Inefficient exogenous loading of a tapasin-dependent peptide onto HLA-B*44:02 can be improved by acid treatment or fixation of target cells

Authors

  • Vincent Stroobant,

    1. Ludwig Institute for Cancer Research, Brussels, Belgium
    2. WELBIO and de Duve Institute, Université catholique de Louvain, Brussels, Belgium
    Search for more papers by this author
    • These authors contributed equally to the work.

  • Nathalie Demotte,

    1. Ludwig Institute for Cancer Research, Brussels, Belgium
    2. WELBIO and de Duve Institute, Université catholique de Louvain, Brussels, Belgium
    Search for more papers by this author
    • These authors contributed equally to the work.

  • Rosalie M. Luiten,

    1. Department of Dermatology, Academic Medical Center, University of Amsterdam, The Netherlands
    Search for more papers by this author
  • Ralf M. Leonhardt,

    1. Department of Immunobiology, Yale University School of Medicine, New Haven, CT, USA
    Search for more papers by this author
  • Peter Cresswell,

    1. Department of Immunobiology, Yale University School of Medicine, New Haven, CT, USA
    2. Howard Hughes Medical Institute, Chevy Chase, MD, USA
    Search for more papers by this author
  • Aude Bonehill,

    1. Laboratory of Molecular and Cellular Therapy, Department of Physiology and Immunology, Medical School of Vrije Universiteit Brussel (VUB), Brussels, Belgium
    Search for more papers by this author
  • Alexandre Michaux,

    1. Ludwig Institute for Cancer Research, Brussels, Belgium
    2. WELBIO and de Duve Institute, Université catholique de Louvain, Brussels, Belgium
    Search for more papers by this author
  • Wenbin Ma,

    1. Ludwig Institute for Cancer Research, Brussels, Belgium
    2. WELBIO and de Duve Institute, Université catholique de Louvain, Brussels, Belgium
    Search for more papers by this author
  • Arend Mulder,

    1. Department of Immunohematology, Leiden University Medical Center, Leiden, The Netherlands
    Search for more papers by this author
  • Benoît J. Van den Eynde,

    1. Ludwig Institute for Cancer Research, Brussels, Belgium
    2. WELBIO and de Duve Institute, Université catholique de Louvain, Brussels, Belgium
    Search for more papers by this author
  • Pierre van der Bruggen,

    1. Ludwig Institute for Cancer Research, Brussels, Belgium
    2. WELBIO and de Duve Institute, Université catholique de Louvain, Brussels, Belgium
    Search for more papers by this author
  • Nathalie Vigneron

    Corresponding author
    1. WELBIO and de Duve Institute, Université catholique de Louvain, Brussels, Belgium
    • Ludwig Institute for Cancer Research, Brussels, Belgium
    Search for more papers by this author

Correspondence: Dr. Nathalie Vigneron, Ludwig Institute for Cancer Research, de Duve Institute, Avenue Hippocrate 74, UCL 7459; B-1200 Brussels, Belgium

Fax: +32-27647590

e-mail: nathalie.vigneron@bru.licr.org

Abstract

Antitumor cytolytic T lymphocytes (CTLs) recognize peptides derived from cellular proteins and presented on MHC class I. One category of peptides recognized by these CTLs is derived from proteins encoded by “cancer-germline” genes, which are specifically expressed in tumors, and therefore represent optimal targets for cancer immunotherapy. Here, we identify an antigenic peptide, which is derived from the MAGE-A1-encoded protein (160-169) and presented to CTLs by HLA-B*44:02. Although this peptide is encoded by MAGE-A1, processed endogenously and presented by tumor cells, the corresponding synthetic peptide is hardly able to sensitize target cells to CTL recognition when pulsed exogenously. Endogenous processing and presentation of this peptide is strictly dependent on the presence of tapasin, which is believed to help peptide loading by stabilizing a peptide-receptive form of HLA-B*44:02. Exogenous loading of the peptide can be dramatically improved by paraformaldehyde fixation of surface molecules or by peptide loading at acidic pH. Either strategy allows efficient exogenous loading of the peptide, presumably by generating or stabilizing a peptide-receptive, empty conformation of the HLA. Altogether, our results indicate a potential drawback of short peptide-based vaccination strategies and offer possible solutions regarding the use of problematic epitopes such as the one described here.

Introduction

A number of cancer vaccination trials currently make use of antigenic peptides, which are recognized by cytolytic T lymphocytes (CTLs) and are displayed on tumor cells by MHC class I molecules (reviewed in [[1]]). Because of their tumor specificity, peptides derived from proteins encoded by cancer-germline genes such as the MAGE, BAGE, GAGE, LAGE/NY-ESO, or SSX genes, are of particular interest for cancer immunotherapy [[1]]. Indeed, these genes are expressed in many tumors of various histological types and in male germline cells, but are silent in other normal tissues. Since male germline cells are devoid of MHC class I molecules, presentation of these peptides is generally considered as strictly tumor specific [[1]].

CTLs that recognize tumor antigens have often been isolated from the blood or tumors of cancer patients [[2-4]]. The antigenic peptides recognized by these CTLs are typically produced in the cytosol through degradation of cellular proteins by a multicatalytic protease called the proteasome [[5]]. Peptides resulting from proteasomal degradation are then transported into the lumen of the endoplasmic reticulum (ER) by the transporter associated with antigen processing (TAP). In the ER, peptide loading onto MHC class I-β2m dimers is facilitated by the peptide-loading complex, which is composed of TAP, tapasin (TPN), ERp57, calreticulin (CRT), MHC class I heavy chain (HC), and β2m (reviewed in [[6]]). Association of MHC class I molecules with a peptide is essential for their stability and transport to the cell surface, and proteins such as tapasin play a crucial role for optimal assembly of peptide-MHC class I complexes [[6]].

We report here the identification of an antigenic peptide derived from tumor protein MAGE-A1 (160-169) and presented to CTLs by HLA-B*44:02. Although the peptide is well presented when processed intracellularly, it almost completely fails to associate with HLA-B*44:02 when pulsed onto target cells from outside. Using a variety of cellular and biochemical approaches, we investigated the conditions required for the exogenous loading of this tapasin-dependent peptide.

Results

Isolation of a CTL recognizing a MAGE-A1 peptide

In order to identify new antigenic peptides derived from the MAGE-A1 protein, CD8+ T cells from a hemochromatosis patient were subjected to four weekly stimulations with autologous presenting cells transduced with a recombinant canarypoxvirus, ALVAC, containing the coding sequence of MAGE-A1. CTL clone LB1801-461/G4.2 (hereinafter called CTL4) was isolated after cloning of the responsive population by limiting dilution. CTL4 specifically lysed autologous EBV-B cells infected with a vaccinia virus construct driving expression of the MAGE-A1 protein (Fig. 1A, left panel). The MAGE-A1 positive melanoma cell line LB373-MEL4.1 was also recognized by CTL4, demonstrating that the MAGE-A1 peptide was presented on melanoma cells after processing of the endogenous MAGE-A1 protein (Fig. 1A, right panel).

Figure 1.

CTL4 recognizes a peptide derived from MAGE-A1 and restricted by HLA-B*44:02. (A) Lysis of melanoma (LB373-MEL) and EBV-B cells expressing MAGE-A1 by CTL4 (at the indicated effector-target ratios) was measured by chromium release assay. Two hours after transduction of autologous EBV-B cells with the vaccinia constructs, cells were labeled with 51Cr and incubated with CTL4. Chromium release was then measured after 4 h. Data are representative of eight experiments. (B) The antigenic peptide recognized by CTL4 is derived from MAGE-A1 and restricted to HLA-B*44:02. COS-7 cells were transfected with constructs encoding MAGE-A1 and the indicated HLA alleles. Transfected cells were then tested for their ability to stimulate TNF production by CTL4. Data are shown as mean of three replicates and are representative of three experiments. (C) Representation of the MAGE-A1 coding sequence, and the minigenes used to identify the region coding for the antigenic peptide. These constructs were transfected in COS-7 together with the cDNA encoding HLA-B*44:02, and the cells were then tested for their ability to stimulate TNF production by CTL4. Data are shown as mean of two replicates and are representative of three experiments. (D) CTL4 does not recognize MAGE-A1 peptides after loading on LB1801-EBV. 51Cr-labeled target cells were loaded using 1 μM of the indicated peptide. After loading, CTLs were added at an effector-to-target ratio of 10 to 1. As control, the CTLMZ2-82/30 was used to verify the integrity of the peptide solutions loaded on HLA-A*01:01+LB1972-EBV. NT: not tested. Data are shown as mean of three replicates.

Identification of the MAGE-A1 antigenic peptide recognized by CTL4

Melanoma cell line LB373-MEL4.1, which is recognized by CTL4, and patient LB1801, from which CTL4 was derived, share HLA-A*02:01, HLA-B*44:02, and HLA-Cw5, suggesting that the peptide recognized by CTL4 is presented by one of these HLA molecules. We therefore transfected COS-7 cells with cDNA constructs encoding MAGE-A1 and either HLA-A*02:01, HLA-B*44:02, or HLA-Cw5 and assessed their recognition in T-cell assays. CTL4 only recognized cells transfected with MAGE-A1 and HLA-B*44:02 (Fig. 1B), demonstrating that its cognate antigenic peptide is presented by HLA-B*44:02.

To further define the peptide-coding sequence, we transfected COS-7 cells with plasmids encoding HLA-B*44:02 and a series of truncated MAGE-A1 plasmid constructs. These cells were then tested for recognition by CTL4 (Fig. 1C). A minigene construct encoding a 9-amino acid fragment (EADPTGHSY) located between positions 161 and 169 of the MAGE-A1 protein and preceded by a start-methionine sensitized transfected cells to CTL recognition. Because the peptide binding motif to HLA-B44 is characterized by a glutamic acid in position 2 and an aromatic amino acid in position 9 or 10 [[7]], it was likely that the endogenous peptide recognized by CTL4 contained the lysine residue present at position 160 of the MAGE-A1 protein instead of the start-methionine of the minigene construct (Fig. 1C). To address this possibility, various synthetic peptides overlapping the MAGE-A1160-169 sequence were loaded exogenously on autologous HLA-B*44:02+ target cells (LB1801-EBV) and tested for their ability to activate CTL4 (Fig. 1D). However, surprisingly none of the peptides tested was recognized by CTL4 even when applied at 1 μM (Fig. 1D, left panel). In contrast, HLA-A*01:01+ target cells loaded with the exact same set of peptides were efficiently recognized by CTL MZ2-82/30, which is specific for peptide EADPTGHSY presented by HLA-A*01:01 [[8]] (Fig. 1D, right panel). Peptide candidate KEADPTGHSY was only weakly recognized by CTL4, at doses of 10 μM and higher (Fig. 2A, higher panel). Strikingly, we observed that CTL recognition of peptide KEADPTGHSY was drastically improved when the HLA-B*44:02+ target cells were fixed with paraformaldehyde (PFA) prior to exogenous loading (Fig. 2A). This was the case when using the autologous EBV-B cells LB1810-EBV or the HLA-B*44:02+ 721.220 cells back-transfected with tapasin (so that they express sufficient amounts of surface HLA-B*44:02). Because fixed antigen-presenting cells (APCs) are devoid of any metabolic and processing capability, this suggested that PFA fixation improved exogenous peptide loading, as previously proposed by Rock et al. [[9]], by cross-linking surface HLA heavy chain and β2m in a peptide receptive state, preventing their denaturation after endogenous peptide dissociation.

To confirm that peptide KEADPTGHSY was presented at the surface of cells expressing MAGE-A1, we eluted the peptides from HLA-B*44:02+ 721.220 cells back-transfected with tapasin and transduced with a retroviral vector encoding the MAGE-A1 protein. Because these cells do not express any endogenous HLA-A or -B alleles, this limited the possibility that the MAGE-A1 peptide KEADPTGHSY could bind to other HLA alleles than HLA-B*44:02. Peptide eluates were separated by high-performance liquid chromatography (HPLC), and the fractions obtained were loaded on PFA-treated cells (Fig. 2B). The fractions able to stimulate CTL4 corresponded to those obtained when synthetic peptide KEADPTGHSY was run under the same HPLC conditions. This indicates that peptide KEADPTGHSY is presented by HLA-B*44:02 on the surface of cells expressing MAGE-A1 and rules out the possibility that the peptide recognized by CTL4 contains a post-translational modification.

Figure 2.

The peptide recognized by CTL4 requires endogenous loading. (A) Activation of CTL4 is improved when peptide KEADPTGHSY is pulsed on PFA-fixed target cells. B*44:02+ target cells — HLA-B*44:02+ 721.220 cells expressing tapasin (top) or autologous LB1801-EBV (bottom) — were fixed (black squares) or not (empty squares) prior to pulsing with the indicated concentration of peptide at 37°C for 1 h. Cells were then washed and tested for their ability to activate CTL4 in an IFN-γ release assay (ELISA). Data are shown as the mean of two replicates and are representative of three experiments. (B) Peptide KEADPTGHSY is present on the surface of cells expressing HLA-B*44:02 and MAGE-A1. Total peptides were eluted by mild acid treatment from B*44:02+ target cells (HLA-B*44:02+ 721.220 cells expressing tapasin) transduced with a MAGE-A1 retroviral construct. The acid eluate was fractionated by HPLC, and the fractions were tested for recognition by CTL4 after pulsing on PFA-fixed target cells, by IFN-γ production. To rule out contamination of the HPLC system, buffer was run on the column before the eluted sample, and the fractions were tested similarly (top). Synthetic peptide was injected under the same HPLC conditions, and the fractions were also tested for CTL recognition (bottom). Data are shown as mean of two replicates. (C) CTL activation is improved when peptide is loaded inside the target cells. HLA-B*44:02+ targets cells (HLA-B*44:02+ 721.220 cells expressing tapasin) were electroporated with the indicated peptides and tested for recognition by CTL4 (black bars). As a control, peptide was pulsed in the same conditions, except that electroporation was omitted (white bars). NT: not tested. Data are shown as mean ± SD of four replicates and are representative of two experiments. (D) Presentation of the peptide KEADPTGHSY requires expression of tapasin. HLA-B*44:02+ 721.220 cells expressing (or not) wild-type tapasin (WT-TPN) were transfected with constructs encoding MAGE-A1 and cultured with the CTL4. After 20 h, IFN-γ was measured in the supernatants. Data are shown as mean ± SD of four replicates and are representative of four experiments.

Recognition of the peptide by CTL4 requires endogenous loading

So far, our results indicate that the peptide KEADPTGHSY is naturally expressed on cells expressing full-length MAGE-A1 and HLA-B*44:02. On the other hand, exogenous loading of the peptide on HLA-B*44:02-expressing cells does not trigger activation of CTL4. In line with this, we observed that HLA-B*44:02+ target cells were much better recognized by CTL4 when peptide KEADPTGHSY was introduced inside the cells by electroporation rather than being applied from outside (Fig. 2C). Shorter peptides, lacking the N- or the C-terminal residues were not recognized after electroporation further supporting that KEADPTGHSY is the minimal epitope recognized by CTL4. The fact that the peptide can be recognized by CTLs after electroporation but not after exogenous loading clearly shows that presentation of the peptide to CTL4 required loading inside the target cell. In that regard, HLA-B*44:02 differs from many other HLA molecules by its strict dependence on chaperone tapasin for peptide loading [[10]]. Tapasin helps maintaining the HLA molecule in a peptide receptive state, facilitating the release of low-affinity peptide and the subsequent stabilization of MHC class I molecules with peptides of greater affinities [[11]]. To assess whether the loading of peptide MAGE-A1160-169 required assistance by tapasin, we transiently transfected MAGE-A1 into 721.220.B*44:02 cells expressing or not tapasin, and observed that cells expressing tapasin were recognized by CTL4, while cells lacking tapasin were not recognized at all (Fig. 2D). Overall, these results indicate that efficient presentation of peptide KEADPTGHSY to CTL4 requires endogenous loading and, like most other peptides loaded on HLA-B*44:02, the presence of tapasin.

The MAGE-A1 peptide poorly binds surface HLA-B*44:02

The fact that tapasin, a factor essentially required to load endogenous MAGE-A1160-169 onto HLA-B*44:02 in the loading complex (Fig. 2D), is subcellularly restricted to the ER, and thus does not localize to the plasma membrane, may explain why exogenously applied peptide KEADPTGHSY does not activate CTL4. In this scenario, peptide pulsed from the outside would simply fail to access and bind surface-HLA-B*44:02, because tapasin-mediated assistance is lacking. Alternatively, tapasin might drive the formation of a CTL-reactive conformation of the HLA/peptide complex. In this alternative scenario, exogenously applied peptide would principally bind surface HLA-B*44:02, but the thereby formed HLA/peptide complex would adopt a distinct conformation that is not recognized by CTL4. In order to discriminate between lack of peptide binding to surface HLA or the binding of the peptide in an alternative conformation, we quantified the amount of peptide found at the surface of cells exogenously pulsed with peptide KEADPTGHSY or expressing MAGE-A1 endogenously (Fig. 3). Peptides acid eluted from cells were concentrated and loaded onto prefixed target cells. As a control, peptides were also eluted from cells that had been fixed with PFA prior to pulsing with peptide KEADPTGHSY (Fig. 3). As expected, unfixed peptide-pulsed cells were not recognized by CTL4 (Fig. 3, left panel). Peptides eluted from the surface of these cells did not appear to contain the epitope, because when they were loaded onto prefixed target cells, they did not activate CTL4 (Fig. 3, right panel). In contrast, peptides eluted from cells expressing MAGE-A1 intracellularly or from prefixed, peptide-pulsed cells clearly activated CTL4 when exogenously applied to PFA-treated targets (Fig. 3, right panel). These results demonstrate that the lack of CTL recognition is not related to the binding of the peptide in an alternative conformation but rather to the fact that exogenously applied peptide KEADPTGHSY does not bind surface HLA-B*44:02 on untreated cells.

Figure 3.

Peptide KEADPTGHSY does not bind surface HLA-B*44:02 when pulsed exogenously. HLA-B*44:02+ target cells 721.220.B*44:02.tapasin expressing (or not) MAGE-A1 were pulsed (or not) with the relevant peptide. Recognition by CTL4 prior to peptide elution was determined by IFN-γ production (left). Total peptides were acid eluted from surface HLA molecules and purified on a C18 column. Concentrated (1:1, black bars) or twofold diluted eluates (1:2, white bars) were then loaded onto PFA-fixed, HLA-B*44:02+ target cells for recognition by CTL4 as measured by IFN-γ production (right). As a control, peptides were also eluted from HLA-B*44:02+ target cells, which were fixed with paraformaldehyde prior to pulsing of the peptide. Data are shown as mean ± SD of four (left) and two (right) replicates and are representative of three experiments.

In order to further elucidate the role of tapasin in the loading of the MAGE-A1 peptide, we made use of cells expressing HLA-B*44:05, an allele that differs from HLA-B*44:02 by a single polymorphic residue in position 116, conferring HLA-B*44:05 with the ability to load peptide independently from tapasin. We loaded the peptide on HLA-B*44:05+ cells, and tested the cells for recognition by CTL4 (Supporting Information Fig. 1A). CTL4 was unable to recognize HLA-B*44:05+ target cells loaded with peptide. Because this lack of recognition could originate from the fact that CTLs and target cells express different polymorphic variants of HLA-B44, we verified whether peptide KEADPTGHSY could bind HLA-B*44:05 as described above. Peptides eluted from exogenously loaded HLA-B*44:05 expressing cells were loaded onto PFA-fixed HLA-B*44:02 cells and tested for their ability to activate CTL4. Surprisingly, the presence of peptide KEADPTGHSY could not be detected in the eluates obtained from peptide-pulsed HLA-B*44:05 even if these cells were fixed with PFA prior peptide loading (Supporting Information Fig. 1B). This suggests that peptide KEADPTGHSY cannot bind HLA-B*44:05. However, this could originate from the fact that the peptide repertoires displayed by HLA-B*44:05 and HLA-B*44:02 do not completely overlap and that both HLA show different peptide binding specificities, with HLA-B*44:05 having a stronger preference for F in P9 while HLA-B*44:02 equally accepts F or Y [[10]].

Lack of peptide binding to surface HLA-B*44:02 is not due to a low affinity of the peptide

To determine whether inefficient loading of the peptide on surface HLA-B*44:02 was due to a particularly low affinity of the peptide for HLA-B*44:02, we measured the ability of peptide KEADPTGHSY to bind and stabilize HLA-B*44:02 using an in vitro stabilization experiment (Fig. 4A) [[12]]. Peptide KEADPTGHSY was able to stabilize HLA-B*44:02 almost as efficiently as the SEIPRVYKF control peptide, which had been previously used as an efficient HLA-B*44:02 binder in peptide transport experiments [[13]] (Fig. 4A). Surprisingly, peptide KEADPTGHSY appeared to have a higher affinity for HLA-B*44:02 than peptide EEKLIVVLF, a MUM-1-derived peptide that can be efficiently recognized by CTLs after exogenous loading on surface HLA-B*44:02 [[14]]. The ability of the peptide to bind HLA-B*44:02 in these in vitro conditions despite the absence of tapasin might be related to the presence of detergent in the assay. Indeed, a low detergent concentration was previously suggested to generate empty HLA molecules or stabilize an HLA structure that is more prone to bind peptide [[15]]. In any case, our results demonstrate that the inability of the MAGE-A1 peptide to associate with surface HLA-B*44:02 is not caused by a particularly low affinity of this peptide for the HLA molecule.

Figure 4.

Conditions favoring peptide exchange on surface HLA-B*44:02. (A) In vitro peptide binding experiment. Tapasin-negative, HLA-B*44:02+ 721.220 cells were radio-labeled with 35S-methionine/cysteine for 30 min and lysed in 1% Triton-X100 in the presence of various concentration of the indicated peptides. As a negative control peptide RRYQNSTEL was used. Stabilized HLA molecules were immunoprecipitated using the W6/32 antibody. The samples were then separated by SDS-PAGE (right), and the intensity of the bands was analyzed using phosphorimaging (left). The amount of W6/32-reactive complexes recovered was plotted as a percentage of that obtained with the highest concentration of peptide KEADPTGHSY. Data are shown as mean ± SD of two experiments. (B) Low pH improves exogenous loading of the peptide on HLA-B*44:02. HLA-B*44:02+ 721.220.B*44:02.tapasin cells were incubated for 20 min at 37°C in the presence of the indicated concentration of peptide KEADPTGHSY at pH 5.5 or 7. After neutralization and washing, pulsed cells were tested for their ability to stimulate IFN-γ production by CTL4. Additionally, cells were incubated at pH 5.5 for 20 min, neutralized then pulsed with peptide, washed and incubated with CTL4. Data are shown as mean ± SD of two replicates and are representative of four experiments. (C) Cells expressing a lower affinity repertoire are more efficiently recognized by CTL4 after pulsing. HLA-B*44:02+ 721.220 cells expressing WT or C95A tapasin mutant were pulsed with peptide KEADPTGHSY at pH 7 for 1 h at 37°C, washed, and cultured with CTL4. After 20 h, IFN-γ produced by CTL4 was measured. Data are shown as mean ± SD of two replicates and are representative of four experiments.

Acidic pH favors the loading of the MAGE-A1 peptide onto HLA-B*44:02

It is known that acidic pH can promote peptide exchange, at least for some MHC/peptide combinations [[16]]. For instance, the peptide repertoire of suboptimally loaded MHC class I can be affinity improved via peptide exchange in post-ER compartments such as the trans-Golgi network (TGN), where the pH is around 5.5–6.0 [[12, 17]]. At this pH, low-affinity ligands tend to dissociate from the MHC [[16, 18]], but MHC class I molecules seem to remain at least temporarily in a peptide-receptive state allowing peptide exchange to occur [[19]]. We therefore speculated that low pH might also improve exogenous loading of the MAGE-A1 peptide. To test this hypothesis, we pulsed HLA-B*44:02+ cells with exogenous peptide KEADPTGHSY at pH conditions resembling the acidic milieu of the TGN and endosomes or at a neutral pH for control [[20]]. After neutralization, cells were washed and tested for CTL recognition (Fig. 4B). Strikingly, decreasing pH to 5.5 drastically favored peptide loading onto surface HLA-B*44:02, suggesting that a low pH environment is sufficient to drive some peptide exchange at the cell surface.

Because acidic pH is known to elute peptides from MHC molecules, we asked whether it favored loading of the MAGE-A1 peptide simply by increasing the number of empty HLA-B*44:02 molecules or whether it modified peptide receptiveness of HLA-B*44:02 in a manner similar to tapasin. Acid-treated HLA-B*44:02+ cells were neutralized before pulsing the MAGE-A1 peptide. In these conditions, pulsing should occur at neutral pH on target cells with presumably empty surface HLA-B*44:02 molecules (Fig. 4B). Peptide loading was not improved in these conditions, suggesting that acidic pH improved loading through a direct effect on HLA-B*44:02 peptide receptiveness.

Decreasing the affinity of the peptide repertoire also facilitates loading of peptide KEADPTGHSY

Previous reports showed that one of the rate-limiting step for peptide exchange is the dissociation of the pre-existing peptide-MHC complexes, which basically depends on the affinity of the preloaded peptide for the HLA molecule [[21]]. We therefore hypothesized that cells displaying a lower affinity peptide repertoire might be loaded more efficiently with the exogenous synthetic peptide. Previously, we showed that 721.220.B*44:02 cells expressing defective tapasin mutant C95A express a lower affinity peptide repertoire when compared with those expressing wild-type (WT) tapasin [[22]]. We compared CTL recognition of these two cell types loaded exogenously with peptide KEADPTGHSY. Even though HLA-B*44:02 surface expression was expectedly lower in cells expressing C95A when compared with WT tapasin-expressing cells [[22]], substantially less peptide was needed to activate CTL4 (Fig. 4C). This result suggests that exogenous loading of peptide KEADPTGHSY occurs much more readily when HLA-B*44:02 is not optimally loaded, probably because peptide exchange happens more efficiently under these conditions.

Tapasin-independent endogenous loading of an overlapping peptide is preferred to tapasin-dependent loading

As indicated above, previously described MAGE-A1 peptide EADPTGHSY, which is entirely encompassed within the HLA-B44-restricted peptide described here, is presented by HLA-A*01:01 to the CTL clone MZ2-82/30 [[23]]. Therefore, and because HLA-A*01:01 was described as a partially tapasin-dependent molecule [[24]], we tested tapasin dependence of the presentation of the peptide EADPTGHSY by HLA-A*01:01, since this peptide can be loaded exogenously (Fig. 1D). We observed that HLA-A*01:01+, tapasin-deficient 721.220 cells transfected with MAGE-A1 efficiently stimulated the CTL clone MZ2-82/30, indicating that HLA-A*01:01-retricted peptide EADPTGHSY is not tapasin dependent, contrary to the HLA-B44-restricted peptide KEADPTGHSY (Fig. 5A). Because these two almost identical peptides are presented either by HLA-B*44:02 in a tapasin-dependent manner or by HLA-A*01:01 in a tapasin-independent manner, we wondered whether these two HLA molecules competed for presenting these peptides and whether endogenous presentation of the tapasin-independent peptide was preferred. We transfected 293 cells with increasing concentrations of the MAGE-A1 cDNA in the presence of HLA-A*01:01 and/or HLA-B*44:02, and tested them for recognition by the MAGE-A1/B44 CTL4 and the MAGE-A1/A1 CTL MZ2-82/30 (Fig. 5B). Interestingly, presentation of peptide KEADPTGHSY by HLA-B*44:02 was abolished in the presence of HLA-A*01:01, while presentation of the peptide by HLA-A*01:01 was not affected by the presence of HLA-B*44:02 (Fig. 5B). We conclude that there is competition between HLA-A*01:01 and HLA-B*44:02 for presentation of the peptide, and that presentation by HLA-A*01:01 is favored. This might result from the tapasin-independent loading of the peptide on HLA-A*01:01, potentially combined with a higher affinity of the peptide for HLA-A*01:01.

Figure 5.

Competition for the endogenous loading of the MAGE-A1 peptide on HLA-A*01:01 (tapasin independent) and HLA-B*44:02 (tapasin dependent). (A) Presentation of the peptide EADPTGHSY does not require tapasin. HLA-A*01:01+ 721.220 cells expressing (or not) wild-type tapasin (WT-TPN) were transfected with constructs encoding MAGE-A1 and cultured with the HLA-A*01:01-restricted CTLMZ2-82/30. After 20 h, IFN-γ was measured in the supernatants. Data are shown as mean ± SD of three replicates and are representative of three experiments. (B) Endogenous presentation of peptide (K)EADPTGHSY by HLA-A*01:01 is favored over HLA-B*44:02. HEK-293-EBNA were transiently transfected with HLA-B*44:02 and/or HLA-A*01:01 cDNA in the presence of the indicated amount of MAGE-A1 cDNA. The next day the HLA-B*44:02-restricted CTL4 or the HLA-A*01:01-rectricted CTLMZ2-82/30 were added. After 20 h, IFN-γ was measured in the supernatants. Data are shown as mean of two replicates and are representative of two experiments.

Discussion

Here, we describe an antigenic peptide presented by HLA-B*44:02 and derived from tumor-specific protein MAGE-A1. The peptide identified (KEADPTGHSY) is naturally presented by melanoma cells and requires tapasin for loading on HLA-B*44:02 in the peptide loading complex. Strikingly, however, the corresponding synthetic peptide hardly sensitizes target cells to CTL recognition when pulsed exogenously, because it fails to bind to surface HLA-B*44:02 molecules, which are presumably not in a conformation receptive to the peptide. Binding of the exogenous peptide can be recovered by fixation of target cells or loading at acidic pH, both of which likely change the conformation of surface HLA molecules to a peptide-receptive state. Our peptide elution studies (Fig. 3) formally exclude an alternative scenario that could have explained the inefficient exogenous loading of this peptide and would have involved the binding of the exogenous peptide to HLA-B*44:02 in a conformation that would not be recognized by CTL4. This was observed with a MHC-class II-restricted HEL peptide, which in the absence of HLA-DR, is loaded in an alternative conformation that is not recognized by T cells directed against the normally presented peptide [[25]].

High concentrations of exogenously loaded peptides are often required for recognition of HLA-B44/peptides complexes by antitumor CTLs [[26]]. Among 70 tumor antigenic peptides found in the Cancer Immunity database [[27]], 15 peptides required concentrations higher than 100 nM to induce half-maximal lysis (Supporting Information Table 1), and seven of these are presented by HLA-B44. Although the need for a high concentration of peptide might also be due to the presence of a low-affinity TCR on CTLs, it likely originates from a low efficiency of exogenous loading on surface HLA-B44. Inefficient exogenous loading of surface HLA-B44 was previously suggested by Akatsuka et al. [[26]], who identified an HLA-B*4403-restricted peptide derived from BCL2A1 that sensitized the target cells more efficiently after endogenous loading. Likewise, the MAGE-A1 peptide we described is hardly recognized when loaded exogenously, and only efficiently activates CTL4 when it is introduced inside the target cells by electroporation, or when expressed from a minigene inside the cytosol. Peptide loading inside the target cells is facilitated by tapasin, an ER-localized chaperone that regulates the loading of peptide on MHC class I molecules. Molecular dynamics simulation studies suggested that tapasin stabilizes the HLA molecule in an intermediate conformation state that induces the rapid dissociation of low-affinity peptides bound to the class I molecule, thereby facilitating the stabilization of MHC class I by peptides of greater affinities [[28]]. At the cell surface, the very low amount of such peptide-receptive HLA-B44 molecules might explain the lack of binding of the MAGE-A1 peptide onto surface HLA-B44, as well as the general requirement for high concentration of peptides to obtain half-maximal lysis by B44-restricted CTLs. The inability to produce stable empty HLA-B44 molecules might also explain why HLA-B44 recombinant monomers appear to be so difficult to refold in vitro (D. Colau, LICR Brussels, personal communication).

We observed that the inability of surface HLA-B44 to load peptide KEADPTGHSY could be overcome when the peptide was pulsed at lower pH. In mouse cells, H-2Kk molecules also bound peptides and stimulated CTLs more efficiently when pulsed at pH 5.5 [[16]], while other alleles such as H-2Kb and H-2Dd optimally presented peptides when pulsed at pH 7. Improved pulsing at low pH was not restricted to tapasin-dependent alleles, since we also observed this phenomenon with alleles such as HLA-A*01:01 and HLA-A*02:01 (Supporting Information Fig. 2). However, in those cases, exogenous loading was already efficient at neutral pH. We showed that increased binding at low pH appears to be due to an increased receptiveness of the HLA molecule. The difference in the ability of various HLA alleles to bind peptide at acidic pH might therefore depend on the degree of receptiveness of the corresponding HLA molecule at neutral pH versus acidic pH. Additionally, decreasing the affinity of the peptide repertoire displayed on surface HLA-B*44:02 also appeared to improve the loading of exogenous peptide KEADPTGHSY. It is in agreement with the fact that the rate of exogenous peptide binding is limited by the dissociation rate of previously bound peptides [[21]] and that a suboptimal peptide repertoire is essential for peptide exchange on HLA-B*5101 in the TGN and post-TGN vesicles [[17, 29]].

Taken together, we present and characterize here an HLA-B*44:02-restricted epitope derived from the tumor antigen MAGE-A1 and unravel interesting aspects of this peptide with regard to MHC class I binding. To the best of our knowledge no other MHC class I ligand has been described thus far, whose exogenous loading is that inefficient, but can be so dramatically improved by loading at acidic pH or on prefixed target cells. Since MHC class I restricted peptides are generally identified by screening of potential candidate peptides binding to the HLA molecule, B44-restricted epitopes may have been missed in the past and the approaches outlined in this paper should allow their easier detection. Because HLA-B44 is the most common HLA-B allele in the Caucasian population, being expressed in about 24% of individuals [[30]], and given the limited number of HLA-B44-restricted tumor antigens identified so far, efforts will clearly be needed to expand the list of such ligands and offer to a higher number of patients a perspective for future peptide-based immunotherapies. Peptide loading at acidic pH was already used for direct loading of dendritic cell (DC)-derived exosomes [[31]], and might also prove particularly useful when peptide-pulsed DCs or exosomes are used as vaccine vehicles. Alternatively, vaccination modalities relying on endogenous presentation, such as DNA- or RNA-based vaccines may allow efficient loading of HLA-B44. In summary, our results strongly suggest that specific requirements of particular peptide vaccines must be carefully taken into consideration in order to allow for their maximal effect in patients and for the success of future immunotherapies.

Materials and methods

Cell lines

Melanoma cell line LB373-MEL (HLA-A*02:01, -A28, -B*44:02, -B*53:01/02, -Cw04, -Cw*05:01), line 721.220, a human lymphoblastoid cell line lacking HLA-A, -B, and tapasin genes [[32]], and the Epstein-Barr virus transformed B cells, LB1801-EBV (HLA-A2, -A11, -B22, -B*44:02, -Cw5, -Cw9) and LB1972-EBV were grown in IMDM (Invitrogen Corporation, Carlsbad, CA, USA) containing 10% FCS (Thermo Fisher Scientific Inc., Waltham, MA, USA). 721.220, stably transfected with HLA-B*44:02 [[33]] were transduced as previously reported [[34]], using a retroviral vector M1-CSM that encodes the full length MAGE-A1 fused to the invariant chain and the truncated form of the human low-affinity nerve growth factor receptor (LNGFr) separated by an IRES (internal ribosome entry site) sequence. Transduced cells were positively sorted using an antibody recognizing LNGFr. The culture medium of the 721.220.B*44:02 transfectants (expressing or not MAGE-A1) was supplemented with 0.5 mg/mL G418 (Roche, Basel, Switzerland). COS-7 cells were cultured in DMEM (Invitrogen) supplemented with HEPES and glucose (Invitrogen) and 10% FCS. WEHI 164 clone 13 were cultured in RPMI 1640 medium (Invitrogen) supplemented with 5% FCS. All culture media were supplemented with L-arginine (116 mg/L), L-asparagine (36 mg/L), L-glutamine (216 mg/L), penicillin (100 U/mL), and streptomycin (100 μg/mL) (Invitrogen).

DCs and T-cell precursors

Blood was obtained from hemochromatosis patients, PBMCs were isolated as previously described [[35]] and frozen until further use. Lymphocyte-depleted PBMCs were left to adhere for 2 h. Adherent cells were cultured in the presence of 10 ng/mL IL-4 and 100 ng/mL GM-CSF in RPMI medium, supplemented with amino acids and 10% FCS and fed on day 2 and 4 by adding fresh medium, IL-4 and GM-CSF. The day before the stimulation, T cells were thawed and left overnight in IMDM supplemented with amino acids, 10% human serum and IL-2 (5 U/mL). The next day, CD8+ T cells were isolated using an anti-CD8 monoclonal antibody coupled to magnetic microbeads (Miltenyi Biotech, Köln, Germany) and stimulated with irradiated (100 Gray) autologous APCs as described in the next section.

Activation of MAGE-A1 CTL precursors

Autologous DCs from donor LB1801 were infected with recombinant canarypoxvirus, ALVAC-MAGE-A1 at a multiplicity of infection (MOI) of 30, and washed. CD8+ T cells (150,000) and 30,000 infected DC were cocultured in microwells in IMDM medium supplemented with amino acids, 10% human serum and 1000 U/mL IL-6 and 10 ng/mL IL-12. The CD8+ T cells were then stimulated weekly three times using autologous DCs infected with adenovirus expressing MAGE-A1, and one time with infected autologous PBMC. From day 7, cells were grown with medium supplemented with 10 U/mL IL-2 and 5 ng/mL IL-7. On day 28, MAGE-A1 specific responder CD8+ T-cell microculture was cloned by limiting dilution using autologous EBV-B cells infected with Yersinia-MAGE-A1 as stimulator [[35]], in order to avoid proliferation of CTLs directed against adenoviral vector derived antigens. Stimulation was performed using allogeneic EBV-B cell as feeder cells in medium containing 50 U/mL IL-2. CTL clones were maintained by weekly stimulation with irradiated tumors or 30 ng/mL OKT3, according to previously described protocols [[8, 36]].

Assay for cytolytic activity

The lytic activity of the CTLs was tested in a standard chromium release assay, as described previously [[37]]. Before use, autologous EBV-B target cells were (Fig. 1A) or not (Fig. 1D) infected for 2 h with vaccinia virus encoding MAGE-A1 as described [[35]]. In Fig. 1D, the targets were loaded with the indicated peptides at a concentration of 1 μM before addition of CTLs at an effector-to-target ratio of 10.

Peptides and CTL activation assays

Peptides used were synthesized on solid phase using Fmoc for transient N-terminal protection and characterized by mass spectrometry. The lyophilized peptides were solubilized at 20 mg/mL in DMSO and stored at −80°C. For CTL activation assays, various concentrations of peptides were pulsed on the indicated target cells (30,000 cells per well). After 45 min, cells were washed. The IFN-γ secreted by CTLs (10,000 cells per well) was measured by ELISA, after an overnight incubation. Where indicated, targets were fixed by 10 min incubation in a 1% PFA solution, followed by three extensive washes in PBS. In Fig. 4B, peptides were pulsed for 20 min in 50 mM sodium citrate buffer pH 5.5 or 7. Cells (40,000 cells per well) were washed three times before addition of CTLs (10,000 cells per well). As a control, acid-treated cells were neutralized by adding extensive amounts of medium prior pulsing with the peptide.

Transient transfection of COS-7 cells and HEK-293-EBNA

Transient transfection of COS-7 cells was performed using the DEAE-dextran-chloroquine method [[38]]. COS-7 cells (15,000) were transfected in duplicate with 50 ng of plasmids. The HLA-A*02:01, -B*44:02, -Cw5 cDNA, the full-length MAGE-A1 cDNA, and the truncated fragments of MAGE-A1, which were obtained by PCR, were cloned into plasmid pcDNA1/amp (Invitrogen). After 24 h, transfected cells were incubated with 1500 CTLs in 200 μL of culture medium supplemented with 25 U/mL rIL-2. After 20 h of coculture the supernatant was collected. TNF production was estimated by testing the cytotoxicity of the supernatant on WEHI 164 clone13 in a MTT colorimetric assay [[39]]. HEK-293 were transiently transfected with 50 ng of HLA-B*44:02 (in pcDNA1), HLA-A*01:01 (in pcDNA1), and/or empty vector using lipofectamin according to the manufacturer protocol (Invitrogen). Twenty-four hours after transfection, 15,000 CTL4 or 5000 CTL MZ2-82/30 were added in 200 μL of culture medium supplemented with 25 U/mL rIL-2. After 20 h of coculture the supernatant was collected and IFN-γ content was measured by ELISA.

Electroporation of 721.220 cells

A total of 4 × 106 721.220 cells were electroporated (450 V, 150 μF, 4 mm Cuvette, Gene pulser) with 30 μg of MAGE-A1/pcDNA1 or empty vector together with 15 μg of a plasmid DNA encoding monomer DsRed or EGFP (Clontech, Mountain View, CA, USA) [[22]]. After 20 h, electroporated cells (25,000 cells per well) were cocultured with 20,000 CTLs and 25 U/mL IL-2. Stable transfectants 721.220.A1 were obtained by electroporating (1 pulse, square wave 500 V, 1 ms in a 4 mm Cuvette) 721.220 expressing HLA-B*44:05 with a pCEP4 construct containing the cDNA encoding HLA-A*01:01. Transfected cells were then selected using 400 μg/mL hygromycin, and sorted for HLA-A*01:01+ cells using the anti-HLA-A*01:01 antibody GV5D1 [[40]].

Electroporation of the peptides

A total of 4 × 106 LB1801-EBV cells were electroporated in 400 μL electroporation buffer (K2HPO4/KH2PO4 10 mM pH 7.4, MgCl2 1 mM, saccharose 250 mM) with 30 μM peptide in 0.4 cm Genepulser cuvettes (Biorad), using a PA4000 electroporator (Cyto Pulse Sciences) functioning in rectangular wave mode with 10 pulses of 5 ms at 310 V separated by 995 ms intervals [[41]]. Cells were diluted to a peptide concentration of 3 μM and incubated for 5 h. Cells were then washed and plated at 10,000 cells per microwell, before addition of 5000 CTLs.

Peptide stabilization assay

721.220.B*44:02 cells were starved in methionine- and cysteine-free medium for 40 min and pulsed with (35S)-methionine/cysteine (Perkin Elmer, Waltham, MA, USA) at 500 μCi/mL/5 × 106 cells for 30 min. Cells were then washed in ice-cold PBS, and lysed in 1% TRITON X-100 (American bioanalytical, Natick, MA, USA) in TRIS buffer solution (pH 7.4) in the presence of the indicated concentration of peptide. After 30 min lysis, postnuclear supernatants were precleared overnight at 4°C using protein A and rabbit serum. Precleared samples were then immunoprecipitated using W6/32 antibody. Washed precipitates were resolved on a 9% SDS-PAGE gel. Quantifications were performed with ImageQuant software 5.2 (GE Healthcare, UK).

Peptide elution

Peptides were acid eluted from 25 × 106 cells (Fig. 3) in 50 mM sodium citrate buffer pH 3.0 for 90 s, then neutralized using three volumes of 150 mM Na2HPO4, pH 10.5. Eluates were cleaned by centrifugation, and concentrated on a Sep-pak Plus C18 column (Waters, Milford, MA, USA). Peptides were eluted in acetonitrile 60% and dried by speed vacuum. Dried peptides were resuspended in 250 μL X-VIVO-10 (Lonza, Köln, Germany) and purified on a 3 KDa microcon YM3 column (Millipore, Billerica, MA, USA) before pulsing on PFA-fixed APC (Fig. 3). In Fig. 2B, HLA were immunopurified from 1 × 109 cells using W6/32 antibody and fractionated by HPLC after injection on a narrow-bore 2 × 150 mm Deltapack C18 column (Waters) as described [[41]].

Acknowledgments

We thank Aline Depasse and Sabrina Ottaviani for technical assistance and Julie Klein for aiding in the preparation of this manuscript. 220.B4405.TPN cell line is a kind gift from Pr. J. McCluskey. Support for this work was provided by grants from the European Union under the Sixth Programme (CancerImmunotherapy; LSHC-CT-2006-518234), the Belgian Programme on Interuniversity Poles of Attraction initiated by the Belgian State (Prime Minister's Office, Science Policy Programming), the Fonds National de la Recherche Scientifique (FNRS, Belgium), the Fondation contre le Cancer (nonprofit organization, Belgium), the Fonds J. Maisin (Belgium), and the Fondation Salus Sanguinis (Belgium), the Howard Hughes Medical Institute, Yale SPORE in Skin Cancer Grant 5P50 CA121974, the NIH/National Institute of General Medical Sciences. NV was supported by a Marie Curie Outgoing International Fellowship (OIF) from the European Union and a post-doctoral fellowship from the F.R.S.-FNRS, Belgium, RaML by a Cancer Research Institute Fellowship, and RoML by a Marie Curie postdoctoral fellowship from the “Training and mobility of researchers” program of the European Commission. AB is a post-doctoral fellow with Fonds Wetenschappelijk Onderzoek-Vlaanderen (FWO-V).

Conflict of interest

The Ludwig Institute for Cancer Research has filed a patent application on the peptide described in this manuscript.

Abbreviations
CRT

calreticulin

DEAE

diethylaminoethyl

FMOC

fluor-enylmethyloxycarbonyl

HC

heavy chain

IRES

internal ribosome entry site

MTT

3-(4,5-dimethylthiazol-2-yl)-2,5-diphenyltetrazolium bromide

PFA

paraformaldehyde

TPN

tapasin

TGN

trans-Golgi network

Ancillary