated antigens. In support of this hypothesis, a recent study documented marginally greater endosomal membrane damage in splenic cDC1s than in cDC2s at homeostasis [140]. Mass spectrometry-based proteomics of endosomes identified several members of the ESCRT-III complex enriched in cDC1 endosomes as compared to that of cDC2s, suggestive of more frequent endosomal membrane damage in cDC1s. Using mutuDCs, shRNAs against Chmp4b and Chmp2a ESCRT-III subunits enhanced cytosolic transfer of endocytosed β-lactamase and increased cross-presentation of soluble OVA [140].

A recent study proposed a model of P2C antigen transfer in which receptor-mediated NADPH oxidase activation causes lipid oxidation leading to localized phagosome rupture [141]. CLEC9A is a cDC1-specific C-type lectin receptor that facilitates optimal cross-presentation of cell associated-antigens via its recognition of F-actin exposed on necrotic cells [107,142]. Upon recognition of F-actin, CLEC9A induces SYK activation with its intracellular hemITAM motif. Such SYK signaling was found to activate NADPH oxidase and generate ROS specifically in the CLEC9A⁺ phagosomes, resulting in membrane damage and cytosolic transfer of antigens from the phagosomes [141]. FLT3L-cDC1s deficient in NOX2, the catalytic subunit of NADPH oxidase, showed reduced phagosomal staining of galectin-3, a marker of phagosomal damage, but resulted only in a partial reduction in cross-presentation of dead cell-associated antigen in vitro. A microscopic 3D reconstruction of phagosomes in HEK293T cells, under conditions of chimeric CLEC9A intracellular domain activation, showed rupture of phagosomal membranes [141]. However, in vivo analysis of cross-presentation of dead cell-associated antigens in mice with cDC1-specific NOX2 deficiency (in the setting of a mixed BM chimera) revealed only a partial reduction in expansion of endogenous antigen-specific CTLs. The responses to tumors or viral infections has not been examined in vivo in mice with cDC1-specific NOX2 deficiency, but its role in phagosomal rupture and cross-presentation remains an active ongoing area of investigation.

5. In vivo genetic requirements of cross-presentation

In vitro studies using different combinations of antigenic forms and cell types have identified many factors regulating various aspects of cross-presentation. However, in vivo evidence for many is incomplete. While the use of GM-DCs has aided in investigations of cross-presentation, the differences between GM-DCs and cDC1s suggest a need for continued in vivo work [13,65,67]. Induction of CD8 T cell responses in vivo against viruses and tumors relies largely on cross-presentation by the Batf3/Irf8-dependent cDC1 lineage of cDCs [36,37]. In this section, we review the in vivo data for various genetic factors that may regulate cross-presentation and contribute to CD8 T cell responses.

5.1. NOX2

NOX2 (gp91^phox/CYBB) is a member of the NOX family of NADPH oxidases, which are enzymes capable of transporting electrons across membranes and generating superoxides and other ROS [143]. We have discussed the possible role of NOX2 in phagosomal rupture suggested to mediate cytosolic antigen transfer [141]. But, NOX2 was also shown previously to be recruited to phagosomal membranes of GM-DCs, generating ROS that acted to prevent phagosomal acidification and thereby reducing antigen degradation [92]. In agreement, NOX2-deficient mice showed nearly 2-fold reduction in in vivo cross-presentation of OVA conjugated to anti-DEC205 antibody [92]. Subsequently, a study implicated RAB27A, a small GTPase localized to lysosomal vesicles of GM-DCs, in recruitment of NADPH oxidase to the phagosomes [144]. RAB27A-deficient GM-DCs showed similar reductions in in vitro cross-presentation of OVA-coated beads, but no in vivo data was reported [144]. In a follow-up study, the role of NOX2 in phagosomal acidification was confirmed using splenic CD8α⁺ DCs from Cybb^−/− mice and implicated small GTPase Rac2 in the assembly of NADPH oxidase to the phagosomes [145]. Indeed, NOX2-deficient CD8α⁺ DCs exhibited similar reductions in in vitro cross-presentation of soluble OVA [145].

5.2. IGTP

IFN-γ-induced GTPase (IGTP) is a member of the immunity-related GTPases, which are structurally related to the dynamin-like GTPases implicated in membrane events, and localizes to the ER and lipid droplets in GM-DCs [146,147]. Expression of IGTP is inducible on CD11c⁺ DCs in the spleen by the administration of poly(I:C). In GM-DCs, IGTP deficiency seems to increase phagosomal maturation triggered by phagocytosis of OVA-coated beads. Both splenic cDC1s and cDC2s from Igtp^−/− mice exhibited minor defect in lipid droplet accumulation and failed to accumulate lipid droplets in response to poly(I:C) [148]. In vitro cross-presentation of OVA-coated beads was reduced in Igtp^−/− GM-DCs [148]. Importantly, in Igtp^−/− mice, in vivo cross-presentation of OVA-conjugated to anti-DEC205 antibody was reduced by less than 2-fold and OVA:poly(I:C)-loaded splenocytes was reduced by 2-fold [148]. However, the basis for the impact of Igtp on lipid droplet accumulation or on cross-presentation has remained a mystery. Interestingly, in WT GM-DCs, higher lipid droplet accumulation positively correlated with higher cross-presentation efficiency, suggesting that lipid droplets are required for optimal cross-presentation of OVA-coated beads in vitro [148]. These results suggest that IGTP promotes lipid droplet accumulation in DCs triggered by an unknown signal or by phagocytosis of particulate antigens.

How a storage compartment for neutral lipids might influence cross-presentation remains unknown. Lipid droplet biogenesis has been proposed to promote the exit of misfolded proteins from the ER into the cytosol via a direct pore formation pathway [149]. This explanation presumes a transfer of phagocytosed antigens to the ER, from which they would enter the cytosol in a lipid droplet dependent manner. In this way, Igtp^−/− mice may show reduced cross-presentation due to the reduction in ER to cytosol antigen transfer. However, such a transport pathway for phagocytosed antigens to the ER has not been reported. Alternately, the increased phagosomal maturation in ITGP-deficient GM-DCs suggests that ITGP may play a role in slowing antigen degradation to facilitate cross-presentation [148].

5.3. PPT1

Palmitoyl-protein thioesterase 1 (PPT1) is a lysosomal enzyme that hydrolyzes thioester-linked fatty acyl groups for degradation of S-acylated proteins [150]. In vivo cross-presentation of OVA-loaded splenocytes was enhanced by less than 2-fold in the Ppt1^−/−:Batf3^−/− mixed BM chimeric mice, suggesting that PPT1 in cDC1s may function to suppress cross-priming [151]. Further, these mice presented greater antigen-specific tumor-infiltrating CD8 T cell responses and concordant slower tumor growth [151]. Notably, PPT1 is expressed at high levels in cDC1s at steady-state, but Ppt1 expression is significantly down-regulated upon maturation via poly(I:C) treatment ex vivo [151]. This effect correlates with the enhanced cross-presentation seen in PPT1-deficient cDC1s and activated WT cDC1s. However, the mechanism to explain the 2-fold effect of PPT1 deficiency is unclear.

5.4. SEC22B

Vesicle-trafficking protein SEC22B is a member of the soluble N-ethylmaleimide sensitive factor attachment protein receptors (SNAREs), which regulate the final membrane fusion step in vesicle transport [152, 153]. SEC22B knockdown using shRNAs in JAWSII cells inhibited cross-presentation of soluble, particulate, and microbe-associated antigens in vitro. SEC22B localizes to the ERGIC and pairs with Syntaxin-4 (STX4) on plasma membranes and phagosomes, suggesting that proteins of the MHC-I PLC are recruited to the phagosomes through a SEC22B-dependent mechanism [108]. In vivo cross-presentation of cell-associated antigen, in the form of necroptotic 3T3-RIPK3-OVA cells, was reduced by approximately 3-fold in CD11c-Cre Sec22b^fl/fl mice [14]. Further, OVA-specific CD8 T cell response to EG7-OVA tumor was also reduced by around 3-fold with a concordant increase in tumor growth and worsened survival rate. Notably, in vivo cross-presentation of soluble antigen in CD11c-Cre Sec22b ^fl/fl mice was reported to be SEC22B-independent [154]. While a discrepancy between studies remains regarding the in vitro role of GM-DCs with soluble/particulate antigens [14,154], there is no disagreement as to the in vivo contribution of SEC22B to anti-tumor CD8 T cell responses. Formally, it remains to be shown that SEC22B acts in vivo in cDC1s.

5.5. RAB43

Ras-related protein Rab-43 (RAB43) is a member of the Rab family of small GTPases, which regulate intracellular membrane trafficking [155, 156]. Splenic CD8α⁺ DCs from Rab43^−/− mice showed reduction in cross-presentation in vitro of heat-killed L. monocytogenes, whereas cross-presentation of soluble and peptide antigens remained largely intact [13]. In contrast, GM-DCs from Rab43^−/− mice showed no defect in cross-presentation of both cell- and microbe-associated antigens in vitro [13]. Importantly, in vivo cross-presentation of cell-associated antigens was reduced by 3-fold in Rab43^−/− mice [13]. RAB43 localized to the Golgi and non-lysosomal vesicles of FLT3L-DCs, where it may function in the vesicular transport of endocytosed antigen or possibly the components of the PLC [13]. RAB43 seems necessary for optimal cross-presentation of cell-associated antigens in vivo, but whether RAB43 is required for the induction of CD8 T cells against viruses and tumors remains uninvestigated.

5.6. CLEC9A/DNGR-1

CLEC9A is a member of the Group V C-type lectin superfamily, composed of transmembrane receptors that trigger distinct signaling pathways [157]. CLEC9A is highly expressed on cDC1s and recognizes F-actin-myosin complexes exposed on dead cells to facilitate cross-presentation of dead cell-associated antigens in vitro [107,142]. In particular, CLEC9A engagement with its ligands induced phosphorylation of its cytoplasmic hemITAM motif and activates tyrosine-protein kinase SYK [107]. CLEC9A-deficient (Clec9a^gfp/gfp) mice showed partially reduced in vivo cross-presentation of cell-associated antigen [107]. CLEC9A-deficient cDC1s showed no defect in the uptake of necrotic cells in vivo, suggesting that the role of CLEC9A in cross-presentation is predominantly through its signaling domain and that phagocytosis may be compensated for by other redundant phagocytic receptors on cDC1s. CLEC9A signaling was recently reported to promote phagosomal rupture in HEK293T cells and suggested to promote cross-presentation by allowing NOX2-dependent phagosomal release of captured antigens [141]. Indeed, in vivo cross-presentation of F-actin-myosin II-OVA-conjugated latex beads and irradiated OVA:poly (I:C)-pulsed 5555 BRAF^V600E cells was reduced partially in both CLEC9A-deficient (Clec9a^cre/cre) and Cybb^−/− mice, suggesting a role of NADPH oxidase in optimal cross-presentation. So far, it remains unclear if this mechanism operates in anti-viral or tumor responses in vivo. Notably, in the B16F10 and MC38 tumor models, tumor growth was unchanged in CLEC9A-deficient mice and cross-presentation of tumor-derived antigens ex vivo also occurred independently of CLEC9A [158]. Conceivably, there may be redundancy in receptors and inhibitory mechanisms in cDC1s in the induction of anti-tumor CD8 T cell response in vivo. Indeed, secreted form of gelsolin (sGSN) was shown to competitively block CLEC9A binding to F-actin, such that high levels of sGSN may completely negate CLEC9A function in certain tumor models [159].

5.7. WDFY4

WD repeat- and FYVE domain-containing protein 4 (WDFY4) was discovered in an in vitro CRISPR/Cas9 screen for cross-presentation by FLT3L-cDC1s [15]. Mice lacking Wdfy4 showed a complete loss of cross-presentation of cell-associated antigens in vitro and in vivo and consequently failed to prime antigen-specific CD8 T cells against viral infections and immunogenic fibrosarcoma tumor cell lines [15]. Wdfy4^−/− mice uniformly failed to reject strongly immunogenic tumors in vivo, similar to mice lacking cDC1s [15]. Direct presentation of endogenous antigens and MHC-II presentation to CD4 T cells remained intact in Wdfy4^−/− splenic cDCs [15]. Notably, GM-DCs generated from Wdfy4^−/− mice showed no defect at all in cross-presentation in vitro of cell-associated antigens, suggesting that GM-DCs use a WDFY4-independent mechanism of cross-presentation [15].

WDFY4 is a member of a family of nine relatively large, multi-domain proteins containing a highly conserved Beige and Chediak-Higashi (BEACH) domain (Fig. 2) [160], which was originally identified in lysosomal trafficking regulator (LYST), the causative gene for Chediak-Higashi syndrome (CHS) [161–163]. Crystal structures of the BEACH domains of human NBEA and LRBA showed an unique structure that extensively associated with the preceding pleckstrin-homology (PH) domain, suggesting that these two domains may function as a single unit [164,165]. Additionally, with the exception of NSMAF/FAN and WDR81, BEACH domain proteins contain a concanavalin A (ConA)-like lectin domain upstream of the PH-BEACH domain, as identified by in silico sequence analyses and structure modeling [166]. Members of the ConA-like lectin/glucanase domain superfamily, which includes legume lectins, pentraxins, Clostridium neurotoxins, and Trypanosoma sialidase, can bind carbohydrate structures [167], but it is unknown whether this is true for the ConA-like lectin domain of BEACH proteins.

Fig. 2.

Mouse BEACH domain protein family. The domains of the nine BEACH domain proteins in mice are annotated and drawn to scale. The lighter shades within domains represent a short linker between two adjacent domains (e.g. PH-BEACH domains) or insertions unique to the subset of BEACH domain proteins (e.g. ARM). Domain boundaries were determined by protein structure prediction using I-TASSER [184] and RoseTTAFold [185]. Armadillo repeats (ARM); Concanavalin A-like lectin (ConA-ll); Pleckstrin homology-like (PH); Beige and Chediak-Higashi (BEACH); WD40 repeats (WD40); Fab1, YOTB, Vac1, and EEA1 (FYVE); Glucosyltransferases, Rab-like GTPase activators and myotubularins (GRAM).

Several of the BEACH domain proteins are associated with human diseases and are suggested to function as scaffold proteins in various membrane events. Patients with CHS and the beige mice harboring mutation in Lyst are characterized by enlarged granules and defects in protein sorting [168]. Similarly, mutations in LRBA lead to impaired recycling of internalized CTLA4 and consequent lymphoproliferation and autoimmunity [169,170]. Mutations in NBEA, a paralog of LRBA, is associated with neurodegeneration and autism, although no humans with biallelic mutations have been reported, perhaps due to perinatal lethality as seen in mouse models [171–173]. Patients with WDFY3 mutations have been reported to show neurodevelopmental delay and intellectual disability characterized by macrocephaly and microcephaly [174].

ALFY/WDFY3, a paralog of WDFY4, was initially discovered in Drosophila melanogaster as the homologous gene blue cheese (bchs) [175] and later cloned in humans from a partial sequence containing an FYVE domain of a protein with an unknown function [176]. Mutations in bchs led to reduced life span with age-dependent formation of protein aggregates throughout the CNS [175]. Loss of Wdfy3 in mice led to perinatal lethality due to defects in brain development [177,178]. Functionally, ALFY/WDFY3 facilitates the clearance of aggregated protein products through selective macroautophagy; as a scaffold protein, ALFY/WDFY3 interacts with several proteins involved in auto-phagosome biogenesis (p62, Atg5, and GABARAPs) through its PH-BEACH and WD40 repeats domains [179–182]. Unlike WDFY4, WDFY3 harbors an additional domain at its C-terminus, an FYVE domain with weak binding affinity to phosphatidylinositol 3-phosphates, unlike other FYVE domain-containing proteins [183].

The mechanism of WDFY4 in cross-presentation is completely unknown. It appears nearly absolute in its requirement for cross-presentation of cell-associated antigens in cDC1s, but not in GMDCs, suggesting substantial differences in how these cells carry out the process. Knowledge of this requirement also does not favor or exclude any other models of cross-presentation and is essentially agnostic to whether the process in cDC1 is vesicular or cytosolic. In any case, determining how WDFY4 functions in cDC1 cross-presentation may help in further defining this process biochemically.

6. Concluding remarks

Cross-presentation underlies the ability to prime CTLs against antigens arising exogenously to cDCs. During the past decades, many studies of cross-presentation have led to the identification of several apparently different pathways, for example, TAP-dependent and TAP-independent mechanisms, and pathways that appear to differ genetically were carried out by different types of cell models. The mechanisms operating in cell models are incompletely understood, as is the case for in vivo cross-presentation of viral or tumor-derived antigens. Physiologically, the cDC1 represents a cell highly specialized for efficient cross-presentation of exogenously derived cell-associated antigens. Many genes, compartments and routes of antigen trafficking have been proposed to explain the mechanism of cross-presentation. These have been based on results gathered from many studies using many different forms of antigen, different types of APCs, and different read-outs to define cross-presentation. While the in vivo requirement for the cDC1 lineage has been established for efficient cross-presentation in many settings, much about the molecular mechanism for this particular cell remains to be determined. This review has not covered contributions by direct presentation or cross-dressing that may occur in special settings, but rather has focused on what we consider to be the predominant pathway relevant to responses to tumors and most viral infections. From our perspective, cross-presentation by cDC1 is now a problem whose solution will be found in the realm of cell biology, for example, in determining the mechanism by which the enigmatic WDFY4 protein or the important CLEC9A receptor enact their contributions to cross-priming in vivo. The challenge, as we see it, is to ensure that the proposed mechanisms are stringently validated in appropriate model systems in vivo.

Data availability

No data was used for the research described in the article.