| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
|
|
|
|||||||
|
|||||||||
* Ludwig Institute for Cancer Research and
Department of Pathology, Austin Health, Heidelberg, Victoria, Australia;
Ludwig Institute for Cancer Research, Lausanne, Switzerland; and
Ludwig Institute for Cancer Research, New York, NY 10021
 |
Abstract |
|---|
 Top Abstract IntroductionMaterials and MethodsResultsDiscussionDisclosuresReferences |
|---|
|
Introduction |
|---|
|
TopAbstractIntroductionMaterials and MethodsResultsDiscussionDisclosuresReferences |
|---|
The immunosurveillance hypothesis, where the immune system controls the early development of cancer, was proposed by Burnet decades ago and has gained strong experimental support in recent years (4). Once a tumor is established, the immune system again plays an active role to suppress the tumor, while at the same time the tumor actively evolves mechanisms to escape this pressure. This leads to changes in the antigenic phenotype of the developing tumor, a process described more recently as "immunoediting" (5), emphasizing the interplay between the immune system and tumor cells. Tumor cells have evolved numerous mechanisms to evade immune attack, such as down-regulation of MHC molecules, expression of immunoinhibitory cytokines, or expression of FasL (6). The vigor of such reactive responses in turn suggests that the immune response against specific tumor Ags is often a potent one. However, the extent and breadth of such antitumor T cell responses have not been well-characterized previously.
Since the discovery of the first human tumor Ag-derived peptide in the early 1990s (7), many more tumor Ags and their respective T cell epitopes have been defined and incorporated into cancer clinical trials (updated listings of tumor-derived T cell epitopes, 
www.cancerimmunity.org/peptidedatabase/Tcellepitopes.htm
). However, most of these studies concentrated on characterization of a single epitope with a single T cell line or clone (7, 8, 9, 10). Although many reports studied or identified more than one T cell epitope, those T cell responses were often pooled from multiple subjects (11, 12, 13, 14) or from a single subject after repeated in vitro Ag stimulation (15) (which could lead to in vitro T cell priming and an undesirable skewing of the natural responses) (16). Thus, the overall pattern of naturally occurring antitumor responses within given individuals has not been studied in detail before. So, although tumors often express many tumor Ags from various gene families, we know very little about how often and how extensively an individual’s immune system simultaneously responds to multiple tumor Ags.
Immunodominance is a well-described and -studied phenomenon in which cellular immune responses focus only on a few epitopes and can be ordered as a reproducible hierarchy. Although immunodominance holds true for immune responses to single artificial Ags (17), viruses such as influenza (18, 19) and vaccinia (20), and intracellular bacteria that encode >2000 gene products (21), it has not been properly demonstrated for antitumor T cell responses. There are rare examples where multiple epitopes were shown to be generated from a single tumor Ag and restricted to the same HLA allele (9, 12), but these T cells were not assessed under controlled conditions and the immunodominance hierarchy could not be established. Most tumor Ags are "self Ags" and the T cell repertoire directed against them is thought to have been shaped by various tolerance mechanisms. In addition, if tumor Ags arise as a consequence of gene mutation, the entire immune response may only involve a single or very few epitopes, which would not result in a true hierarchy. Finally, previous studies have failed to use a systematic approach to characterize the full breadth of T cell responses associated with tumor Ags and this would have undoubtedly undermined the detection of broad hierarchies.
NY-ESO-1 (hereafter ESO) is one of the best-characterized cancer Ags in terms of its immunogenicity although little is known about its biological function. Patients who develop anti-ESO Abs often have detectable specific CD8+ (10, 22) and CD4+ T cell responses (13). We recently conducted a cancer clinical trial using recombinant full-length ESO protein Ag formulated with ISCOMATRIX (CSL). CD8+ and CD4+ T cell responses for a large number of ESO epitopes were observed (3). However, it is not clear whether the broad T cell responses induced by vaccination are similar to naturally occurring anti-ESO responses. Indeed, there is emerging evidence that the T cell repertoire following vaccination may differ from the naturally induced one (23).
To gain more insights into host antitumor T cell responses and to explore the possible breadth and depth of the T cell responses specific to a given tumor Ag, we comprehensively studied naturally occurring anti-ESO T cell responses in one patient from one sampling point and showed that the CD8+ and CD4+ T cell responses specific to ESO were broad and hierarchical with most of the overall immune responses focused on a few immunodominant epitopes. We further confirmed our observations using PBMCs from a second patient. Our findings have important implications for understanding antitumor immune responses, immunoediting, and rational vaccine development.
|
Materials and Methods |
|---|
|
TopAbstractIntroductionMaterials and MethodsResultsDiscussionDisclosuresReferences |
|---|
Patient A was initially diagnosed with melanoma in 1987, when the patient was then 46 years of age. He received adjuvant postoperative therapy (bacillus Calmette-Guérin injections) for 12 mo and remained disease-free for 6 years. Melanoma relapse in iliac and supraclavicular lymph nodes was diagnosed in late 1993 and the patient underwent lymph node resection followed by chemotherapy. His disease progressed and he underwent radiotherapy combined with chemotherapy toward the end of 1995, which was followed by vitiligo and extensive tumor necrosis detected on computed tomography scan. The subsequent course of the patient was of a slow progression nature with predominantly nodal metastatic disease. The patient died in November 1997 due to septicemia related to chronic infection of his metastatic disease. Blood samples used in this study were collected in March 1996, with written informed consent of the patient. His HLA typing is shown in Table I. Patient B presented at age 46 with a level IV melanoma in January 1996, treated with amputation. Disease first relapsed in July 2001 when she developed an isolated adrenal metastasis. This was fully resected with a right nephroadrenalectomy. The tumor cells expressed Melan A, tyrosinase, and ESO. These studies were approved by the Austin Health Human Research Ethics Committee.
|
All epitope prediction was conducted at http://syfpeithi.bmi-heidelberg.com/ and http://bimas.dcrt.nih.gov/molbio/hla_bind/. ESO 18-mer peptides overlapping by 12 aa were individually synthesized, and 13-mer peptides overlapping by 11 aa were synthesized as cleaved pin peptides, both by Chiron Mimotopes. All other peptides were synthesized and purified (purity >95%) by Auspep. The sequence of ESO96–104 is FATPMEAEL; ESO124–133 is KEFTVSGNIL; and the previously reported ESO92–100 is LAMPFATPM (14). All peptides were dissolved in DMSO at 1 or 10 mM.
Anti-CD4 (PE), anti-CD8 (CyChrome), and anti-IFN-
(FITC) were purchased from BD Biosciences. Pan anti-HLA-DR (L243), anti-HLA-DP (B7/21), and anti-HLA-DQ (SPV-L3) Abs were used as culture supernatants (24). Anti-ESO mAb ES121 has been reported elsewhere (25). The anti-HLA-A, B, C Ab w6/32 was purchased from Sigma-Aldrich. Abs used in the ELISPOT assay were purchased from Mabtech.
Cell culture
EBV-transformed B lymphocyte cell lines (B-LCL; made available from the International HLA Workshop and from the Victorian Transplantation and Immunogenetics Service (VTIS, Melbourne, Australia); courtesy of Associate Prof. B. Tait; Table I), melanoma tumor cell lines, and the Chinese hamster ovary (CHO) cell line transfected with full-length ESO were all cultured in "RP-10" consisting of RPMI 1640 with 10% FCS (CSL), L-glutamine (2 mM), 2-ME (5 x 10–5 M).
Recombinant tumor Ags and ELISA
Detailed methods have been reported elsewhere (26). Briefly, all recombinant tumor Ags used in Fig. 1A were produced in Escherichia coli. One microgram of purified recombinant protein was adsorbed to microwell plates (Nunc) overnight at 4°C. Plates were washed with PBS and blocked with 2% FCS/PBS. Patient serum was diluted in 2% FCS/PBS and added for 2 h. Plates were washed and goat anti-human IgG-AP (Southern Biotechnology Associates) was added. Plates were washed, incubated with Attophose substrate (JBL Scientific) for 25 min, and immediately read (CytoFluor 2350; Millipore).
|
Tumor biopsy specimens were finely minced then placed in a T25 flask in RP-10. Nonadherent cells were discarded the next morning. The melanoma cells normally grew out within 2–4 wk (27). Once established, they were handled as normal cell lines (see Table I).
Generation of ESO-6His protein in CHO cell line
To generate a mammalian version of ESO, the IFN-
signal sequence and a 6His-tag coding cDNA was attached to the 5' and 3' ends of the ESO cDNA, respectively, with a BamHI site on the 5' end and a HindIII site on the 3' end. The PCR-amplified fusion gene, named IEH, was inserted into the pcDNA 3.1 vector. CHO cells were transfected with the IEH plasmid by electroporation (Bio-Rad Gene Pulser II). Stable transfectants were selected with G418 (Invitrogen Life Technologies) and cloned by limiting dilution.
Purification of native ESO from IEH.CHO cells
Trypsinized IEH.CHO cells were sonicated and centrifuged. The clear phase of cell extracts were loaded onto a Ni-NTA Superflow column (Qiagen) and the ESO protein was eluted with Tris-phosphate (pH 8.0) buffer containing 250 mM imidazole.
IFN- treatment of tumor cell lines
The melanoma cell lines were cultured in either RP-10 or RP-10 plus 100 ng/ml recombinant human IFN- (PeproTech) for 48 h before being used as APCs.
Generation of short-term T cell lines
The T cell culture method has been previously reported (28). Briefly, 5 x 106 PBMCs were pulsed with 10 µM ESO 18-mer peptide (or pooled 18-mer peptides as indicated), 2 ml of RP-10 containing IL-2 (10 IU/ml) was then added, and the cells were cultured for the indicated times. When minimum peptides were used as stimulating Ags, 1 µM was used to pulse PBMCs. For the T cells used in Fig. 1B, to minimize the number of cells used, 1 x 106 PBMCs were pulsed with a known HLA-A2 peptide at 1 µM for 60 min; the cells were then washed and five groups were combined in one culture.
Generation of DCs and Ag-specific CD4+ T cells
Autologous monocytes were isolated from PBMCs and cultured in RP-10 containing GM-CSF and IL-4 for 7 days (24). The resulting DCs were then pulsed with ESO immune complexes (IC; 1:1 mix of purified IEH.ESO and purified anti-ESO ES121 at 50 µg/ml) then matured with CD40-L at 1 µg/ml (Amgen) for 2 h and cocultured with autologous PBMCs at a ratio of 1:10 (DC:PBMC) with 10 IU/ml human rIL-2.
Intracellular cytokine staining (ICS) and tetramer staining
These methods have been described previously (28). Briefly, peptides to be tested were directly added into cultured T cells (T cells and other cells served as APC) in the presence of 10% FCS. Brefeldin A (BFA; 10 µg/ml; Sigma-Aldrich) was then added. When assessing T cell avidity, peptides were added to the cultured T cells (3 x 105/well) at various concentrations ± FCS, as indicated. The cells were harvested 4 h later and stained with anti-CD4-PE and anti-CD8-CyChrome, washed, and fixed with 1% paraformaldehyde in PBS and further stained with anti-IFN- in the presence of 0.2% saponin (Sigma-Aldrich).
Tetramers indicated in Figs. 1A and 5 were all produced by our tetramer facility (Ludwig Institute for Cancer Research, Lausanne branch, Switzerland). The cultured T cells were first stained with the indicated tetramer (diluted at 1/200, except ESO124–133 tetramer 1/50) at room temperature for 20 min; anti-CD8+ Ab was then added and stained for a further 30 min. The cells were then washed and 100,000 acquired on a FACSCalibur (BD Biosciences) and analyzed with Flowjo software (Tree Star).
|
ELISPOT assay
A 96-well nitrocellulose plate (MAHA S4510; Millipore) was coated with 5 µg/ml capture Ab at 4°C overnight. The wells were washed and blocked with 5% human AB serum (Valeant Pharmaceuticals). PBMCs and peptides were added and incubated for 18 h. The wells were washed with PBS containing 0.05% Tween 20. Biotinylated detection Ab, streptavidin-conjugated alkaline phosphatase, and its substrate (Sigma-Aldrich) were used to develop the IFN- spots. Spots were counted on an SZ-CTV Olympus microscope.
MHC class I restriction and MHC class II Ab-blocking assays
In class I restriction assays, partial MHC class I-matching B-LCL were pulsed with the indicated peptides then washed extensively before being used as APC. In the MHC class II-blocking assay, target B-LCL were pulsed with 10 µM peptide at 37°C for 1 h. The cells were washed and 20 µl of anti-HLA-class II Ab supernatant was added for another hour before addition of T cells and BFA.
|
Results |
|---|
|
TopAbstractIntroductionMaterials and MethodsResultsDiscussionDisclosuresReferences |
|---|
Patient A had a relatively indolent course of metastatic disease with concurrent vitiligo. We wished to assess whether this was associated with an antitumor immune response. Serological assessment of Abs against several tumor Ags was first performed. Significant titers of Abs against multiple Ags were detected (Fig. 1A). In this patient the most prominent Ab responses were against the CT Ags SSX-2 and ESO.
CD8+ T cell responses to multiple tumor Ags
It has been reported previously that Ab and T cell responses to tumor Ags often occur concurrently (10, 13). In patient A, the development of extensive vitiligo suggested that an immune response had been induced against melanocyte differentiation Ags. Because the patient was HLA-A2 positive, we reasoned that stimulating PBMCs from this patient with available known HLA-A2 tumor antigenic peptides would provide a useful measure of his antitumor CD8+ cellular responses. As shown in Fig. 1B, after 13 days of in vitro stimulation, we were able to detect T cells to multiple HLA-A2-restricted epitopes using specific tetramers. Surprisingly the anti-ESO157–165 response was clearly larger than responses to Melan-A27–36 and gp100457–466. Although signals for tyrosinase1–9 and MAGE-A4230–239 responses were low, the staining patterns were discrete and specific (data not shown), suggesting that these Ags were also targeted. We did not detect T cells specific to MAGE-A10254–262 nor to gp100280–288. It should be noted that the responses to some peptides may have been underestimated using our methods here. This is because the T cells were stimulated with a mixed population and relatively limited number of PBMCs (106 instead of the typical 5 x 106 cells were used to conserve the precious PBMC samples from this patient).
Systematic detection of broad anti-ESO CD8+ T cell responses
In view of the high Ab titer to ESO and relatively large CD8+ T cell response against ESO157–165, we decided to systematically explore the full breadth of potential ESO-specific CD8+ and CD4+ T cell responses. CD14+ monocytes were enriched and monocyte-derived DCs (MoDCs) were generated. The CD14-negative cells were stimulated with pooled, overlapping ESO 18-mer peptides (pools indicated in Fig. 2 legend). Eleven days later, the T cells were screened against all the stimulating peptides as well as an irrelevant control 18-mer peptide. On the day, all the T cell cultures showed negligible background responses to the control 18-mer peptide (<0.1%), so all the signals from each separate culture were plotted on one graph (Fig. 2A). Clearly, the ESO157–165-specific response was relatively large, but notably there were two other responses specific to peptides within ESO91–108 and ESO121–138 that were 3
4-fold larger. There were also very small subdominant responses detected during the 18-mer screen, such as ESO43–60 and ESO61–78. The former was subsequently located within ESO46–58 using overlapping 13-mer peptides (Fig. 2A, inset). The other was located in 60–72 (data not shown). No T cells specific to other ESO peptides were detected in those cultures, suggesting that the smaller specific T cell populations were not the results of T cell competition.
|
10% of the total CD8+ T cells (ESO91–100 and ESO92–101 at high concentrations were as potent as ESO92–100, data not shown). In contrast, ESO87–96 stimulated nearly 20% of the total CD8+ T cells while the peptides shared by ESO96–104 stimulated 4564% of the total CD8+ T cells in the same culture. Assuming that all the T cells with different specificities expanded proportionally in vitro, this response was five times bigger than the T cell response to the previously identified ESO92–100. There was also clearly an epitope contained in ESO82–91, which was missed by the corresponding 18-mer screen (Fig. 2A, ESO79–96). Within that sequence, the 9-mer ESO82–90 GPESRLLEF and 10-mer ESO82–91 GPESRLLEFY are predicted as the best candidate peptides to bind HLA-A1, which patient A expressed. We subsequently showed, using overlapping 9-mer and 10-mer peptides within 82–92, that ESO82–90 is not an epitope and that the minimum epitope is likely ESO82–91 although the nonamer ESO83–91 remains a possibility (Fig. 2B, inset). Due to limited sample availability, we were unable to confirm the restricting HLA. Using similar characterization strategies with 13-mer peptides, we found the core sequence responsible for the other immunodominant epitope to lie within ESO124–136 (KEFTVSGNILTIR; data not shown).
Detailed characterization of the immunodominant anti-ESO CD8+ T cell responses
To confirm that 96–104 was the shared minimum peptide by the three 13 mer shown in Fig. 2B and knowing 95–103 was not recognized by the same T cell line (data not shown), pure peptides 95–104 and 96–105 were synthesized and screened using T cells expanded in vitro with ESO91–108 (18 mer) in the absence of FCS to prevent peptide processing by serum proteases (29) (Fig. 3A). It was clear that ESO96–104 was the minimum peptide for the epitope within ESO91–108 as the T cells were able to recognize this peptide at a subnanomolar concentration (10–10M for 50% maximum recognition) and the recognition decreased dramatically as the peptide sequence was either extended on the N terminus or truncated at the C terminus by a single amino acid.
|
3-fold more potent than ESO125–133, the difference in the presence of FCS was only marginal, possibly due to some ESO124–133 being converted to ESO125–133 by serum proteases (Fig. 3B). Under these conditions, the peptides could be titrated down to the submicromolar range (10–7M for 50% maximum recognition), which is still a substantially higher concentration than that recognized by ESO96–104-specific T cells (Fig. 3A). However, ESO125–134 and ESO123–133 were much less potent, indicating positions 123 and 134 were not part of the minimum epitope. From these data, we concluded that the minimum sequence for this epitope is ESO124–133. Having identified the minimum peptides for both immunodominant T cells, we used these peptides to pulse a set of partially HLA-matched B-LCL lines to define the restricting HLA molecules. The peptide-pulsed B-LCL were then extensively washed before being used as APC for the respective T cell lines, this time raised against either minimum peptide. The results in Fig. 3, C and D, showed that both T cell epitopes were restricted by HLA- Cw*0304.
Both novel immunodominant T cell epitopes are naturally presented by tumor cell line
Once the minimum epitopes and their restricting HLA molecules were identified, we sought to determine whether these epitopes were naturally processed and presented by ESO-expressing melanoma cell lines. The two T cell lines were used to screen five LAR tumor cell lines established in our laboratory (for HLA typing, see Table I). Of these tumor lines, LAR14, LAR31, and LAR39 expressed HLA-Cw3, and LAR14 and LAR31 also expressed ESO. Only LAR31 expressed HLA-Cw*0304. The T cell lines, which each contained 60% Ag-specific CD8+ T cells as judged by peptide-induced activation (
, Fig. 4, A and B), recognized LAR31 slightly differently. The ESO96–104-specific line failed to recognize LAR31 cultured under standard culture conditions but was activated by the LAR31 cells pretreated with IFN-. Importantly, when exogenous ESO96–104 peptide was added to IFN--treated LAR31 cells (Fig. 4A, ), there was no further gain in T cell activation, indicating the T cells activated by either the tumor cells (Fig. 4A, 
) or the synthetic peptide were the same population. Interestingly, although the ESO124–133-specific line failed to respond to low peptide concentrations (Fig. 3B), it recognized the same tumor cell line even without IFN- treatment. Upon IFN- treatment, the percentage of T cell activation was doubled (Fig. 4B) reaching 50% of total Ag-specific T cells revealed by peptide addition. IFN- treatment for 48 h generally doubled surface HLA expression (Fig. 4C). The same T cell lines used in Fig. 4, A and B, successfully lysed the LAR31 tumor line treated with IFN- to an extent similar to that seen when the same cell line was pulsed with the respective peptide (Fig. 4, E and F). The two T cell lines did not cross-react to the other’s cognate peptide (data not shown).
|
Longer peptides require trimming by the serum proteases to generate the suitable minimum peptides for T cell stimulation, which might be a sequence-dependent process (22). To formally establish the immunodominance hierarchy for the three HLA-Cw3-restricted T cell responses, T cell cultures were established separately with the above-characterized minimum peptides and then assessed with specific tetramers. It was clear that ESO96–104-specific T cells were most abundant and ESO92–100-specific T cells were relatively fewer (Fig. 5A). Because of the reasonably large signals detected for the two immunodominant T cells (Fig. 5A, also Figs. 2–4), we reasoned that we might be able to detect these T cells directly ex vivo. PBMCs were either directly stained to gain a CD8+ T cell percentage of total PBMCs (Fig. 5C) or were stimulated in an ELISPOT assay for 18 h again with the minimum peptides. The enumerated ELISPOT data were further converted to Ag-specific T cells per total CD8+ T cells. As shown in Fig. 5B, ESO96–104-specific T cells had an unstimulated frequency in PBMCs of 0.4% total CD8+ T cells and the ESO124–133-specific T cells of 0.2%. ESO157–165-specific T cells had a frequency 0.015%, which is 10- to 20-fold lower than the two immunodominant T cell frequencies. Interestingly, the previously identified HLA-Cw*0303-restricted T cells specific to ESO92–100 were undetectable under identical conditions.
Systematic detection of anti-ESO CD4+ T cell responses
It has been well-demonstrated that B cell and CD8+ T cell priming requires CD4+ T helpers. In some cases these CD4+ T cells recognize epitopes derived from the same protein Ag, which is described as "cognate help" (30). We and others have proposed that an optimal vaccine may require the incorporation of the most immunodominant CD8+ and CD4+ T cell epitopes even if the vaccine contains components which can otherwise provide a potent "danger" signal such as TLR ligands, adjuvants, or cytokines (24, 31). We assessed Ag-specific CD4+ Th responses using a full-length Ag-based strategy, which should only identify T cell responses to naturally presented epitopes. Importantly, this method also made it possible to reveal the immunodominance hierarchy that resulted from processing the full-length Ag by DCs. Immature MoDCs were incubated in the presence of preformed IC using purified ESO protein and the anti-ESO mAb ES121 (25). The ESO protein purified from IEH.CHO was a single band on a SDS-gel and its specificity was confirmed with ES121 in a Western blot analysis (data not shown). The Ag-loaded MoDCs were matured with CD40L and used to stimulate autologous PBMCs in our standard bulk culture. Twelve days later, the bulk T cell line was screened with overlapping 18-mer ESO peptides (Fig. 6A). Two days after the 18-mer screening, the same bulk line was used to screen the overlapping 13-mer ESO peptides (Fig. 6B). Clearly, the most immunodominant CD4+ T cell response was located within ESO157–174. There were also multiple other CD4+ T cell epitopes identified within the 18-mer peptides. All those epitopes were confirmed by the overlapping 13-mer screening and narrowed to a single 13 mer as the core sequence. Interestingly, a few T cell epitopes that were not revealed by the 18-mer peptides were further detected by the 13-mer screening (ESO98–110, ESO102–114, ESO128–140, ESO132–144, ESO146–158). Potentially, the minimum or core peptides contained within those 18 mer were not properly processed by serum proteases within the 4-h assay time. It is very unlikely that intracellular peptide processing took place within the activated T cells, because BFA was added at the same time when peptides were added to these cells.
|
Having identified the 13-mer peptides containing likely minimum epitopes, we used pan class II Abs in a blocking assay to identify the restricting class II molecules. The T cells were first incubated with one of the three Ab supernatants to block the presenting class II molecules on their surface; 13-mer peptides were then added and ICS was performed. It was evident that most CD4+ T cell responses were blocked by the anti-HLA-DP mAb, including the previously identified immunodominant CD4+ T cell epitope ESO157–170 (Fig. 7) (22). The anti-HLA-DQ Ab partially blocked responses to several peptides, a finding which could potentially be due to more then one T cell epitope being contained within a 13-mer sequence (for example, ESO42–54 and ESO52–64). Nonspecific blocking by the anti-HLA-DP Ab could be excluded because the ESO120–132-specific T cells were clearly not affected by the anti-HLA-DP Ab; instead, their activation was blocked by anti-HLA-DR Ab (Fig. 7B). T cells specific to ESO54–66 and ESO118–130 were not blocked by any of the three Abs. This observation could potentially be due to a relatively higher avidity interaction which could overcome the Ab blocking. The T cell responses detected in patient A are summarized in Fig. 8.
|
|
To demonstrate that the immunodominance hierarchy in patient A is not an isolated event, we identified a second subject (patient B) who had four HLA class I alleles in common with patient A (Table I). We screened her CD8+ T cell response in a very similar fashion and a similar T cell response was revealed (Fig. 9A). There is a CD8+ T cell epitope within ESO11–23 (Fig. 9B). There are likely three epitopes within ESO85–108 (Fig. 9C) in which ESO85–94 seemed to be independent and contains the predicted HLA-A3 epitope ESO86–94. The other two epitopes ESO92–100 and ESO 96–104 were confirmed by the corresponding HLA-Cw3 tetramers (Fig. 9E). There are also likely three epitopes within ESO121–144 (Fig. 9D), including ESO124–133 which is also confirmed by tetramer (Fig. 9E). However, the most dominant response from this patient is against ESO92–100 rather than ESO96–104 and ESO124–133 as in patient A. There were other epitopes detected in our screen including a peptide located within ESO43–60, as well as the previously identified ESO157–165 restricted by HLA-A2 (Fig. 9A and data not shown). Collectively, although patient B also responded to ESO extensively, her major CD8+ T cell responses focused on a few epitopes located within ESO85–104. We also demonstrated that the most immunodominant CD4+ T cell response in patient B was toward HLA-DP4/ESO157–170 and she also responded to Melan A27–36 vigorously (data not shown).
|
|
Discussion |
|---|
|
TopAbstractIntroductionMaterials and MethodsResultsDiscussionDisclosuresReferences |
|---|
Although the concept of immunosurveillance against tumors was proposed nearly five decades ago, the strongest experimental evidence came only recently (4) from mouse tumor models. The presence of broad spontaneous T cell responses against a single Ag suggests that ESO was an important tumor Ag in these patients. The relatively indolent clinical course with patient A may have been related to these immune responses, although it is not possible to conclude that a causal relationship existed. We have, however, demonstrated that the T cell responses were potentially clinically relevant because these T cells were capable of recognizing naturally processed Ags, raising the possibility that the antigenic profile of the cancer might have been "edited" by the T cell responses.
We have characterized in detail two novel CD8+ T cell responses specific to ESO96–104 and ESO124–133. Both are presented by HLA-Cw*0304 and are more dominant than the previously defined epitope ESO92–100 restricted by HLA-Cw*0303 (14) in patient A. Interestingly, in patient B, although both HLA-Cw*0303 and Cw*0304 are expressed, the ESO92–100 is clearly more dominant, which further confirms the previously published data (14). There is only a single amino acid difference between HLA-Cw*0304 (91G) and HLA-Cw*0303 (91R). This substitution is located outside of the peptide-binding cleft and therefore predicted not to affect the presentation of these peptides. Whether that difference preferentially biased Ag presentation for the three HLA-Cw3-restricted epitopes and whether the seemingly HLA suballele-dependent immunodominance is truly associated with that single amino acid substitution warrant further investigation.
ESO92–100 is predicted to be the HLA-Cw*0301 epitope with the highest score (http://bimas.dcrt.nih.gov/molbio/hla_bind/). The HLA-Cw*0301 allele was confirmed to have the same sequence as HLA-Cw*0304 and was subsequently deleted from the database, http://www.ebi.ac.uk/imgt/hla/. Our novel epitope ESO96–104 is ranked seventh according to the same prediction and the ESO124–133 is ranked also seventh as a 10-mer candidate. Although the prediction programs have been successfully used to predict many epitopes, generally only 20–50% of predicted epitope peptides possess sufficient binding ability to the HLA molecules; in addition, many of those are not created by the Ag-processing machinery. It has been surmised that only 1 in 2000 potential peptides will eventually achieve immunodominant status (34). In contrast, many peptides are still able to bind to MHC while containing noncanonical sequences (9); this does not take into consideration peptides that may be too long (35), coded by an alternative reading frame (36), carrying posttranslational modifications (37), or joined together through posttranslational splicing (38). In our case, both immunodominant epitopes restricted by HLA-Cw3 are predicted with relatively low scores. Had we used the epitope prediction algorithms and only selected the highest scored peptides for our assays rather than a systematic method, these immunodominant T cells would not have been detected. We suggest that a role still remains for more systematic efforts to reveal antitumor T cells using either full-length Ag-based or overlapping peptide stimulation.
The novel CD8+ T cell epitopes were presented by melanoma cells, especially after IFN- treatment which boosts Ag-processing machinery and enhances HLA expression. It was interesting to note that the ESO96–104-specific T cells have a higher avidity for their cognate peptide than that of the T cells specific to ESO124–133, yet most of the Ag-specific T cells failed to recognize tumor cell line LAR31. Upon IFN- treatment, Ag presentation became saturated and was as good as stimulation induced by pulsing exogenous peptide onto the same tumor cells. The seemingly "black and white" switch pre- and post-IFN- treatment for the ICS assay may indicate some degree of immunoproteasome dependence for that epitope, which would be worth further investigation. In contrast, the recognition of ESO124–133 on the same cells was much less dramatic. Although that T cell did not show very high avidity for its minimum peptide (Fig. 3B), some of the T cells within that T cell line were able to recognize the naturally presented peptide on LAR31 without IFN- treatment (Fig. 4B). Upon IFN- treatment, the total T cell activation, although doubled, only reached half the maximum signal achieved by saturating ESO124–133 peptide. It might be the case that this epitope is very efficiently processed and presented but with a relatively poor HLA-Cw3-binding ability. This was indirectly implied by the fact that ESO124–133-containing HLA-Cw*0304 tetramers stained specific T cells relatively weakly (data not shown) and required higher concentrations (Fig. 5A). In contrast, HLA class I tetramers complexed to either ESO96–104 or ESO92–100 stained the respective T cells strongly (Fig. 5A and data not shown), suggesting higher binding affinity.
The initial 18-mer screening strategy depended on the capacity of serum proteases to trim longer peptides to smaller ones capable of binding to surface HLA on the APC. This trimming has been reported to be sequence dependent (29) and may generate a different peptide repertoire than the intracellular Ag-processing machinery of tumor cells or other APC. Consequently, the initial ESO 18-mer screening could then have potentially created a false impression for the peptides which were dominant in vivo (Fig. 2A). To exclude that possibility, we further confirmed the novel T cell responses for their immunodominant status by the following two means using samples from patient A. First, we used the minimum peptides, including the previously published ESO92–100, as stimulating Ags at saturating concentration and confirmed that ESO96–104 and ESO124–133 were indeed much more immunodominant (Fig. 5A; data not shown). Second, we performed an ex vivo ELISPOT assay using these minimum peptides separately. The T cell responses specific to the two novel epitopes were clearly much greater than that specific to ESO92–100, which fell below the detection threshold (Fig. 5B).
So far, there are nine fully characterized ESO CD8+ T cell epitopes (two coded by alternative reading frame) restricted to various HLA-A, B, or C molecules according to the most recent update of the tumor-specific T cell epitope database (www.cancerimmunity.org/peptidedatabase/tumorspecific.htm). Three of those are within ESO92–102 and are restricted by B*3501, B51, and Cw3, respectively. Now, we have added another epitope, ESO96–104, largely overlapping that region. That region matches one of three most hydrophobic regions (ESO86–100, ESO115–136, and ESO150–173) according to Hopp-Woods hydrophobicity plot (not shown). We speculate that this region not only has the hydrophobic residues required by many HLA-binding motifs as anchor residues but also is very efficiently processed to potentially achieve multiple immunodominant epitopes out of this minimum sequence. It would be very interesting to test T cell responses to these epitopes in individuals expressing all above-mentioned HLA alleles and also to analyze Ag presentation from tumor cell lines expressing these alleles. That should help us to determine first if all the epitopes are generated simultaneously, and second, if so, which epitope is presented most efficiently.
Despite the great effort invested into identifying HLA-A2-restricted ESO-derived T cell epitopes, there is still only one epitope (ESO157–165) (10, 39) shown to be naturally presented by this molecule. In patient groups, either with or without HLA-A2 expression, vaccinated with ESO formulated with ISCOMATRIX adjuvant (3), we found that T cell responses specific to ESO157–165 were often much smaller and less frequent than other detectable responses restricted to non-HLA-A2 molecules. Although it is still likely that this peptide represents the most immunodominant epitope restricted by the HLA-A2 molecule, it is certainly not the most immunodominant T cell epitope as demonstrated in this report for both patients. We believe that there are other more immunodominant T cell epitopes presented by other major HLA alleles in addition to the two epitopes presented by HLA-Cw*0304 (Fig. 2). If that is generally true, then we should move away from the idea of trying to cover the biggest population with a single epitope (restricted by the most popular HLA-A2 molecule) to a more practical one where the aim is to achieve the same or even a larger population coverage with a group of immunodominant epitopes presented by other major HLA alleles. For instance, it might be favorable to use multiple immunodominant T cell epitopes restricted by the most common HLA-A, -B and -Cw alleles as well as the most immunodominant helper epitopes to form the core of a vaccine, which could ultimately provide broad HLA coverage yet still be relatively easy to produce compared with a full-length recombinant protein-based vaccine. Although the latter could bring advantages by stimulating broad T cell responses and automatically covering all presenting HLA alleles, it might also stimulate deleterious tumor-specific regulatory T cells (40). Most importantly, incorporation of extraimmunodominant epitopes may prevent CTL evasion (41).
Patient A also mounted very vigorous responses to multiple CD4+ T cell epitopes derived from ESO and most responses were restricted by HLA-DP4 (Figs. 6 and 7). The most immunodominant T cell epitope in this patient, also in patient B (data not shown), was ESO157–170, which was reported as an immunodominant response simply because it was repetitively detected within a group of patients (22). Both patients expressed HLA-DPB1*0401 and HLA-DPB1*0402. The two alleles differ by four amino acids p36A of *0401 to 36V *0402, p55, 56AA to DE, p178L to M. There is no published HLA-DP4 structure. According to the HLA-DRB1*0101 crystal structure which has very similar sequence to HLA-DPB1*0401 (42), p36 located on the peptide-binding floor and p56 on the 1 
-helical region, both could interact with the carboxyl anchor residue of the bound peptide. However, p56 is located outside the -helical region and p178 is located on the 2 domain; neither is likely to affect peptide binding or TCR recognition. It might be the case that these differences did not affect the overall Ag presentation, which means these patients were functionally homozygous for HLA-DP4; alternatively, each HLA-DP4 molecule in our patient might have presented a subset of HLA-DP4-binding peptides which might have diversified the anti-ESO CD4+ T cell responses (Fig. 6). There have been reported examples showing multiple CD4+ T cell responses specific to various tumor Ags from a single patient (13, 33) and multiple CD4+ T cell responses specific to the same tumor Ag ESO from either a single (24) or multiple individuals (13). It will be very important in future ESO peptide-based vaccines to incorporate the most immunodominant Th and CTL epitopes.
Perhaps the most intriguing feature of the antitumor T cell response from both patients was that it targeted multiple tumor Ags and had a broad and hierarchical response for both CD4+ and CD8+ T cells to ESO, which is very similar to a typical antiviral response. Although a relatively small protein, ESO has been well-demonstrated for its outstanding immunogenicity both in a natural setting (10, 13, 22) and after full-length Ag vaccination (3). ESO is normally well-sequestered in the germ cells which do not express MHC class I molecules (43). Therefore, it is possible that it behaves like a neoantigen and may have all the characteristics of a foreign Ag, such as viral Ags. However, how it elicits such broad and intensive immune responses in the absence of innate danger signals is interesting and warrants more investigation.
|
Acknowledgments |
|---|
|
Disclosures |
|---|
|
TopAbstractIntroductionMaterials and MethodsResultsDiscussionDisclosuresReferences |
|---|
|
Footnotes |
|---|
1 This work was supported by an International Senior Research Fellow Fellowship from the Wellcome Trust (066646/Z/01/Z) (to W.C.). I.D.D. was supported in part by an Australian National Health and Medical Research Council Career Development Award. 
2 Address correspondence and reprint requests to Dr. Weisan Chen, Ludwig Institute for Cancer Research, Austin Health, Studley Road, Heidelberg, Victoria 3084, Australia. E-mail address: weisan.chen{at}ludwig.edu.au
3 Abbreviations used in this paper: DC, dendritic cell; CT, cancer testis; B-LCL, B lymphocyte cell line; CHO, Chinese hamster ovary; ICS, intracellular cytokine staining; IC, immune complex; BFA, brefeldin A; MoDC, monocyte-derived DC.
Received for publication October 14, 2005. Accepted for publication February 24, 2006.
|
References |
|---|
|
TopAbstractIntroductionMaterials and MethodsResultsDiscussionDisclosuresReferences |
|---|
, previously described epitopes; 