The Vacuolar Pathway of Long Peptide Cross-Presentation Can Be TAP Dependent

Cross-presentation is the process through which exogenous Ags are presented by MHC class I (MHC-I) molecules to CD8⁺ T cells. This contrasts with classical Ag presentation, in which peptides derived from exogenous proteins are presented by MHC class II molecules, whereas peptides derived from intracellular proteins are presented by MHC-I molecules. Two main cross-presentation pathways have been described: cytosolic and vacuolar (1–5). In the cytosolic pathway, the internalized Ag needs to reach the cytosol and be processed by cytosolic proteases, such as the proteasome. Then, the resulting peptides are transported by TAP into a vesicle like the endoplasmic reticulum (ER), phagosome, or endolysosome, where they are loaded onto MHC-I molecules and presented at the cell surface. As opposed to the cytosolic pathway, in the vacuolar pathway, there is no need for a cytosolic protease to process the peptide, nor for TAP to transport the cross-presented peptide across membranes, because both the processing and MHC-I loading steps occur in vacuoles. TAP-dependency has therefore been considered as a key criterion to distinguish the cytosolic from the vacuolar pathway of cross-presentation. However, TAP may also indirectly affect cross-presentation by controlling the availability of MHC-I for cross-presentation, irrespective of transport of the cross-presented peptide. In the absence of TAP, most MHC cannot load peptides in the ER and therefore cannot leave the ER because of ER quality control. Some MHC-I, however, like HLA-A2, can load peptides in the ER in the absence of TAP because they can accommodate hydrophobic peptides, among which are signal peptides, which are delivered in the ER by the signal peptide peptidase.

We recently showed that the long gp100_184–227 peptide was cross-presented following a vacuolar pathway, making use of newly synthesized HLA-A2 molecules that are loaded with suboptimal peptides in the ER and then reach the cross-presenting vacuole following an alternative secretion pathway (6). In this article, we studied the cross-presentation of another long peptide, derived from tumor Ag MAGE-A3 and comprising the HLA-A*0101–restricted epitope, MAGE-A3_168–176, by human monocyte-derived immature dendritic cells (Mo-iDCs).

T2 cells were given by A. Hill (Oxford), and T1 cells were given by P. Cresswell (Yale University). The MAGE-A1_161–179–specific CTL clone MZ2-CTL-82/30 and RU1_34–42–specific CTL clone 381/84 were derived in house as described previously (7, 8). The MAGE-A3_168–176–specific CTL clone ERL3-Exvivo-3E5 and gp100_209–217–specific CTL clone LB2686-CTL-811/327.4 were given by P. Coulie (Université catholique de Louvain). Melanoma cell line EB-81-MEL.2 was derived in house as described previously (9). Human Mo-iDCs were derived as previously described from PBMCs (6).

Alexa 568-conjugated goat anti-rabbit Ab was from Life Technologies. Nonconjugated human anti-human HLA-A1 GV5D1 was made in the Department of Immunohematology and Blood Transfusion of Leiden University Medical Center (10). Alexa 633 (Life Technologies) conjugation to GV5D1 was performed in house. Mouse anti-human HLA class I Ab HC10 was given by P. Cresswell (Yale University). Rabbit anti-RAB11A polyclonal Ab was from Abcam (ab3612). FITC-conjugated mouse anti-human CD107a/CD107b and APC-conjugated mouse anti-human CD8 were from BD Pharmingen. Mouse anti-human β-actin AC15 and cytochalasin B were purchased from Sigma-Aldrich.

Target cells were pulsed with 5-μM long peptide for 2–3 h at indicated temperatures in complete medium and cocultured with CTL after washing. CTL activation was evaluated by a cytokine production assay or a degranulation assay. For the cytokine production assay, 45,000 (in IFN-γ assay) or 15,000 (for TNF-α assay) Ag-loaded target cells were cocultured with 15,000 CTLs in round-bottom 96-well plates in triplicates. IFN-γ or TNF-α in the supernatant was then quantified by ELISA after 18 h of coculture. Degranulation assay was performed by coculturing 12,000 target cells with 4000 CTLs at 37°C for 40 min unless indicated otherwise, and then cells were labeled with FITC-conjugated mouse anti-human CD107a and CD107b Ab, and APC-conjugated mouse anti-human CD8 Ab at 4°C for 30 min. The percentage of FITC-positive T cells among the CD8⁺ T cells was determined by FACS. To study the effect of 20°C incubation, degranulation assay was done by coculturing 12,000 target cells with 4000 CTLs in the presence of monensin and FITC-labeled mouse anti-human CD107a and CD107b Ab at 20°C for 2.5 h, as indicated, followed by CD8 labeling with Ab at 4°C for 30 min. To study the effect of inhibitors, unless stated otherwise, dendritic cells (DCs) were treated with drugs for 30 min before and during the 2–3 h of pulse with the long precursor peptides.

EGFP-, HLA-A1-, ICP47-, and full-length gp100-encoding cDNAs were first cloned into the backbone of pST1-A(120) (11). Then, the plasmids were linearized with SapI, and mRNA was prepared as described (12). Electroporation was performed in a 4-mm gap electroporation cuvette using gene pulse Xcell electroporation system (Bio-Rad Laboratories) with 2.5 million DCs resuspended in 200 μl Opti-MEM I no phenol red (Life Technologies) containing 10 μg mRNA, under the following conditions: mode, exponential decay; voltage, 300 V; capacitance, 150 μF; resistance, ∞ Ω, resulting in a pulse time of ≈10 ms.

The recombinant vaccinia virus that expresses full-length MAGE-A3 was provided by V. Cerundolo (Molecular Immunology Group, University of Oxford) and amplified in house by transfecting 293T cells at a multiplicity of infection of 5. Supernatant of transfected 293T cells was collected and titered by plaque-forming assay. To mimic endogenous cytosolic Ag presentation, Mo-iDCs were infected with recombinant MAGE-A3-expressing vaccinia virus at a multiplicity of infection of 25 for 2–2.5 h. Infected cells were then washed thoroughly and tested for their ability to stimulate MAGE-A3_168–176–specific CTL with the same E:T ratio as the cross-presentation assay.

A pool of 3 insulin-degrading enzyme (IDE)–targeting small interfering RNA (siRNA) (IDE-HSS105-178, IDE-HSS105-179, IDE-HSS105-180) was from Invitrogen. Its nontargeting control D-001210-01-20 was from Dharmacon. RAB11a-targeting siRNA (on-target plus SMARTpool siRNA L-004726-00-0005) and its control (nontargeting siControl) were from Dharmacon. DCs were transfected with 1 μM of siRNA on day 3 after differentiation and assayed for their cross-presentation ability 96 h after transfection. Efficacy of knockdown was analyzed by Western blot with cells collected at the same time as the cross-presentation assay. siRNA was delivered into cells by electroporation as described above for mRNA transfection.

Fewer than 2 × 10⁶ DCs were treated with 1 ml of 50 mM citric acid (pH 3) for 2 min on ice, then neutralized with 50 ml of complete medium. After washing, cells were used for cross-presentation or HLA-I expression assay.

A total of 1 × 10⁶ HLA-A1–transfected T2 cells (T2-A1 cells) were prepared in 200 μl of Opti-MEM I (no phenol red). Cells were then mixed with 2 μl of peptide (10 mM stock in DMSO) or DMSO (as mock control) and electroporated immediately under the following conditions: cuvette, 4 mm; mode, square wave; voltage, 500 V; pulse time, 2 ms. Electroporated cells were quickly diluted into 50 ml of culture medium at room temperature to wash away remaining peptides. In the condition of mock control, 2 μl of corresponding peptide stock were added into the cell suspension immediately after electroporation, right before cells were transferred into 50 ml of washing medium. Cells were then collected by gentle spin (100 × g for 10 min) and used for cross-presentation assay.

To study long peptide cross-presentation by HLA-I molecules different from HLA-A2, we chose a 23-aa long peptide (MAGE-A3_158–180) that derives from tumor Ag MAGE-A3 and comprises the HLA-A*0101–restricted epitope MAGE-A3_168–176. After being incubated with Mo-iDCs at 37°C, this long peptide can be cross-presented efficiently (Fig. 1A). We excluded direct loading of the long peptide on HLA-A1 molecules by showing that this cross-presentation was blocked by low temperature (4°C) incubation and cytochalasin B treatment, which inhibits endocytosis by interrupting actin polymerization (Fig. 1A, 1B).

Because the two key features that distinguish the cytosolic and vacuolar pathways of cross-presentation are dependence on cytosolic proteases and dependence on TAP, we first studied the role of cytosolic proteases in cross-presentation of the long MAGE-A3 peptide. We previously reported that processing of the endogenous MAGE-A3_168–176 antigenic peptide does not depend on the proteasome but on a cytosolic metallopeptidase named IDE (13). We reasoned that if the cross-presented long MAGE-A3 peptide was processed in the cytosol, IDE should be involved. Yet, the cross-presentation was not affected by IDE knocking down (>90%), which inhibited the presentation of MAGE-A3_168–176 from endogenously expressed MAGE-A3 protein introduced by vaccinia virus infection (Fig. 2A–C). Cross-presentation of the long MAGE-A3 peptide was also not dependent on aminopeptidases, because it was not blocked by leucinethiol, which did block endogenous presentation of gp100_209–217, presumably by blocking N-terminal trimming of the epitope by ERAP (14, 6) (Fig. 2D). Not surprisingly, proteasome inhibitor MG132 did not inhibit cross-presentation of the long MAGE-A3 peptide either (Fig. 2E, 2F). We thus postulated that, like the long gp100 peptide, cross-presentation of the long MAGE-A3 peptide followed a vacuolar pathway. In agreement with this notion, the cross-presentation of the long MAGE-A3 peptide was inhibited by lysosomotropic agent, chloroquine (Fig. 2G).

We then blocked TAP by introducing viral inhibitor ICP47 into Mo-iDCs through mRNA electroporation. Surprisingly, we observed that cross-presentation of the long MAGE-A3 peptide was largely inhibited (Fig. 3A). To explain this paradox, we considered that, in the vacuolar pathway that we previously described for the long gp100 peptide, the HLA class I molecules that are used for cross-presentation are newly synthesized and sorted directly from the ER to the cross-presenting vacuole. Because of quality control in the ER, it is likely that HLA-I can only exit the ER if they are properly folded and loaded with peptides. In fact, our previous results suggested that HLA-I that followed this alternative secretion route were loaded with suboptimal peptides. ER loading of HLA-A2 with suboptimal peptides occurs readily in the absence of TAP because of the presence of hydrophobic signal peptides that can bind to HLA-A2. However, other HLA-I molecules that, like HLA-A1, do not preferentially bind hydrophobic peptides cannot easily bind peptides in the ER in the absence of TAP and therefore may fail to exit the ER and reach the vacuole. This may explain the paradoxical effect of ICP47 in vacuolar cross-presentation of the long MAGE-A3 peptide: TAP would not be needed for transport of the cross-presented peptide, but to allow HLA-A1 to load peptides in the ER and thereby escape quality control and traffic to the cross-presenting vacuole.

To test this hypothesis, we first tried to increase traffic of HLA-A1 from the ER to the vacuole in TAP-blocked cells. It is known that overexpressing an ER-retained protein may lead to its artifactual escape from ER quality control in mammalian cells (15, 16). We therefore overexpressed HLA-A1 in TAP-blocked DCs by mRNA electroporation, so that the overexpressed HLA-A1 molecules may escape ER quality control and arrive in cross-presenting vacuoles. We evaluated escape of HLA-A1 molecules from the ER by measuring their expression at the cell surface (Fig. 3B). TAP-blocked cells expressed 20–25% less surface HLA-A1 as compared with control cells. The retained surface expression of HLA-A1 likely reflected the presence of pre-existing stable HLA-A1 molecules. HLA-A1 electroporation almost fully restored surface HLA-A1 expression in TAP-blocked cells, confirming ER escape. We then tested cross-presentation of the long MAGE-A3 peptide in these conditions, considering that if the decreased cross-presentation we observed in TAP-blocked cells was due to a lack of transfer of HLA-A1 molecules from the ER to the cross-presenting vesicle, then overexpression of HLA-A1 should at least partially restore the cross-presentation. On the contrary, if the inhibitory effect of TAP blockade on cross-presentation was due to impaired transport of cross-presented MAGE-A3 short peptides across the membrane, then cross-presentation should not be restored by HLA-A1 overexpression. As shown in Fig. 3A, overexpression of HLA-A1 in ICP47-transfected DCs increased cross-presentation of the long MAGE-A3 peptide by ∼3-fold, whereas endogenous presentation of the same MAGE-A3 epitope, expressed in the cytosol from a MAGE-A3-expressing vaccinia viral construct, was not restored at all (Fig. 3A, 3C). This strongly suggests that TAP is required for cross-presentation because it conditions the supply of HLA-A1 to the cross-presenting vacuole.

In our previous study, we showed that TAP-deficient T-B lymphoblast hybrid T2 cells and their parent TAP-competent T1 cells could cross-present the long gp100 peptide in the same pathway as Mo-iDCs (6). We therefore tested HLA-A1–transfected T1 (hereafter referred to as T1-A1) and T2-A1 cells for their ability to cross-present the long MAGE-A3 peptide. For these tests, we measured CTL activation with degranulation assays, because they provide a quicker readout (40 min) as compared with cytokine production assays (overnight), and can be performed at low temperature. As expected, T1-A1 cells displayed good surface expression of HLA-A1 and cross-presented the long MAGE-A3 peptide well (Fig. 3D, 3E, Supplemental Fig. 1). Interestingly, T2-A1 cells, which are TAP deficient and have barely detectable surface HLA-A1, could also cross-present MAGE-A3, although to a lower extent than T1-A1 cells (Fig. 3D, 3E, Supplemental Fig. 1). As discussed above, the retained ability of T2-A1 cells to cross-present the MAGE-A3 long peptide may result from the escape of transfected HLA-A1 from ER retention because of overexpression initiated by the powerful CMV promoter (15, 16). Moreover, this cross-presentation was further increased by incubation of T2-A1 cells at 26°C, which was reported to stabilize MHC-I molecules in TAP-deficient cells (17, 18), even though we did not detect increased surface HLA-A1 by FACS upon 26°C incubation (Fig. 3D, 3E). These results were consistent with the notion that TAP was required for the supply of HLA-I to the vacuole but not for transport of the cross-presented peptide.

To further confirm this notion, we tried to rescue empty HLA-A1 molecules from the ER of TAP-deficient cells by delivering HLA-A1–binding peptides into the ER directly through electroporation. We presumed that this was possible with T2 cells, because they have a small cytosolic volume filled with ER (Supplemental Fig. 2). As shown in Fig. 4A, T2-A1 cells electroporated with HLA-A1-restricted short peptides MAGE-A1_161–169 (7) or MAGE-A3_168–176 and washed could activate the relevant CTLs very well, whereas cells that were electroporated without peptides, then similarly snap-pulsed (1.5 min) with the peptide, or cells that were snap-pulsed without electroporation, could not (Fig. 4A, Supplemental Fig. 3). These results strongly suggested that at least part of the electroporated short peptides was delivered directly into the ER to load and stabilize empty HLA-A1 molecules. Because T2-A1 cells express very little HLA-A1 at the cell surface (Fig. 3E), organelles such as endosomes and recycling vesicles should contain few HLA-A1 molecules. Therefore, even if the electroporated peptide could also reach these organelles, it would not be able to efficiently load HLA-A1. We then studied the effect of different ER-delivered peptides on cross-presentation of the MAGE-A3 long peptide by T2-A1 cells. We used two peptides predicted to bind HLA-A1 with a moderate affinity (MAGE-A1_161–169 and MAGE-A3_{168–176I172C}, a modified MAGE-A3_168–176 that is no longer recognized by the CTL) and two HLA-A2–binding peptides (MAGE-C2_191–200 and LLAAWTARA), which should not bind HLA-A1 (9, 6). We observed that cross-presentation of the MAGE-A3 long peptide by T2-A1 cells was increased by electroporated HLA-A1–binding peptides but not by HLA-A2–binding peptides (Fig. 4B). This observation further confirmed that TAP-dependency of the long MAGE-A3 peptide cross-presentation reflected the need for TAP to make HLA-A1 molecules available for cross-presentation. In other words, the processed epitope itself does not need TAP-mediated transport, in agreement with the notion that the peptide is processed in the vacuole.

To investigate the source of HLA-I for cross-presentation of the long MAGE-A3 peptide, we tested the effect of 20°C incubation, which blocks both recycling and classical secretion of HLA-I molecules (6). We found that 20°C incubation did not affect cross-presentation of the MAGE-A3 long peptide by Mo-iDCs (Fig. 5A), whereas it successfully blocked the classical secretion pathway of HLA class I, which goes from the ER through the Golgi stack and trans-Golgi network to the cell surface (Fig. 5B). Our previous results also showed that 20°C incubation fully blocked presentation of conventionally processed endogenous Ags to CD8 T cells (6). Because we previously showed that 20°C incubation fully blocked both internalization and recycling of HLA class I molecules, these results indicated that the cross-presentation of the long MAGE-A3 peptide made use of nascent HLA-I that follow a nonclassical secretion pathway. In line with this notion, the cross-presentation was also inhibited by protein synthesis inhibitor, cycloheximide (Fig. 5C), and protein secretion inhibitor, brefeldin A (BFA) (Fig. 5D). This inhibitory effect could not result from indirect inhibition of HLA-A1 recycling due to possible decreased surface HLA-A1 level, because surface HLA-I of Mo-iDCs were barely decreased after treatment with cycloheximide or BFA for 2 h, the time needed for the cross-presentation assay (Fig. 5E).

To further confirm the dispensability of recycling HLA-A1 molecules for cross-presentation, we stripped off surface HLA-I of Mo-iDCs with citric acid (pH 3) and performed the cross-presentation assay at 20°C to limit the reappearance of surface HLA-A1 occurring at 37°C. We found that, although more than 90% of surface HLA-A1 were successfully stripped off, cross-presentation of the long MAGE-A3 peptide was not affected (Fig. 5F, 5G).

In our previous report, we also excluded the involvement of recycling HLA-I molecules in cross-presentation of the long gp100 peptide (6), and we monitored surface HLA-I molecules using conformation-specific anti-HLA-I mAb W6/32, which can only detect peptide-loaded HLA-I molecules. However, there is also a small proportion of free heavy chains (FHC) of HLA-I molecules at the cell surface. Although internalized FHC are probably destined for degradation, recycling FHC have been proposed to play a role in Ag cross-presentation (19, 3). To investigate the role of recycling FHC in the cross-presentation of the MAGE-A3 long peptide, we checked the level of cell surface FHC before and after acid strip with FHC-specific murine mAb HC-10 (20). We found a dramatic increase of FHC signal after acid strip (Fig. 5H). Because the cross-presentation was not influenced by acid strip (Fig. 5F), we conclude that recycling FHC should not be the source of HLA for cross-presentation.

Another possible source of recycling HLA-I is the “slow recycling” route, which is proposed to be regulated by RAB11A, with a half-time of around 20 min or even longer. This RAB11A-modulated slow recycling route of HLA-I has recently been shown to play an important role in Ag cross-presentation (21, 22). They showed that in murine bone marrow–derived dendritic cells, upon TLR activation, phagosomal MHC-I enrichment and cross-presentation of peptides derived from phagocytic cargo were abolished by expression of dominant negative RAB11A or by Rab11a silencing with siRNA. Because the recycling assay we used in our previous studies can only monitor the “fast recycling” route, with a half-time of around 4–5 min (23, 6), we further studied the role of slow recycling HLA-I in cross-presentation by knocking down RAB11A in Mo-iDCs with siRNA. No inhibition of cross-presentation of the long MAGE-A3 peptide was observed with more than 95% knocking down of RAB11A (Fig. 6A, 6B). At the same time, we also checked the effect of RAB11A knocking down on the cross-presentation of gp100_184–227 long peptide by HLA-A2. Again, no significant difference was observed between RAB11A knocked down cells and the control cells (Supplemental Fig. 4). These observations showed that the slow recycling route of HLA-I does not play a role in long peptide cross-presentation. We can therefore conclude that the cross-presentation makes use of nascent HLA-I.

Altogether, we conclude that, similar to the long gp100 peptide we reported previously, cross-presentation of the long MAGE-A3 peptide occurs in the vacuolar pathway, making use of newly synthesized HLA-I molecules that follow an alternative secretion pathway. Its paradoxical TAP-dependency results from the fact that HLA-A1 molecules cannot be loaded in the ER in the absence of TAP and therefore cannot exit the ER to proceed to the cross-presenting vacuole. But the cross-presented peptide itself does not need to be transported by TAP.

We previously showed that cross-presentation of the long peptides that derive from gp100 or Melan-A and comprise HLA-A*0201–restricted epitopes follows a vacuolar pathway. It makes use of peptide-preloaded nascent HLA-I molecules that follow an alternative secretory pathway to reach the cross-presenting vacuole, where preloaded peptides are exchanged for the cross-presented peptides before transfer of the final peptide/HLA complexes to the cell surface. In this article, we show that the cross-presentation of another long peptide that derives from tumor Ag MAGE-A3 and comprises an HLA-A*0101–restricted epitope follows the same vacuolar pathway, suggesting that it can be a general pathway used by DCs for long peptide cross-presentation. This pathway makes use of an alternative secretory route for nascent HLA-I molecules. The fact that this route can be followed by both HLA-A*0201 and HLA-A*0101 molecules in both Mo-iDCs and T2 cells suggests that it may be a widely used secretory route worthy of further studies.

Another interesting observation is that the cross-presentation of the long MAGE-A3 peptide is sensitive to TAP blockade, even though it follows a vacuolar pathway. This is due to the fact that newly synthesized HLA-A1, which are needed to cross-present the epitope derived from the long MAGE-A3 peptide, can only be loaded with TAP-dependent peptides in the ER. When TAP is blocked, newly synthesized HLA-A1 molecules in the ER are kept empty and therefore cannot exit the ER and traffic to the vacuole for cross-presentation. In line with this is the fact that, unlike HLA-A2, HLA-A1 surface expression cannot be restored by incubation of TAP-deficient cells at 26°C, indicating that in the absence of TAP, very few HLA-A1 molecules can escape ER retention (Fig. 3E). Our results illustrate a dissociation between dependency on cytosolic proteases and TAP-dependency, two key features that have been used in association to distinguish the cytosolic and the vacuolar pathways of cross-presentation. The only exception so far in that respect came from the group of Peter van Endert who studied the cross-presentation of phagocytosed Ags by murine DC. They described a proteasome-dependent and “TAP-independent” pathway, in which TAP-deficient cells were unable to cross-present, but this inability could be relieved by incubating cells at 26°C (24). However, this pathway of cross-presentation was different from the one we describe in this article, as it was cytosolic (proteasome-dependent) and implied the involvement of an unknown transporter to import peptide from the cytosol into the loading vacuole, in which recycling MHC-I molecules were proposed to be loaded. The authors proposed shortage of surface MHC-I molecules as the reason TAP-deficient DC could not cross-present at 37°C, whereas 26°C incubation increased the amount of surface MHC-I available for recycling. Our pathway rather seems to involve newly synthesized MHC-I that follow a nonclassical secretory route, and whose supply to the loading vacuole is impaired in the absence of TAP due to ER quality control. Yet, a common conclusion can be drawn from the study of these two different model systems, which is that TAP-dependency can no longer be used as a key criterion to distinguish the cytosolic and the vacuolar pathways of cross-presentation: the work of van Endert et al. showed that cytosolic cross-presentation can be TAP independent, whereas our work shows that vacuolar cross-presentation can be TAP dependent. We believe this conclusion could have major implications in the field of cross-presentation and could lead to reconsidering the conclusions of previous studies regarding TAP-dependency of cross-presentation in a number of model systems.

The pathway used for cross-presentation may vary according to the form of the Ag, the nature of the APCs, and their species of origin (reviewed in Ref. 25). This may explain the different conclusions from studies performed in different model systems. In general, the cytosolic pathway of cross-presentation was the most frequently reported in published studies, most of which were performed in mice. In our study, we used human DCs, and we selected Mo-iDCs. The reason for this choice was mostly practical: Mo-iDCs can reproducibly be isolated from blood monocytes, whereas it is much more difficult to work with human DC from other sources, even though some recent reports describe such studies (26). Although they are differentiated in vitro, Mo-iDCs are representative of a subset of DCs present in vivo, named inflammatory DC, which differentiate from monocytes recruited to inflammatory sites and play important roles in eliciting immune response against pathogens, vaccines, and tumors undergoing chemotherapy or radiotherapy. Our study of the mechanism of Ag cross-presentation by Mo-iDCs can help better understand the role of inflammatory DCs in the aforementioned conditions and may provide guidance to optimize therapies. In fact, a recent report published during revision of this manuscript confirmed that human in vivo-generated monocyte-derived inflammatory DCs cross-present exclusively through a vacuolar pathway, confirming the relevance of our findings (27).

Our results further stress the importance of the vacuolar pathway of cross-presentation in Mo-iDCs and inflammatory DCs, particularly for long peptides, and provide a detailed analysis of the cellular biology aspects of this pathway, highlighting the existence of an alternative secretory route for MHC-I molecules, the characterization of which will be worthy of further studies.

We thank the Centre d’hématologie of the Cliniques universitaires Saint-Luc (Brussels) for providing blood samples from hemochromatosis patients, and Auriane Sibille for editorial assistance.

The authors have no financial conflicts of interest.