Heat shock proteins (HSPs) have two unique roles as constituents of tumor vaccines: (i) to shuttle associated tumor antigens into professional antigen-presenting cells (APCs) and (ii) to activate professional APCs. Here we investigated the shuttle function of the HSP gp96 (glycoprotein 96) for a human melanoma peptide antigen MART-1 that was noncovalently bound to gp96 in vitro. This in vitro complexing reaction was optimized using the radioiodinated MART-1 peptide and human gp96. Up to 20% of gp96 molecules could bind the peptide, assuming a 1:1 molar ratio. The binding was temperature-dependent and thus reversible. At -20°C, 95% of the peptide remained complexed after 24 h, but 25% and 60% of the peptide dissociated at 37°C within 6 and 24 h, respectively. This observation suggests that under the physiological conditions in APCs, spontaneous peptide dissociation from gp96 complexes may facilitate the delivery of peptide antigen into antigen presentation pathways. The gp96/MART-1 complexes stimulated an HLA A2-restricted MART-1-specific CTL clone dependent on the amount of complexed peptide and the presence of HLA-A2-positive APCs. The reaction was peptide-specific and could be blocked by an excess of untreated native gp96. These results show for the first time that peptide antigens from in vitro reconstituted gp96/peptide antigen complexes can be cross-presented by human APCs. These findings extend the scientific basis for further evaluating the use of either endogenous or in vitro reconstituted gp96/tumor-antigen complexes as tumor vaccines.
This article was published in Cancer Immunity, a Cancer Research Institute journal that ceased publication in 2013 and is now provided online in association with Cancer Immunology Research.
Srivastava and coworkers have shown that vaccinating inbred mice with certain HSPs such as gp96, hsp90, and hsp70 purified from a syngeneic murine tumor elicits preventive transplantation immunity only against that specific tumor (1, 2, 3, 4). Later it was shown that this immune response also had therapeutic effects against established tumors (5, 6). In recent years, studies have provided accumulating evidence that the specific immunogenicity of an HSP results from its chaperone function for endogenous peptide antigens. First, various experimental settings demonstrated that vaccinating inbred mice with gp96 preparations induced CTL-specific responses against the peptide antigens expressed by the cells from which the gp96 vaccine had been purified (7, 8, 9). Second, immunogenic peptide antigens have been isolated from HSP preparations derived from antigen-positive cells (10). For example, Meng and coworkers succeeded in isolating hepatitis B virus (HBV) -specific peptides from gp96 that were purified from HBV-induced hepatocellular cancer (11). Third, peptide transport experiments have identified gp96 as a peptide-binding protein in the endoplasmatic reticulum (12, 13). Fourth, in vitro studies have demonstrated that gp96 and hsp70 bind peptides under appropriate conditions (14). Furthermore, a possible peptide-binding region in the gp96 molecule has been identified (15, 16), and the peptide-binding mechanisms have been elucidated (17, 18).
One precondition for the efficiency of HSP/peptide antigen complexes as vaccines consists of a receptor-mediated endocytosis of these complexes by APCs and the channeling of the bound peptides into the MHC class I antigen re-presentation pathway. Recently, evidence for receptor-mediated endocytosis of HSPs by APCs and cross-presentation of HSP-associated antigens by APCs has been reported (19, 20). In addition, Srivastava and coworkers identified the alpha2-macroglobulin receptor, CD91, as a receptor for the endocytosis of gp96, hsp70, and hsp90 (21, 22). However, a more recent study provides evidence for an additional CD91-independent gp96 internalization pathway that functions in peptide antigen re-presentation (23). Another important basis for the efficient use of HSP/peptide antigen complexes as vaccines is their ability to activate the immune system. For example, they can activate a professional APC and induce its maturation (24, 25, 26, 27). So far, only two reports showed cross-presentation of human HSP-associated tumor peptide antigens by human APCs utilizing hsp70 and gp96 preparations from human melanoma cell lines (28, 29).
In the present work, we investigated whether MART-1, bound to gp96, can be cross-presented by human APCs to a peptide-specific CTL clone. To control the amount of antigen bound to gp96, we complexed defined amounts of the synthetic MART-1 peptide with gp96 in vitro. Under optimized conditions, it was possible to load up to 20% of gp96 molecules with the MART-1 peptide, assuming a molar ratio of 1:1. These gp96/MART-1 peptide complexes were tested in cross-presentation tests with HLA-A2-positive APCs and an HLA-A2-restricted MART-1-specific CTL clone. We could reproducibly show a peptide-specific and APC-dependent CTL stimulation with these gp96/MART-1 complexes. In addition, this stimulation was HLA-A2-restricted as well as dependent on the amount of complexed peptide; an excess of uncomplexed gp96 blocked this stimulation.
In summary, we showed for the first time that peptide antigens from in vitro reconstituted gp96/peptide antigen complexes could be cross-presented by human APCs. These findings extend the scientific basis for a personalized tumor vaccine with endogenous or in vitro reconstituted gp96/peptide antigen complexes.
In vitro reconstituted gp96/MART-1 complexes
The HSP gp96 was purified from a mycoplasma-free, human EBV-transformed B-cell line that did not express the MART-1 antigen recognized by CTL 2/9. The high purity and integrity of each gp96 preparation was tested by silver-stained SDS-PAGE gels and by immunoblot analysis (Figure 1). The lipopolysaccharide concentration of the gp96 preparations was determined by limulus amebocyte lysate assays and was found to be generally low, with endotoxin activity between 0.001 and 0.008 EU/µg gp96. Lipopolysaccharide, added at comparable concentrations, never had any biological effect in the experiments described here. In addition, all gp96 preparations were controlled for remaining traces of concanavalin A (Con A) using a Con A-specific ELISA. Our gp96 purification method eliminated Con A from the gp96 isolates below detectable levels. Gp96 was complexed with the iodine-125-labeled synthetic nonameric MART-127-35 peptide (AAGIGILTV) as described previously (14). Under these experimental conditions, only 0.14 ng of peptide antigen could be complexed with 1 µg gp96. The complexing reaction was carefully optimized by systematically varying the experimental conditions (Figure 2). The following conditions proved to be optimal: a molar ratio of 1:200 for gp96 and MART-1 peptides, a gp96 concentration of 300 µg/ml, an incubation time of 30 min at 50°C, a reaction temperature of 50°C, and a DMSO concentration of 20%. Within the tested range, the sodium concentration of the sodium phosphate buffer did not improve the complexing reaction; thus, the 0.7 M sodium concentration was retained. After an incubation time of 30 min, the equilibrium between peptide and HSP had not yet been reached, but the incubation time was not extended to avoid aggregation and/or degradation. A DMSO concentration below 20% could not be used because of the high hydrophobicity of the MART-1 peptide.
Under these optimal conditions, up to 1.7 ng peptide complexed with 1 µg gp96, a 10-fold increase in the complexing efficiency as compared with the initial results described above. To test the efficiency of these gp96/peptide complexes in cross-presentation tests, the remaining free peptides had to be removed (Figure 3). Up to 150 times more unbound peptides than complexed peptides per 1 µg of gp96 were detected directly after the complexing reaction. On average, 96% of the unbound peptides could be removed by a repeated PD-10 gel filtration. After this step, there were still 6 times more unbound peptides per 1 µg gp96 than complexed peptides. The second step was an extensive ultrafiltration of the complex solution with a 30-kDa cut-off membrane. This led to a further reduction of the amount of free peptides to less than 10% of the amount of complexed peptide per 1 µg of gp96. In total, more than 99.9% of unbound peptides could be removed by sequential gel filtration and ultrafiltration. By itself, extensive ultrafiltration for separating unbound peptides, as described previously (9, 14), was not enough to reduce the concentration of unbound peptides to a level that was not active in the cross-presentation tests (data not shown). This was due to aggregation of the hydrophobic MART-1 peptide at higher concentrations during the ultrafiltration process, leading to a less efficient ultrafiltration. This problem could be circumvented by primary gel filtration followed by ultrafiltration.
The gp96/MART-1 complexes were subjected to different temperatures in order to determine their stability (Figure 4). In a representative experiment, the complex proved to be relatively stable at -20°C; only about 5% of the complexed peptides were dissociated within a 24-h period. The amount of peptides dissociating increased with increasing temperature, whereas the gp96 itself remained stable as determined by gel electrophoresis. After 24 h at 20°C, almost 30% were dissociated, and at the physiological temperature of about 37°C, more than 60% of the complexed peptides became unbound (Figure 4A). The peptide dissociation at 37°C was studied even more precisely. After 1 h had elapsed, only 2% of the peptides were dissociated, whereas the amount of dissociated peptides had increased to 25% after 6 h and 60% after 18 h (Figure 4B).
The immunological activity of the gp96/MART-1 complexes that were reconstituted with the nonameric MART-127-35 peptide (AAGIGILTV) was studied in cross-presentation tests in a human cell system. Suboptimal complexes were used as controls: They were reconstituted at a reaction temperature of 4°C. These suboptimal control complexes contained fewer than 10% of the complexed peptides than when the complexes were generated at 50°C (Figure 5) and were used as controls for the effects of complexed peptides. The gp96/MART-1 complexes, reconstituted at optimal (50°C) and suboptimal (4°C) conditions in vitro, were coincubated with HLA-A2-positive APCs and the MART-1-specific CTL clone 2/9 and were analyzed for their ability to induce MART-1-specific and APC-dependent CTL activation (Figure 6). The optimal gp96/MART-1 complexes, reconstituted at 50°C, induced a CTL stimulation that was highly dependent on the amount of added complex and on the presence of APCs (Figure 6).
In contrast, the suboptimal complexes reconstituted at 4°C induced only a marginal CTL stimulation, which was not dependent on the presence of APCs. These results indicate that the CTL stimulation was highly dependent on the amount of complexed peptides. For example, 1.25 µg/ml of the optimal complex, containing 2.1 ng/ml of complexed MART-1 peptide, led to a CTL activation with an IFN-gamma secretion of 230 pg/ml. The suboptimal complex, containing less than 10% of complexed peptide compared with the optimal complex, induced a comparable CTL stimulation only at 10 times the concentration (data not shown).
These results also rule out a false-positive CTL stimulation by copurifying contaminations of gp96 or free peptides, because the optimal complexes only differed from the suboptimal complexes by the complexing temperature, as they were processed identically. The gp96/MART-1 peptide complexes were about 12 times more efficient than free MART-1 peptides in the cross-presentation assays, based on the amount of complexed peptides. This was shown repeatedly by peptide titration in cross-presentation assays. In these assays the CTL stimulation by gp96/MART-1 peptide complexes required at least 0.4 ng/ml of complexed peptide, whereas the threshold peptide concentration for the CTL stimulation induced by free MART-1 peptides was at least 5 ng/ml (data not shown). In an additional effort to avoid CTL stimulation by free nonameric MART-1 peptides in cross-presentation assays, an attempt was made to generate complexes of gp96 with elongated MART-1 peptides. However the 15mer MART-121-35 peptide (YTTAEEAAGIGILTV) still directly stimulated the CTL clone and therefore did not provide any advantage compared to the nonameric peptide. The 20mer MART-121-40 peptide (YTTAEEAAGIGILTVILGVL) could not successfully be complexed with gp96 because it precipitated during complexing reactions due to its high hydrophobicity.
Cross-presentation requires the presentation of the peptide antigen by the MHC molecules of the APC. The CTL stimulation was induced by the gp96/MART-1 complexes and was largely MHC class I-dependent, as there was only a marginal CTL stimulation when HLA-A2-negative APCs, but not HLA-A2 positive APCs, were used (Figure 7). Further evidence for the HLA-A2-restriction of the CTL stimulation came from experiments in which HLA-A2-positive APCs were preincubated with the anti-HLA-A2 antibody MA2.1 for 1 h before the gp96/MART-1 complex was added. This led to a reduced CTL stimulation, like the marginal CTL stimulation obtained with HLA-A2-negative APCs (data not shown).
Another requirement for the cross-presentation of HSP-associated antigens by APCs is the binding of the HSP/antigen complexes to an APC receptor and the subsequent endocytosis of the HSP/peptide complexes. A strong indication that binding of the gp96/MART-1 complexes to the APCs was necessary for the induction of the CTL stimulation in this study came from cross-presentation tests in which a 4-fold to 8-fold excess of uncomplexed gp96 was added. This excess of uncomplexed gp96 reduced the CTL activation by as much as 50% (Figure 8). In this experiment, toxic side effects of the uncomplexed gp96 were excluded by a control, showing that the activation of CTLs and APCs through phytohemagglutinin (PHA) was not influenced by a coincubation with the same amount of uncomplexed gp96.
In general, for all these experiments only fresh in vitro reconstituted complexes were used. Nevertheless, gp96/MART-1 complexes incubated for 7 d at -20°C still showed significant CTL activation in cross-presentation tests (data not shown). Thus, these results are in concordance with our finding that only relatively small amounts of peptide dissociate from the complex stored at -20°C, as described above.
In recent years, several studies presented evidence for a receptor-mediated endocytosis of HSP by APCs (19, 20) and of cross-presentation of HSP-associated peptide antigens from autologous HSP preparations or in vitro reconstituted HSP/peptide antigen complexes by murine APCs (9, 14, 30). In many of these studies, the in vitro reconstitution of complexes of murine gp96 or HSP70 with peptide antigens has been described (9, 14, 30, 31, 32, 33). Here we present the first study of cross-presentation of a human melanoma peptide antigen from in vitro reconstituted human gp96/peptide complexes that employs human CTLs and APCs. Following previously published experimental conditions, the complexing efficiency of the human melanoma peptide antigen MART-1 and human gp96 was comparable to published efficiencies (14, 30). However, by systematic optimization of the reaction conditions, we were able to improve the complexing efficiency 10-fold. Under optimal conditions, the gp96/MART-1 complexes contained about 1.7 ng peptide antigen per 1 µg of gp96. Assuming a molar ratio of gp96 to peptide of 1:1, about 16% of the gp96 molecules could be complexed with a peptide antigen. The main factors for this improved complexing efficiency were the increase in the molar ratio between the MART-1 peptide and gp96, the increased reaction concentration of gp96, and the complexing time. To avoid aggregation and/or degradation of the proteins and peptides, the complexing time could be increased from 10 to a maximum of 30 min, although an equilibrium between free and complexed peptides had not yet been reached at 30 min. In addition, the high hydrophobicity of the nonameric MART-1 peptide required a DMSO concentration of 20% during the complexing reaction to avoid aggregation. However, we found that increasing the DMSO concentration above 20% decreased the complexing efficiency. From these results, it follows that for individual peptide antigens the complex formation with HSPs may need to be quantified and optimized.
Previous studies in murine models showed an in vitro CTL stimulation by gp96/peptide antigen complexes which was about 100 times more effective with respect to the amount of bound peptide than the CTL stimulation by free peptides (9, 14). In contrast, the gp96/MART-1 complexes in our experiments were only about 12 times more effective in CTL stimulation than free MART-1 peptides. However, this lower efficiency in CTL stimulation was compensated for by the higher complexing efficiency in our study as compared with the previous studies. The reasons for this difference in effectiveness may lie in differences among the types of APCs and/or in species differences. Additionally, there is recent data that suggests that the lower effectiveness of the in vitro reconstituted gp96/MART-1 complexes per amount of complexed peptide in this study as compared to previous studies (9, 14) is a result of the elimination of detectable Con A in our gp96 preparations (34). It was reported that endogenous gp96 preparations devoid of Con A lose their antigen re-presentation activity in vitro and in vivo, whereas the presence of low-level Con A in gp96 preparations potentiates the gp96-specific immunological responses (34). Interestingly, our results therefore suggest that a high amount of in vitro complexed peptides with gp96 can overcome a dependency of cross-presentation on low levels of Con A within gp96 preparations.
The monocytic APCs used in our study were purified from PBMCs by plastic adherence according to the study by Castelli et al. (28). This study used hsp70 preparations from a melanoma cell line and is, thus far, the only study that was able to show the cross-presentation of endogenous human tumor antigens by human APCs (28).
It was particularly important in this study that we excluded a contamination of the complexes by residual free peptides as causing CTL stimulation and that we showed a clear dependence of the CTL stimulation on the amount of complexed peptide. Another study avoided the problem of contaminating free peptides by using prolonged peptides with the antigen epitope, which only stimulated CTLs after being complexed to the HSP, but not as a free peptide (22). This approach was not feasible for the hydrophobic MART-1 peptide antigen since a MART-1 15mer peptide still significantly stimulated the CTLs and a 20mer precipitated during in vitro complexing reactions due to its hydrophobicity. Therefore we had to rule out CTL stimulation by contaminating free nonameric MART-1 peptides by (i) maximally removing free peptides and by (ii) demonstrating that CTL stimulation in our experiments was strictly dependent only on the amount of complexed peptide. The removal of free peptides was optimized by using iodine-125-labeled synthetic nonameric MART-1 peptides for quantitation of complexed and free peptides. During the course of these experiments, we recognized that ultrafiltration with a 30-kDa cut-off membrane as previously described (9, 14, 30) was not sufficient to reduce free peptides to levels not inducing CTL stimulation. Most likely this was due to the high hydrophobicity of the MART-1 peptide and its tendency to aggregate, which in turn led to a less efficient ultrafiltration. Therefore, to achieve a maximal complex purification, gel filtration, as in another study (15), had to be repeated and followed by repeated ultrafiltration up to a theoretical dilution factor of 1:108. With this approach, about 99.9% of the free peptides was removed, reducing the amount of free peptide to less than 10% of the amount of complexed peptide. The strict dependence of the CTL stimulation on the amount of complexed peptide was demonstrated by the observation that complexes reconstituted under optimal conditions at 50°C caused CTL stimulation, whereas complexes reconstituted under suboptimal conditions at 4°C at identical concentrations did not. The optimal complexes contained 10 times more complexed peptide than the suboptimal complexes, while both complexes contained the same minimal amount of free peptide because of identical purification conditions.
We verified two important characteristics of T-cell stimulation by HSP/peptide complexes. First, the CTL stimulation by gp96/MART-1 peptide complexes was MHC class I-restricted (9, 14, 28). This was shown by blocking the complex-induced stimulation of the MART-1-specific, HLA-A2-restricted CTL clone with anti-HLA-A2 antibodies and by demonstrating the dependence of CTL stimulation on the HLA-A2 expression of the APCs. Second, an excess of native gp96 could block the CTL activation induced by gp96/MART-1 complexes. This observation is in concordance with the postulated HSP receptor on APCs and with receptor-mediated endocytosis of HSPs (19, 35). Recently, this receptor was described as the CD91 receptor (alpha2-macroglobulin receptor; LDL receptor-related protein), a member of the low-density lipoprotein (LDL) family of scavenger receptors (21). CD91 seems to be a common receptor for the endocytosis of gp96, hsp70, hsp90, and calreticulin (22). However, a more recent study provides evidence for a CD91-independent, receptor-mediated gp96 internalization directing gp96 into the MHC class I antigen re-presentation pathway (23).
In some of our cross-presentation experiments, the control complex or even irrelevant gp96 preparations induced a minimal IFN-gamma secretion, for example, with HLA-A2 negative APCs. This observation has been previously described as an unspecific and MHC class I-independent stimulation of CTLs by gp96 (36).
Another new aspect of our study was the stability of the in vitro reconstituted gp96/MART-1 complexes. Within 24 h and at -20°C, only about 5% of the complexed peptide antigens dissociated from the complex, while about 30% dissociated at room temperature and more than 60% dissociated at 37°C. The stability of the complexes at -20°C is underlined by the cross-presentation experiments using gp96/MART-1 complexes that had been incubated for 7 d at -20°C and still induced a specific CTL stimulation (data not shown). The temperature sensitivity of the complexes can be explained by the same mechanisms that are crucial for the in vitro reconstitution of HSP/peptide complexes. Increasing temperatures result in a conformational change of the gp96 molecule that leads to an open configuration. In this configuration, hydrophobic binding pockets are opened, which facilitate peptide dissociation or peptide exchange in the presence of excess free peptides (37). However, apart from increasing storage temperatures, these complexes are extraordinarily stable, which is also demonstrated by their resistance to the denaturing conditions of SDS-polyacrylamide gel electrophoresis, as shown by us in this study and in previous reports (1, 15). The observed temperature sensitivity demonstrates the importance of careful handling and storage of tumor tissue-derived HSP preparations for vaccine purposes. On the other hand, the temperature sensitivity of the complexes facilitates the dissociation of bound peptides at physiological temperatures after the endocytosis of complexes by APCs, a prerequisite for channeling peptide antigens into the MHC class I-dependent antigen presentation pathway.
The enhanced dissociation of peptides from complexes at 37°C might raise the concern that our results in cross-presentation assays are due to external peptide loading of APCs. However, this can be ruled out for the following reasons: (i) the endocytotic uptake of gp96 seems to occur very rapidly, within 1 h (19, 20, 38, 39), at which time 98% of the complex is still intact; (ii) even if after 24 h at 37°C 60% of the peptide has been dissociated from the gp96/MART-1 complexes at active concentrations between 0.4 and 1.25 µg/ml, the resulting concentrations of free MART-1 peptides are at most between 0.5 and 1.5 ng/ml, which is nowhere near enough to induce CTL stimulation; and (iii) the specific blocking of the gp96/MART-1-induced CTL stimulation by native gp96 cannot be explained by external peptide loading of APCs.
In summary, our study shows for the first time that peptide antigens from in vitro reconstituted gp96/MART-1 complexes can be cross-presented by human APCs and specifically stimulate human HLA-A2-restricted antimelanoma T cells. These results confirm the concept of the cross-presentation of gp96-associated peptide antigens in the context of the human cellular immune system and thus extend the scientific basis to further evaluate the use of HSPs as tumor vaccine. We were able to show that the amount of complexed peptide antigen plays a crucial role in the CTL stimulation by gp96/peptide antigen complexes. In this case, our data support a vaccine approach using in vitro reconstituted complexes of HSP with defined tumor antigens. This approach would at least offer the possibility of conclusive immunomonitoring in pilot clinical trials and, therefore, the chance to optimize the HSP vaccination strategy with respect to dose, injection mode, intervals, and adjuvants. In addition, peptide binding could be monitored during the purification process.
Materials and methods
SK-MEL-29.1 is a clonal cell line from the melanoma cell line SK-MEL-29 (40) which was maintained in DMEM (GIBCO-BRL, Eggenstein, Germany), 10% FCS (Biochrom, Berlin, Germany), and 1% glutamine (GIBCO-BRL). SK-29-EBV is the Epstein-Barr-Virus (EBV) -transformed lymphoblastoid B-cell line autologous to SK-MEL-29.1 (40), which was maintained in RPMI 1640 (GIBCO-BRL), 10% FCS, and 1% glutamine. The HLA-A2-restricted CTL clone 2/9, specific for the Melan-A/MART-127-35 peptide (AAGIGILTV), was kindly provided by Thomas Wölfel III, Medizinische Klinik, University of Mainz. The specificity and culture requirements of this CTL clone, which recognizes the autologous melanoma cell line SK-MEL-29.1 in an HLA-A2-restricted way, have been described previously (41, 42).
Gp96 was purified from the human EBV-transformed B-cell line, SK-29-EBV, as described originally for murine gp96 (1), with slight modifications as described later by the same group (14). Briefly, cells were homogenized in hypotonic buffer (30 mM NaHCO3, 0.2 mM PMSF, pH 7.1) by Dounce homogenization. The lysate was centrifuged at 105 x g for 90 min. The supernatant was fractionated by ammonium sulfate precipitation (50-80% saturation) and applied to a Con A-Sepharose column (Amersham Pharmacia, Uppsala, Sweden). Glycoproteins were eluted by 10% methyl-alpha-D-mannopyranoside. The next purification step, anion exchange chromatography with DEAE Sepharose (Amersham Pharmacia), was modified from previously published methods (1, 14) in that the sample was loaded onto the column in the presence of 10% methyl-alpha-D-mannopyranoside. This proved to be necessary to remove traces of Con A bound to gp96 (43). Briefly, the Con A-Sepharose eluate was applied without buffer exchange to the DEAE Sepharose columns that were equilibrated with 5 mM sodium phosphate buffer (pH 7.0) containing 300 mM NaCl and 10% methyl-alpha-D-mannopyranoside. The columns were washed with at least 5 column volumes with the equilibration buffer and another 5 column volumes without methyl-alpha-D-mannopyranoside. Finally, gp96 was eluted by 700 mM NaCl in 5 mM sodium phosphate buffer. Protein concentrations were determined by Bradford assays photometrically. The purity of the protein preparations was controlled by SDS-PAGE and silver staining of the gels. The identity of the gp96 protein was controlled by immunoblot analysis with rat anti-grp94 antibody (Stress Gen, Victoria BC, Canada) as the first antibody and a rabbit anti-rat antibody (Dako, Glostrup, Denmark) as the secondary antibody. The amount of Con A bound to gp96 was controlled by a Con A-specific ELISA using a primary mouse anti-Con A monoclonal antibody purified from the 71A7 hybridoma (ATCC, Manassas, VA) as capture antibody, a polyclonal rabbit anti-Con A antibody (Sigma, Darmstadt, Germany) as detecting secondary antibody, and a peroxidase-conjugated goat anti-rabbit IgG (Dianova, Hamburg, Germany) (43). The Con A concentration in the gp96 preparations was always below a detection limit of 0.5 pg/ml, employing the modified purification method as described above. These gp96 preparations contained up to 300 times less Con A compared to the gp96 purified without this modification. The amount of endotoxin present in the gp96 stock solutions was determined by means of the quantitative chromogenic limulus amebocyte lysate assay (QCL-1000, BioWhittaker, Walkersville, MD) according to the manufacturer's instructions. Mean endotoxin concentrations in the HSP preparations were calculated from an E. coli 0111:B4 lipopolysaccharide standard curve and expressed in endotoxin units (EU) ml. The endotoxin activity relative to the amount of gp96 was low (0.001-0.008 EU/µg protein).
In vitro reconstituted gp96/peptide antigen complexes
The gp96 complexing reaction with the nonameric MART-127-35 peptide (AAGIGILTV) (Affina Immunotechnik GmbH, Berlin, Germany) was optimized using iodine-125-labeled peptides (Amersham Pharmacia, Freiburg, Germany). In the optimized complexing reaction, the MART-1 peptides were incubated with 300 µg/ml gp96 at a peptide/gp96 molar ratio of 200:1 in sodium phosphate buffer containing 0.7 M NaCl for 30 min at 50°C, followed by incubation at room temperature (21°C) for another 30 min. To determine the quality of the gp96/peptide complexes, they were electrophoretically separated by SDS-PAGE and visualized by Coomassie staining and autoradiography. For the quantification of the complexing reaction and for the analysis of the complex stability, the gel band of the gp96/MART-1 complex and, as background controls, gel slices of the same size above and below the gp96/MART-1 band were cut out and counted separately in a gamma-counter. The average background radioactivity of the control gel slices was subtracted from the radioactivity of the gp96/MART-1 band. The quantity of peptides in this band could be calculated using the known radioactivity of the radiolabeled peptides as reference. For the analysis of the complex stability, the amount of complexed peptide was analyzed from samples of gp96/MART-1 complexes after incubation at different temperatures and for different time intervals. The percentage of bound peptide was calculated by dividing the amount of complexed peptide after incubation of the complex by the amount of complexed peptide in the freshly prepared complex. Before complexes were used in cross-presentation tests, contaminant-free peptides were removed by repeated PD-10 gel filtration (Sephadex 25, Amersham Pharmacia, Uppsala, Sweden), followed by repeated ultrafiltration at 4°C with a 30-kDa Biomax filter (Millipore, Eschborn, Germany) and PBS buffer until a dilution factor of 1:108 was reached. With this method, and by using radiolabeled peptides for complex formation with gp96, we could calculate that more than 99.9% of free peptides was removed.
Plastic adherent peripheral blood monocytes served as APCs. Briefly, peripheral blood mononuclear cells (PBMCs) were obtained by Ficoll Hypaque density gradient centrifugation from buffy coats from healthy and HLA-A2-positive blood donors from the blood bank of the University Hospital of Mainz, Germany. The 96-well, U-bottomed plates were incubated for 5 min at 37°C and contained 200 µl/well of coating medium [X-Vivo 15 (BioWhittaker, Verviers, Belgium) supplemented with 1% glutamine, 1% penicillin-streptomycin, and 1.6% human serum]. Subsequently, 8x104 PBMCs were incubated in the coated wells with 50 µl/well of adherence medium [X-Vivo 15 (Bio Whittaker) supplemented with 1% glutamine, 1% penicillin-streptomycin, and 0.5% human serum] for 45 min. Finally, nonadherent cells were completely removed by washing the plates 3 times with 200 µl/well test medium [X-Vivo 15 (Bio Whittaker) supplemented with 1% glutamine and 1% penicillin-streptomycin]. The adherent monocytes were incubated with gp96/MART-1 complexes at the concentrations indicated for 1 h at 37°C. In blocking experiments with the HLA-A2- and HLA-B17-specific monoclonal antibody MA2.1 (44), the antibody was added at a concentration of 32 µg/ml and incubated for another hour at 37°C. For blocking experiments with noncomplexed gp96, the gp96 was added instead of the antibody MA2.1 at the concentration indicated. To test if the gp96 preparations had toxic side effects, the CTL 2/9 cells were stimulated together with the APCs with 1 µg/ml phytohemagglutinin (PHA) (Sigma) in the presence and absence of the noncomplexed gp96 at the same concentrations. Finally, 2x104 CTL 2/9 cells were added for a final volume of 200 µl/well. After a 24-h incubation at 37°C in a humid incubator with 5% CO2, supernatants were collected for a standard cytokine ELISA measuring the IFN-gamma concentration as a measure of CTL activation, as described recently (45). In all cross-presentation experiments, positive controls showed that the MART-1 specific CTL clone reacted with an excess of the MART-1 pepide (10 µg/ml) as the antigenic stimulus.
This work was supported by the Deutsche Forschungsgemeinschaft (Sonderforschungsbereich 432, A2 to M. H.) and Antigenics Inc.
- Received September 17, 2003.
- Accepted March 17, 2004.
- Copyright © 2004 by Michael Heike