Long-Peptide Cross-Presentation by Human Dendritic Cells Occurs in Vacuoles by Peptide Exchange on Nascent MHC Class I Molecules Free

Cross-presentation is the process whereby antigenic peptides derived from exogenous Ags are presented by dendritic cells (DCs) on their MHC class I (MHC-I) molecules. This process is needed for the induction of CD8 T cell responses to Ags derived from tumors or from pathogens that do not infect DCs (1–3). Cross-presentation is key to the development of vaccines aimed at inducing CD8 responses against cancer or HIV. A promising vaccine approach pioneered by Melief et al. (4, 5) is based on synthetic long peptides, which need to be cross-presented and therefore results in efficient presentation of the antigenic peptide by DCs only, as opposed to short peptides, which can load any cell and therefore induce tolerance. Therapeutic vaccines based on long peptides have shown clear promise in preclinical models (6), and recently showed clinical efficacy in patients with human papillomavirus–induced neoplasia (7). Contrasting with the advanced clinical development of long-peptide vaccination, relatively little is known about the pathway used for cross-presentation of long peptides (8–11).

The pathway used for cross-presentation appears to vary according to the nature of the exogenous Ag. In the mouse, two main cross-presentation pathways have been proposed: cytosolic and vacuolar. In the cytosolic pathway, the Ag internalized by endocytosis is transferred to the cytosol, from which it follows the classical MHC-I processing route. This involves processing by the proteasome and translocation of the resulting peptide by transporter associated with Ag processing (TAP) into the endoplasmic reticulum (ER), where it combines with newly synthesized MHC-I molecules. A proposed variation of the cytosolic pathway involves TAP-mediated transport of the cytosolic peptide back to the phagosome or the endolysosome, where it would combine with MHC-I molecules. The latter can be routed to phagosomes or to endolysosomes by SEC22B or CD74 (invariant chain), respectively (12, 13). Until recently, the missing piece in the cytosolic pathway was the transporter in charge of the transfer of Ags from the endosome to the cytosol. However, recent evidence has implicated SEC61 in this process (14). In the vacuolar pathway, there is no need for Ag transfer across membranes, because the whole processing happens in vacuoles, including cleavage of the internalized protein by endolysosomal peptidases and loading of the resulting peptide onto MHC-I molecules. This pathway is therefore independent from the proteasome and from TAP. Mechanistic studies of the vacuolar pathway are limited. It is unclear whether MHC-I molecules used in the vacuolar pathway are recycled from the cell surface or are newly synthesized (13, 15–19).

T2 cells were given by A. Hill (Oxford University), and T1 cells were given by P. Cresswell (Yale University). We used two gp100_209–217-specific CTL clones: EB81-CTL-606C/2.1 (clone 7) (20), which was used in Figs. 1 and 2A–C, and LB2686-CTL-811/327.4 (given by P. Coulie, de Duve Institute), which was used in Figs. 2D–G, 3, and 4. The RU1_{34 – 42}-specific CTL clone 381/84 was described previously (21). The Melan-A_26–35-specific CTL clone CTL CP50-549/18 was given by P. Coulie. CTL clone A10, which recognizes peptide MAGE-A3_168–176 presented by HLA-A1, was derived in-house (22).

PBMCs were isolated from whole blood of hemochromatosis patients with Ficoll gradient (Lymphoprep; Axis-Shield PoC), after approval from the Commission d’Ethique Biomédicale Hospitalo-Facultaire from the Université Catholique de Louvain institutional review board. Monocytes were enriched on Percoll as described (23) and isolated by adherence. After washing away nonadherent cells with PBS, adherent cells were kept in culture in RPMI 1640 medium supplemented with 10% FCS, 200 U/ml human IL-4 (made in-house) and 70 ng/ml GM-CSF (Leukine [sargramostim]). The cultures were fed with fresh medium and cytokines every 2–3 d. The monocyte-derived immature DCs (Mo-iDCs) were harvested and used on days 5 to 7 of differentiation.

Fluorochrome-conjugated secondary Abs were obtained from Life Technologies. Alexa-633 (Life Technologies) conjugated and nonconjugated murine mAb W6/32 were produced in-house. FITC-conjugated mouse anti-human CD107a/CD107b, and FITC-conjugated anti-HLA-A2 murine mAb BB7.2 were from BD Pharmingen. Mouse anti-human β-actin AC15 and mouse anti-human SEC22B were from Sigma-Aldrich. Rabbit anti-human CD74 polyclonal Ab Matilda was given by P. Cresswell (24). All inhibitors were obtained from Sigma-Aldrich. Long peptide gp100_184–227 was given by P. Cresswell. The other peptides were synthesized in-house.

Target cells were pulsed with long peptide (5 μM) for 2–4 h at the indicated temperatures in complete medium, and cocultured with CTL after washing. CTL activation was evaluated in a cytokine production assay or a degranulation assay. For IFN-γ and TNF production assay, 40,000–50,000 Ag-loaded target cells were cocultured with 10,000–20,000 CTLs in round-bottom 96-well plates in triplicate. Then IFN-γ or TNF was quantified by ELISA in the supernatant after 18 h. Degranulation assay was done by coculturing target cells with CTLs in the presence of monensin and FITC-labeled mouse anti-human CD107a and CD107b Ab at 20°C. The percentage of FITC-positive T cells among the CD8⁺ T cells was determined by FACS after 3 h of coculture. To study the effect of inhibitors, unless stated otherwise, DCs were treated with drugs for 30 min before and during the 2 h of pulse with the long precursor peptides.

EGFP-, gp100-, and ICP47-encoding cDNAs were first cloned into the backbone of pST1-A(120) (25). Next, the plasmids were linearized with SapI, and mRNA was prepared as described (26). Electroporation was performed in a 4-mm gap electroporation cuvette using gene pulse Xcell electroporation system (Bio-Rad) 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.

CD74-targeted 3-RNAi mix (CD74HSS190576, CD74HSS190577, CD74HSS190578, equally mixed) and its control (nontargeting negative control medium GC 12935-300) stealth small interfering RNA (siRNA) were from Invitrogen. Sec22b-targeted siRNA (on-target plus SMARTpool siRNA L-011963-00-0005) and its control (non-targeting SiControl) were from Dharmacon. To knock down CD74, DCs were transfected with 1 μM siRNA on day 3 after differentiation and assayed for their ability of cross-presentation 72 h after transfection. To knock down Sec22b, DCs were transfected with 1 μM siRNA on day 4 after differentiation and assayed for their ability of cross-presentation 48 h after transfection. Efficacy of knocking down was analyzed by Western blot with cells collected at the same time as the cross-presentation assay. siRNA were delivered into cells by electroporation as described above for mRNA transfection.

Mo-iDCs were first incubated with murine mAb W6/32 (40 μg/ml) at 4°C. After thorough washing, cells were incubated at 20°C or 37°C for 12 min in the presence of 220 μM of primaquine. Next, the cells were stripped with 50 mM citric acid (pH 3) for 2 min on ice and neutralized immediately with 150 mM NaH₂PO₄ (pH 10.5). After washing, cells were incubated at 37°C or 20°C. Aliquots were removed at various intervals and placed on ice. Recycled HLA-I/Ab complexes were detected with flow cytometry using Alexa-647 conjugated Donkey anti-mouse IgG.

For HLA-class I secretion analysis, DCs were first stripped with 50 mM citric acid for 2 min. After neutralization and wash, cells were incubated at 20°C or 37°C in culture medium. Aliquots were removed at various intervals and placed on ice. Reappearance of HLA-class I molecules was detected with Alexa-633 conjugated mAb W6/32 and analyzed by flow cytometry.

Lentiviral vectors were derived from pCCLsin.PPT.hPGK.GFP (27). Control pTM895 was generated by inserting, in the backbone pCCLsin.PPT.hPGK.GFP, an IRES from TMEV (28) and the eGFP coding sequence downstream of the PGK promoter. The ICP47-coding cDNA was cloned from pBJ1-Neo (gifted by H.G. Rammensee, University of Tübingen) into pTM895 upstream the eGFP-IRES sequence. pTM945 vector, which was used to deliver peptides into the ER in a TAP-independent manner, was generated by inserting in the backbone of the same vector: a CMV promoter, a multicloning site, an IRES from TMEV and an mCherry coding sequence (28). The coding sequences of IL-2 signal sequence-peptide fusion polypeptide were cloned into the multicloning site of pTM945. Vectors pTM897, pTM895, and pTM945 were provided by Prof. Thomas Michiels (de Duve Institute, Brussels, Belgium). Lentiviral particles were produced upon cotransfection of HEK-293T cells with vector plasmid, and plasmids pMDLg/pRRE#54, pRSV-Rev and pMD2.VSVG (provided by Luigi Naldini, Ospedale San Raffaele, Milan, Italy) (27). Helper lentiviral particles SIVmac were produced upon cotransfection of HEK-293T cells with envelope plasmid pMD2.VSVG and SIVmac package construct pSIV3+ (gifted by Prof. Cimarelli, Ecole Normale Supérieure de Lyon, France) (23). To transduce Mo-iDCs, cells were incubated with both viral particles on day 4 of differentiation. Next, cells were harvested and analyzed on day 7. T2 cells were transduced with lentiviral particles by spin infection at 2,400 rpm and 32°C for 90 min in the presence of 8 μg/ml of polybrene. Forty-eight hours after transduction, mCherry-positive cells were sorted, seeded to 2.5 10⁵ cells/ml, and cultured for 24 or 48 h. Cross-presentation assay was then performed.

For the reconstitution of cell surface HLA-A2, T2 cells were first stripped with 50 mM citric acid for 2 min. After neutralization and washing, cells were incubated with β2-microglobulin (2.5 μM) and short peptides (10 μM) at 18°C for 20 h. For peptide affinity estimation after acid-stripping, T2 cells were incubated with β2-microglobulin (2.5 μM) and serially diluted peptides at 18°C for 3 h. Next, surface HLA-A2 was detected with FITC-conjugated conformation-dependent anti–HLA-A2 murine mAb BB7.2, and analyzed by flow cytometry. Peptide concentrations were then plotted against the mean fluorescence intensity of surface HLA-A2. The concentration needed to restore surface HLA-A2 molecules to 50% of the maximum was calculated (EC₅₀). The maximum level of HLA-A2 stabilization achieved with each peptide relative to that obtained with peptide gp100_205-217 was also calculated (relative efficacy).

We used a 44-aa-long peptide derived from melanocytic protein gp100 (gp100_184–227) to study cross-presentation of long peptides by hDCs (Fig. 1). After a 2-h incubation with this peptide at 37°C, HLA-A2–positive Mo-iDCs stimulated the production of IFN-γ by a CTL clone recognizing the HLA-A*0201–restricted epitope gp100_209–217 comprised in long peptide gp100_184–227. We observed no cross-presentation when the long peptide was incubated with Mo-iDC at 4°C, or with fixed Mo-iDC (Fig. 1A, B). We performed a two-step experiment to exclude the possibility that the long peptide was processed in the extracellular milieu and loaded directly onto surface HLA-A*0201 molecules in our experimental setup. We first incubated the long peptide for 2 h at 37°C with HLA-A*0201–negative Mo-iDCs. We then collected and incubated the supernatant with HLA-A*0201–positive Mo-iDCs in a cross-presentation assay performed at 4°C or 37°C. If the long peptide was processed in the extracellular milieu during the first incubation, the supernatant of this first step should activate the CTL when loaded at 4°C onto HLA-A*0201-positive DC in the second step. This was not the case (Fig. 1C). This result therefore excluded the processing of the long peptide by extracellular proteases. To strengthen this conclusion further, we blocked endocytosis by inhibiting actin polymerization with cytochalasin B (Supplemental Fig. 1A). We observed a dose-dependent inhibition of cross-presentation of the long peptide. Because cytochalasin B inhibits endocytosis but not proteolysis, this result further confirmed that cross-presentation of the long peptide in our system depends on endocytosis and involves an intracellular processing step.

To determine whether long-peptide cross-presentation followed a cytosolic or a vacuolar pathway, we tested the effect of proteasome and TAP inhibition. We observed that cross-presentation of the long gp100 peptide was not reduced by inhibition of the proteasome with epoxomicin (Fig. 1D) or MG132 (Supplemental Fig. 1B), whereas endogenous presentation of the same gp100 peptide introduced by mRNA electroporation was blocked by epoxomicin (Fig. 1D). TAP blockade with viral inhibitor ICP47, which was introduced into the cells by mRNA electroporation, did not inhibit the cross-presentation of the long gp100 peptide either, whereas endogenous presentation of an ubiquitous self-peptide (RU1_34–42) was blocked by ICP47 (Fig. 1E). We concluded that cross-presentation of the long peptide occurred in the vacuolar pathway.

We were intrigued by the consistent 2-fold increase of cross-presentation observed after TAP inhibition (Fig. 1E). We performed an extensive phenotyping of Mo-iDCs to verify that ICP47 transfection did not activate the production of IL-10 and IL-12, nor did it modify the expression of costimulatory molecules (Supplemental Fig. 2). Although it is well established that DCs are the most efficient cross-presenting cells, other cells might cross-present with a low efficiency. We reasoned that if the cross-presentation of long peptides was indeed favored by TAP blockade, it might be increased or revealed in other cells by the loss of TAP. We therefore compared the cross-presentation of the long gp100 peptide by TAP-deficient T2 cells and their parental TAP-competent T-B lymphoblast hybrid cells T1 (29). We found that T2 cells cross-presented long peptide gp100_184–227 quite efficiently in a proteasome-independent manner, and much better than its parental cells T1 (Fig. 2A, B). The cross-presentation ability of T2 cells was not modified by ICP47 transfection, which excluded off-target effects of ICP47 and further validated TAP inhibition as responsible for the increased cross-presentation observed after transfection of ICP47 in Mo-iDCs (Supplemental Fig. 3A). Because TAP blockade results in a shortage of peptide supply in the ER, we considered two possible explanations for the increased cross-presentation: either more empty HLA-A*0201 molecules became available in the ER to load the cross-presented epitope, provided it can reach the ER, or the ER contained more HLA-A*0201 molecules loaded with suboptimal peptides derived from signal peptides (30–33). Those suboptimal HLA-peptide complexes might then exit the ER and reach another compartment, where the preloaded suboptimal peptides would be exchanged for the cross-presented epitope.

If the increased cross-presentation after TAP inhibition was due to more empty HLA-A*0201 molecules in the ER, then blocking the release of signal peptides into the ER lumen by inhibiting signal peptide peptidase (SPP) should result in even more empty HLA-A*0201 molecules and a further increased cross-presentation. Conversely, if increased cross-presentation was due to more HLA-A*0201 molecules loaded with suboptimal signal peptides, inhibiting signal peptide release from the ER membrane should reduce suboptimally loaded HLA-A*0201 molecules and reduce cross-presentation in TAP-deficient cells. To inhibit SPP, we treated Mo-iDCs with (Z-LL)2-ketone before electroporating ICP47 or EGFP mRNA and incubating with long peptide gp100_184–227. SPP inhibition completely prevented the increase of cross-presentation induced by TAP inhibition (Fig. 2C, left), which was confirmed by the failure of the very same ICP47-transfected DCs to present the HLA-A*0101–restricted epitope MAGE-A3_168–176 provided endogenously with a vaccinia vector (22, 34) (Fig. 2C, right). The lack of effect of (Z-LL)2-ketone on control cells transfected with EGFP instead of ICP47 indicated that the effect of (Z-LL)2-ketone on ICP47-transfected DC was not caused by toxicity or inhibition of the secretion of HLA-I or costimulatory signals. This was further confirmed by analyzing the expression of costimulatory molecules such as CD80 and CD86, and the production of cytokines such as IL-10 and IL-12 by (Z-LL)2-ketone treated cells (Supplemental Fig. 2). That (Z-LL)2-ketone efficiently reduced the number of suboptimally loaded HLA-A2 molecules in TAP-deficient cells was confirmed by the decreased expression of HLA class I—which are mainly HLA-A2 in these cells—at the surface of T2 cells treated with (Z-LL)2-ketone (Fig. 2D). Thus, these results supported the second model positing that increased cross-presentation observed after TAP inhibition resulted from increased abundance of suboptimally loaded HLA-A*0201 molecules available for peptide exchange. The cross-presented epitope is seemingly not delivered into the ER, otherwise it would be able to load empty HLA-A*0201 molecules in the ER and cross-presentation should be increased by (Z-LL)2-ketone. It follows that peptide exchange on suboptimally loaded HLA-A*0201 molecules should occur in a post-ER compartment. In line with this, the cross-presentation of long peptide gp100_184–227 by T2 cells was also inhibited dramatically by (ZLL)2-ketone (Fig. 2E), indicating that, like in DC, it relied on suboptimally loaded HLA-A*0201 molecules.

To determine whether suboptimally loaded HLA-A*0201 molecules were also needed for cross-presentation of long peptides in TAP-competent cells, we took advantage of the fact that ERAP, an aminopeptidase that trims antigenic peptides in the ER, is required to produce an optimal repertoire of stable peptide–MHC-I complexes (35). We observed that ERAP inhibition by l-leucinethiol dramatically increased cross-presentation of the gp100 epitope by Mo-iDCs, whereas endogenous presentation of the same gp100 epitope after electroporation of Mo-iDCs with gp100 mRNA was almost completely blocked (Fig. 2F). Again, no significant difference was observed concerning the expression of costimulatory signals by cells after l-leucincethiol or vehicle treatment (Supplemental Fig. 2). These results further supported the essential role of suboptimally loaded HLA-A*0201 molecules for cross-presentation of long peptides.

To confirm further the role of suboptimally loaded HLA-I molecules, we used lentiviral constructs enabling the delivery of peptides into the ER in a TAP-independent manner, through fusion to the signal sequence of IL-2 (36). From the list of HLA-A*0201-binding peptides eluted from TAP-deficient T2 cells (37), we selected two short peptides with low/intermediate affinity for HLA-A*0201 and one peptide with high affinity (Fig. 2G). As a negative control, we also used a peptide known to bind to HLA-A1 but not to HLA-A2 (MAGE-A3_168-176; EVDPIGHLY). We reasoned that after introduction of the constructs in T2 cells, the high amounts of ER-delivered peptides should modulate cross-presentation efficiency, with the low–intermediate HLA-A2 binders (suboptimal peptides) favoring cross-presentation, as opposed to the high HLA-A2 binder and the nonbinder peptides. This is exactly what we observed (Fig. 2G). As a control for the ER delivery of the peptides, we observed the HLA-A1–restricted peptide was efficiently presented to specific CTL in a TAP-independent manner (Supplemental Fig. 4). These results supported our model that cross-presentation occurs by peptide exchange on suboptimally loaded HLA-I molecules.

The results obtained so far suggest that cross-presentation of the long gp100 peptide occurs through peptide exchange on suboptimally loaded HLA-A*0201 molecules in a post-ER compartment containing the processed epitope derived from the engulfed long peptide. Because the vacuolar pathway of cross-presentation has been considered to make use of recycling MHC-I molecules that are derived from the cell surface (15–18), we asked whether those suboptimally loaded HLA-I were recycled or were newly synthesized. To avoid unspecific effects of inhibitors like primaquine on vesicle trafficking, we chose to block HLA-I recycling by incubating cells at low temperature, which is reported to block recycling of surface molecules (38). We observed that cross-presentation was not reduced at 20°C (Fig. 3A). We then confirmed that low temperature efficiently blocked HLA-I recycling in our Mo-iDCs cellular system using a two-step approach. We first showed that internalization of HLA-I, which conditions recycling, was almost completely blocked at 20°C (Fig. 3B, left), in line with the observation that endocytosis of MHC-I depends on membrane rafts, which are highly temperature sensitive (39), as opposed to the clathrin-dependent endocytosis of mannose receptor CD206 (38, 40, 41) (Fig. 3B, right). We then evaluated the recycling step at different temperatures (Fig. 3C). Surface HLA-I molecules of Mo-iDCs were first labeled with conformation-dependent anti–HLA-I murine mAb W6/32. Next, HLA-I/Ab complexes were allowed to internalize for 12 min at 37°C or 20°C in the presence of primaquine (220 μM) to prevent recycling. After stripping non-internalized HLA-I/Ab complexes and washing away primaquine, we followed the reappearance of HLA-I/Ab complexes at the cell surface at 20°C or 37°C by FACS using labeled anti-murine IgG Ab. When both internalization and recycling were performed at 37°C, we observed a fast kinetics of recycling, which was consistent with the observation that HLA-I are recycled through the RAB-35–dependent fast recycling pathway (42, 43). When recycling was performed at 20°C, we observed a striking reduction in the number of HLA-I recycled to the cell surface. When both internalization and recycling were performed at 20°C, HLA-I recycling to the cell surface was completely blocked. Because cross-presentation was not blocked at 20°C but recycling of HLA-I was, we concluded that cross-presentation of long peptides did not use recycled HLA-I for peptide exchange. To confirm further that cell surface HLA-I molecules are dispensable for long-peptide cross-presentation, we acid-stripped T2 cells and reconstituted cell surface HLA-A2 with short peptides of different HLA-A2 binding affinities before the cross-presentation assay. Although the extent of surface HLA-A2 reconstitution varied according to the peptide affinity for HLA-A2, the cross-presentation ability of the cells remained unaffected (Fig. 3D). If cross-presentation does not use recycled HLA-I molecules, it must use nascent HLA-I. This was confirmed by the inhibitory effect of protein synthesis inhibitor cycloheximide on cross-presentation of the long gp100 peptide (Fig. 3E).

Low temperature incubation not only blocks recycling of surface molecules; it also blocks the exit of secretory and surface proteins from the trans-Golgi-network in the classical secretion pathway (44, 45). We confirmed in our cellular system that incubation at 20°C prevented classical secretion of HLA-I molecules (Fig. 4A) and the presentation of endogenously derived class I–restricted Ags (Fig. 4B). Because cross-presentation was not affected by low temperature, we concluded that the nascent HLA-I molecules used for cross-presentation followed a nonclassical secretion pathway.

The ER-Golgi Intermediate Compartment is an important early secretion compartment involved in the sorting of ER-synthesized proteins. Cebrian et al. (12) reported a role for SEC22B in escorting ER-resident molecules from the ER-Golgi Intermediate Compartment to the phagosome, making it a good candidate to escort HLA-I for long-peptide cross-presentation. However, we observed that knocking down Sec22b with siRNA in Mo-iDCs did not prevent cross-presentation (Fig. 4C). Recently, CD74 was reported to escort MHC-I from the ER to the endolysosome compartment to promote cross-presentation (13). However, knocking down CD74 in Mo-iDC did not prevent cross-presentation of the long peptide either (Fig. 4D).

Taken together, the results reported above suggest that cross-presentation of the long gp100 peptide makes use of peptide-loaded 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 peptide before transfer of the final peptide/HLA complexes to the cell surface.

We used another long peptide, derived from melanocytic protein Melan-A to determine whether other long peptides followed the same cross-presenting pathway as gp100_184–227. Long peptide Melan-A_15-40A27L (₁₅KGHGHSYTTAE₂₆ELAGIGILTV₃₅ILGVL₄₀), which comprises the heteroclitic antigenic peptide Melan-A_26-35A27L with an alanine-to-leucine substitution in position 27 to increase binding to HLA-A*0201 (46, 47), was efficiently cross-presented to Melan-A_26–35-specific CTL by Mo-iDCs in a proteasome- and TAP-independent manner (Fig. 5A, B). Although ICP47 efficiently blocked TAP, as indicated by the reduced HLA-I surface expression (Supplemental Fig. 3B), it did not increase cross-presentation of the long Melan-A peptide as it did for the long gp100 peptide. Consistently, there was no effect of l-leucinethiol on the cross-presentation either (Fig. 5C). This might result from the higher affinity of Melan-A_26-35A27L for HLA-A*0201 as compared with gp100_209–217 (Supplemental Fig. 3C). The higher the affinity of the cross-presented epitope, the easier peptide exchange can take place, so that the limited pool of suboptimally loaded HLA-I molecules available in TAP- or ERAP-competent cells might become sufficient for optimal cross-presentation. This is supported by the fact that l-leucinethiol strongly increased cross-presentation of the wild-type Melan-A long peptide (Melan-A_15–40) (Fig. 5C), whose epitope (Melan-A_26–35) has a much lower affinity for HLA-A2 (Supplemental Fig. 3C) and is otherwise not cross-presented efficiently. As with the gp100 peptide, low temperature (20°C) did not block cross-presentation of the long Melan-A peptide (Fig. 5D), further supporting the involvement of an alternative secretion pathway of the HLA-I molecules used for cross-presentation. These results indicate that the cross-presentation of other long peptides follow the same pathway as the gp100 long peptide, involving vacuolar peptide exchange on nascent suboptimally loaded HLA-I molecules, which follow a nonclassical secretion pathway.

The first major conclusion from our work is that long-peptide cross-presentation follows a vacuolar pathway. This conclusion contrasts with a number of previous reports supporting a cytosolic pathway (8–11). Our proposed model implying a key role of suboptimally loaded HLA-I molecules provides a potential explanation for those discrepancies. Indeed, the support for a cytosolic pathway usually comes from the observation that cross-presentation is impaired by TAP inhibition. This is interpreted to indicate that the cross-presented epitope needs to be transported across membranes, in line with the cytosolic pathway (1–3). However, if, as we propose, cross-presentation critically depends on suboptimally loaded MHC molecules, TAP inhibition may block cross-presentation indirectly by reducing the availability of suboptimally loaded MHC-I. In fact, HLA-A2 is an exception among MHC-I in the fact that it can load signal peptides in the absence of TAP and therefore increase its pool of suboptimal peptides. Other MHC class I molecules need TAP to load suboptimal peptides. Therefore, in non–HLA-A2–restricted cross-presenting model systems, TAP inhibition may indirectly—and paradoxically—block long-peptide cross-presentation even though it follows a vacuolar pathway.

This alternative interpretation may apply to the study reported by Rosalia et al. (11) who studied cross-presentation of long peptides by mouse and human DC. They studied the role of the proteasome and TAP in murine bone marrow-derived DCs, and observed reduced cross-presentation using epoxomicin (1 μM) or using cells derived from TAP-KO mice. Although it is difficult to compare results obtained in different species using different peptides and DC types, the latter observation might be explained by an indirect effect of TAP inhibition in a vacuolar pathway, as discussed above.

Other reports studied cross-presentation of HLA-A2–restricted peptides similar or identical to the ones we analyzed (8–11). Faure et al. (10) used proteasome inhibitor lactacystin (40 μM) to evaluate cross-presentation of Melan-A and gp100 long peptides by human Mo-DCs. In line with our results, they observed no inhibition of the cross-presentation of the gp100 peptide. For Melan-A, they observed partial inhibition, with high donor-to-donor variation. They did not evaluate the effect of TAP inhibition. In Segura et al. (9),the same group used lactacystin (2.5 μM) to evaluate cross-presentation of a long Melan-A peptide by various DC, using an IFN-γ production assay. They observed some inhibition, again with donor-to-donor variation, but the IFN-γ signal was weak (∼100 pg/ml), probably at the limit of sensitivity of the IFN-γ ELISA. Again, they did not evaluate the effect of TAP inhibition. If the effects reported by Segura et al. (9) are confirmed, the difference with our data might result from the fact that different DC types were used.

Contrary to these studies, Ménager et al. (8) did evaluate the effect of TAP inhibition on the cross-presentation of a long HLA-A2–restricted Melan-A peptide. They did so using a synthetic peptide corresponding to the N-terminal 35-amino acid residues of TAP inhibitor ICP47. They reported inhibition of cross-presentation using this long ICP47 peptide at 50 or 100 μM. However, this experiment was lacking a control with an irrelevant peptide. We have observed competition effects between long peptides for cross-presentation. Because the TAP inhibitor used by Ménager et al. (8) was, in essence, a long peptide, it is expected to compete with the long peptide for cross-presentation. Therefore, it is likely that the inhibition observed was due to competition rather than TAP inhibition. Ménager et al. (8) also studied the effect of proteasome inhibitor epoxomicin at high doses (1–5 μM), and observed significant inhibition only at 5 μM. We used epoxomicin at 0.5 μM, and provided controls demonstrating complete inhibition of endogenous presentation of gp100 at this dose (Fig. 1). It is known that epoxomicin at high doses induces a number of other effects in the cell, including ER stress. Therefore, the effect observed by Ménager et al. (8) might be unspecific because of the high dose used. In line with this conclusion is the fact that the authors did not observe any modulation of cross-presentation when they changed the proteasomal subunit composition of DCs, whereas the Melan-A antigenic peptide is known to be processed differently by the standard proteasome and the immunoproteasome (21, 48–50).

To our knowledge, our study on long-peptide cross-presentation is the first to evaluate both the effect of TAP inhibition and the effect of proteasome inhibition in carefully controlled conditions. Our results clearly indicate a vacuolar pathway of cross-presentation of these long peptides by human monocyte-derived DCs.

The second major conclusion of our work is that long-peptide cross-presentation depends on newly synthesized HLA-I molecules that are loaded with suboptimal peptides. In our view, suboptimal peptides mean peptides that are able to bind HLA-I, but with a low affinity so that they can be readily exchanged. The binding of a low-affinity peptide may be needed to stabilize the MHC-I and allow it to exit to ER and reach the post-ER compartment, where the low-affinity peptide would be exchanged for the cross-presented peptide. The exact nature of the cross-presenting compartment is unclear at this stage and is under study in our laboratory. Candidates include the late endosomal MHC class II compartment, the phagosome and the IRAP-containing vesicles. Another highly relevant question we are currently addressing is to delineate the exact secretion pathway followed by the nascent MHC-I molecules used for cross-presentation. Answering such questions will bring new insights into intracellular trafficking pathways and may influence the development of vaccines based on long peptides to trigger CD8 T cell responses in patients with cancer.

We thank the Centre d’hématologie of the Cliniques universitaires Saint-Luc (Brussels) for providing blood samples from hemochromatosis patients; Aline Depasse for technical help; Aude Bonehill and Carlo Heirman (Vrije Universiteit Brussel, Brussels) for help with mRNA preparation for electroporation; Zhaojun Sun, Didier Colau, and Florence Depontieu for help at various stages of the project; Julie Klein and Mandy Macharis for editorial assistance; and Pierre Coulie for critical reading of the manuscript.

The authors have no financial conflicts of interest.