The polycomb group proteins, BMI-1 and EZH2, are tumour-associated antigens

We used SEREX technology to identify novel tumour-associated antigens in patients with primary hepatocellular carcinoma and found serological responses to the polycomb group (PcG) protein BMI-1, which is overexpressed in a range of different tumour types. Further studies identified T-cell responses to both BMI-1 and another PcG protein, EZH2, in cancer patients and at relatively lower levels in some normal donors. We next identified several CD8+ T-cell epitopes derived from BMI-1 and EZH2 and demonstrated that EZH2-derived peptides elicited more significant interferon-γ (IFN-γ) release than BMI-1-derived peptides. That CD8+ T cells were responsible for the observed responses was confirmed for EZH2 by both IFN-γ capture assays and tetramer staining using an HLA-A0201-restricted, EZH2-derived YMSCSFLFNL (aa 666–674) epitope. The ability of YMSCSFLFNL (aa 666–674) to stimulate the in vitro expansion of specific T cells from peripheral blood lymphocytes was greatly enhanced when the CD25+ T-cell population was depleted. EZH2-specific cytotoxic T lymphocyte clones specific for two HLA-A0201 epitopes were generated and found to recognise endogenously processed EZH2 in both HLA-matched fibroblasts and tumour cell lines. Given the widespread overexpression of PcG proteins in cancer and their critical role in oncogenesis, these data suggest that they may be useful targets for cancer immunotherapy.

Immunotherapeutic approaches to the treatment or prevention of cancer have flourished in recent years (Rosenberg, 2001;Rosenberg et al, 2004). This is largely due to the rapid identification of tumour-associated antigens (TAAs), which has been achieved by examining the specificity of existing T-cell responses (Stevanovic, 2002), or by using screening strategies such as serological identification of antigens by recombinant expression cloning (SEREX) (Pfreundschuh, 2000), proteomic analysis, gene expression profiling or combinations (Schultze and Vonderheide, 2001). The first tumour-specific vaccination was carried out in malignant melanoma which remains a valuable model for investigating cancer immunotherapy in humans (Saleh et al, 2005). With the growing number of TAAs identified, clinical studies have been expanded to other carcinomas (Stevanovic, 2002) and most of these studies used MHC class I-restricted, T-cell epitope peptides derived from TAAs to expand cytotoxic T cells (CTLs) for tumour destruction. Although this approach has the potential to induce antitumour immune responses, few studies have shown significant clinical responses (Rosenberg et al, 2004). Possible reasons for this are that the peptides chosen may not be naturally processed and presented by tumour cells, immune escape and unsolved questions about optimal clinical application. Recent studies implicate CD4 þ CD25 þ regulatory T cells (T REG ) within the tumour microenvironment in suppression of CD4 þ and CD8 þ T-cell activation and proliferation (Shevach et al, 2001). Increased numbers of both peripherally circulating and tumour-associated T REG have been observed in tumourbearing individuals with non-small lung, ovarian, breast, ovarian and hepatocellular carcinoma (HCC) (Woo et al, 2001(Woo et al, , 2002Liyanage et al, 2002;Ormandy et al, 2005;Unitt et al, 2005), and increased numbers of infiltrating tumour T REG has been linked to a striking reduction in patient survival in ovarian cancer (Curiel et al, 2004).
With the exception of melanoma, the number of characterised TAAs for most tumours is limited and there is an urgent need to identify more TAAs and T-cell epitopes derived from their sequences. To be effective vaccines, T-cell epitopes must be naturally processed and presented antigenic targets capable of eliciting T-cell responses. The ultimate goal is to provide multiple tumour-associated epitopes presented by both MHC class I and class II molecules.
Here, we describe the identification by SEREX screening of a new TAA, the polycomb group (PcG) protein BMI-1, in HCC. BMI-1 and a related protein EZH2 are constituents of two distinct classes of PcG proteins. They are important transcriptional repressors that regulate gene activity by the formation of multiprotein complexes, called PcG bodies, and the organisation of chromatin into an inaccessible structure that cannot bind transcription factors (Shao et al, 1999). Interestingly, EZH2 appears to be one of the PcG proteins essential for BMI-1 recruitment to the PcG bodies (Hernandez-Munoz et al, 2005). PcG gene expression is tightly regulated and differs between different tissues and cell types. The proteins ensure correct embryonic development and are also major contributors to the regulation of haematopoiesis in the adult organism (Jacobs and van Lohuizen, 2002). BMI-1 has been shown to be indispensable for the selfrenewal of neural (Molofsky et al, 2003) and haematopoietic stem cells (Raaphorst, 2003), and it also has a key role in regulating the proliferative activity of normal stem and progenitor cells (Lessard and Sauvageau, 2003). EZH2 is essential for cellular proliferation (Bracken et al, 2003). There is a clear connection between altered PcG gene expression and oncogenesis. Several mammalian PcG members have been implicated in the control of cellular proliferation and tumorigenesis (Jacobs and van Lohuizen, 2002), and altered expression of PcG genes is linked to transformation in cell lines and the induction of tumours in mutant mice (Satijn et al, 1997;Satijn and Otte, 1999). Several recent studies suggest that they also contribute to the development of cancer in humans. Aberrant PcG protein expression, such as overexpression, failure of downregulation or coexpression, has been reported in Hodgkin's lymphoma , B-cell lymphoma , mantle cell lymphoma (Bea et al, 2001;Visser et al, 2001), prostate cancer (Sellers and Loda, 2002;Varambally et al, 2002), breast carcinoma (Kleer et al, 2003;Raaphorst et al, 2003;Kim et al, 2004), HCC (Shi et al, 2005) and during lung carcinogenesis (Vonlanthen et al, 2001;Breuer et al, 2004). Some of these altered patterns of expression may be of diagnostic and prognostic relevance (Varambally et al, 2002;Raaphorst et al, 2003;Breuer et al, 2004;Kim et al, 2004). Our study shows that BMI-1 and EZH2 are overexpressed in several epithelial tumours and that both proteins are associated with both humoral and T-cell responses, suggesting that they represent a novel family of TAAs that might be potential targets for immunotherapy.

Patients and subjects
Serum and/or peripheral blood mononuclear cells (PBMCs) were separated from heparinised blood samples (25 -30 ml) taken from patients with a variety of epithelial tumours and from healthy donors. Histocompatibility locus antigen (HLA) class I tissue typing was carried out at the National Blood Service, Birmingham, UK. Tumour tissue was obtained from patients undergoing diagnostic biopsy or surgical resection. Normal tissue was taken from areas of tissue showing no tumour involvement. Approval was obtained from the South Birmingham Research Ethics Committee. Informed consent was obtained from all subjects.
Generation of BMI-1 and EZH2 constructs and recombinant adenoviruses BMI-1 or EZH2 DNA sequences (a kind gift from Professor Thomas Jenuwein) were ligated into the pAdTrack-CMV vector containing the sequence for green fluorescent protein (GFP) and used directly in the nucleofection experiments. The empty vector encoding GFP was used as the negative control. Recombinant adenoviruses were produced by recombining the EZH2-containing pAdTrack-CMV vectors or the empty pAdTrack-CMV with the pAdEasy-1 vector and purified on caesium chloride gradients.

SEREX screening of human sera against a HCC cDNA library
All serum samples were preabsorbed before SEREX screening. The primary screen was carried out using a commercially available HCC cDNA library (Novagen, Darmstadt, Germany). Serum from patients with HCC was tested initially and any positive clones screened against a panel of human serum from non-cancer donors. Samples were tested at a dilution of 1 : 1000 and positive plaques detected by alkaline phosphatase-conjugated goat anti-human IgG antibody (Sigma, Gillingham, Dorset, UK; 1 : 1000) and staining with 5-bromo-4-chloro-3-indolyl phosphate and nitroblue tetrazolium (BioRad, Richmond, USA). Positive plaques were rescreened to eliminate those encoding human immunoglobulin sequences. Phage DNA from positive clones was converted to stable plasmid subclones in BM25.8 cells and then transformed into JM109 cells before sequencing.
CD4 þ and CD8 þ T-cell depletions were carried out using antibody-coated magnetic beads (Dynal, Paisley, UK) according to the manufacturer's instructions and depletion confirmed by flow cytometry. Tumour-infiltrating lymphocytes (TILs) were extracted from liver tumour tissue by homogenisation in RPMI/10% FCS followed by density gradient centrifugation through Lymphoprep.

ELISPOT assay of interferon-c release
Interferon-g (IFN-g) release was measured using a commercially available enzyme-linked immunospot (ELISPOT) kit (ELISPOT assay for human interferon-g; Mabtech, Sweden) using cytokine capture and detection reagents according to the manufacturer's instructions. Peripheral blood mononuclear cells, CD4-depleted or CD8-depleted responder cells were seeded in triplicate wells and peptides added at 10 mg ml À1 . Negative control wells contained irrelevant peptide or diluted DMSO and positive control wells contained 10 mg ml À1 PHA (Sigma) or an HLA-matched peptide epitope derived from Epstein -Barr virus (EBV). Spots were counted using an automated system.

Detection and enrichment of EZH2-specific CD8 þ T cells based on IFN-c secretion
Detection and enrichment of EZH2-specific T cells was carried out using the Miltenyi Cytokine Secretion Assay PE kit for IFN-g (Miltenyi Biotec, Surrey, UK) according to the manufacturer's instructions. Peripheral blood mononuclear cells isolated from HLA-A0201 donors were stimulated overnight at 371C with the HLA-A0201-restricted, EZH2 peptide YMCSFLFNL (aa 666 -674) at 10 mg ml À1 . Negative control cultures contained diluted DMSO and positive controls staphylococcal enterotoxin B (Sigma) at 1 mg ml À1 . Interferon-g-secreting cells were selected on automated columns. Pre-and postselected samples were analysed by flow cytometry using anti-human CD8-PC5 (Beckman Coulter, High Wycombe, Bucks, UK) with propidium iodide to gate out dead cells.

Isolation of EZH2-specific cytotoxic T-cell clones
Peripheral blood mononuclear cells were stimulated with peptide (1 -100 mM) for 1 -2 h at 371C, washed, and resuspended in T-cell medium (RPMI 1640 containing 10% FCS, 2 mM glutamine and penicillin/streptomycin), supplemented with 25 ng ml À1 rIL-7 (Peprotech) and 10 U ml À1 rIL-2 after 3 days. Cultures were restimulated after 1 week using irradiated peptide-loaded autologous PBMCs (4000 Rads), and cloned by limiting dilution to 0.3, 3 and 30 cells well À1 on day 14. Each well also contained 1 Â 10 4 irradiated peptide-loaded autologous lymphoblastoid cells (LCL) generated by in vitro transformation of B cells using the EBV isolate B95.8 (Miller and Lipman, 1973), and 1 Â 10 5 irradiated PHA-stimulated mixed allogeneic buffy coat feeder cells (NBS, Birmingham). Cloning medium comprised T-cell medium containing 1% human AB serum (Gibco-Invitrogen, Paisley, UK), 100 U ml À1 rIL-2 and 30% supernatant from the MLA 144 cell line. Positive clones (initially screened against peptide-loaded autologous LCL target cells in an ELISA of IFN-g release) were transferred to separate 2 ml wells in 1 ml of T-cell medium with 1 Â 10 5 irradiated peptide-loaded autologous LCL and 1 Â 10 6 irradiated PHA-stimulated mixed allogeneic buffy coat feeder cells. Wells were maintained by feeding twice weekly and were restimulated once a month. Cytotoxicity was measured using chromium release assays as previously described (Lee et al, 1997). Target cells labelled with [ 51 Cr]O 4 were used at 2.5 Â 10 3 cells well À1 with T cells added at known effector : target ratios. Histocompatibility locus antigen-A0201-positive primary human fibroblasts (infected with recombinant adenoviruses RAdEZH2 and RAdControl at a multiplicity of infection of 30 for 48 h before labelling) and tumour cell lines (LNCaP, HepG2 and MCF7) were employed as target cells in these assays. In some experiments, targets were cultured with IFN-g (1000 U ml À1 ) for 48 h before assay and/or incubated with the antihuman HLA-A0201 monoclonal antibody BB7.2 (1 mg well À1 ) for 1 h at 371C after labelling and before the addition of T cells.

Identification of the PcG protein BMI-1 as a TAA in HCC using SEREX technology
Screening of an HCC cDNA library with serum from patients with HCC (not associated with viral hepatitis) using SEREX technology revealed an antigen which when sequenced was shown to be a member of the PcG family called BMI-1. Serum from nine HCC patients studied reacted strongly with BMI-1 (Table 2; Supplementary Material), suggesting that immune responses to this antigen are frequent in HCC. The SEREX screening identified three sequences representing different cDNA clones of BMI-1, and all nine HCC serum samples screened had antibody responses to at least one of these. The majority of the noncancer serum samples also showed reactivity to BMI-1, but this was generally weaker as determined by intensity of plaque staining.
Expression of BMI-1 and EZH2 in normal and tumour tissue Before embarking on a more detailed study of cellular immune responses to BMI-1, its expression was examined in a panel of tumour tissue samples and cell lines. All samples, both tissue sections and cell lines, displayed upregulation of the BMI-1 protein (Table 3; Supplementary Material). There was strong, specific, nuclear staining for BMI-1 involving at least 50% of cells in all of the cases studied and in some instances nearly 100% of the tumour cell population were positive ( Figure 1A). There was minimal immunohistochemical staining for BMI-1 (o1% of cells) in normal tissues from breast, stomach, colon or skin (Table 2; Supplementary Material). Three of the four normal liver samples showed slightly more extensive nuclear staining involving o10% of hepatocytes. Occasional infiltrating lymphocytes were also positive in normal tissue specimens. There is currently no commercially available antibody for EZH2, so we used RT -PCR to analyse EZH2 mRNA expression in tumour cell lines. High-level expression of EZH2 mRNA was observed in the prostate cancer cell line LNCaP, the HepG2 hepatoblastoma cell line and three breast cancer cell lines MCF7, BT549 and nDA-MB436. This confirms previous studies demonstrating upregulation of EZH2 expression in tumours of patients with breast cancer (Kleer et al, 2003;Raaphorst et al, 2003;Kim et al, 2004). Variable expression was seen in a number of EBV-transformed LCL, with low expression of EZH2 mRNA in the cervical carcinoma cell line Caski, and in normal fibroblasts ( Figure 1B).

Systemic T-cell responses to BMI-1 and EZH2 can be detected in patients with solid tumours
We tested for immune responses to BMI-1 and EZH2 by measuring IFN-g production from PBMC cultures nucleofected with plasmid DNA (pAdTrack-CMV) encoding the EZH2 and BMI-1 proteins. Background levels of IFN-g production following nucleofection were assessed using the empty pAdTrack-CMV vector, which encodes GFP alone, and were generally low. The presence of GFP in the pAdTrack-CMV vectors expressing BMI-1 and EZH2 also allowed an assessment of the nucleofection efficiency, which was generally between 1 and 2%.
Responses to BMI-1 were seen in one out of 10 HCC patients, two out of six prostate cancer patients and three out of five colorectal cancer patients. EZH2 reactivity was observed in three out of ten HCC patients, one out of five colorectal cancer patients and in none of the prostate cancer samples. One sample from both the HCC group and the colorectal cancer group demonstrated T-cell reactivity to both antigens. No responses were seen to either antigen in the three breast cancer patients investigated. Although the numbers investigated in each group were small, there appeared to be a dominant response to BMI-1 in the patients with colorectal and prostate cancer, whereas stronger reactivity to EZH2 was more common among the HCC patients. Typical responses from each group are shown in Figure 2A where it can be seen that levels of IFN-g production were reasonably high. Figure 2B shows the results of a paired PBMCs and TILs population from a patient with HCC. No T-cell reactivity was seen in the nucleofected PBMC cultures to either BMI-1 or EZH2, but interestingly IFN-g production was extremely high in the TILs culture expressing BMI-1, suggesting that immune responses may be compartmentalized in the tumour. These data demonstrate that T-cell responses to both BMI-1 and EZH2 are evident in cancer patients.
To confirm the T-cell responses in patients with HCC, we used CD8 þ T-cell peptide epitopes derived from both the BMI-1 and EZH2 proteins predicted to bind to HLA-A0201, HLA-B2702/05 and HLA-B4402/03 molecules (Table 1) in ELISPOT assays of IFN-g release. Eleven patients of the appropriate HLA type were investigated and typical results are shown in Figure 3A  Results from paired cultures of TILs and PBMCs from a patient with hepatocellular carcinoma are also shown. Interferon-g production was measured in the supernatants of serially diluted PBMC cultures using an ELISA. The results shown are means of triplicate wells and standard errors were always less than 10%.   Figure 3 ELISPOT responses in patients with hepatocellular carcinoma (HCC) and normal donors to predicted CD8 þ T-cell epitope peptides. Unfractionated PBMCs or CD4-and CD8-depleted responder cell populations were screened against 9-mer peptides derived from BMI-1 and EZH2 proteins predicted to bind to the HLA class I molecules HLA-A0201, HLA-B2702/05 and HLA-B4403 (Table 1)  Responses to BMI-1, all directed against HLA-A201-binding peptides, were detected in three individuals with one showing reactivity to the CLPSPSTPV (aa 271-279), one to TLQDIVYKL (aa 74-82) and one to both peptides (Patient 001 shown in Figure 3A). Two out of the 11 gave a positive response to the HLA-A201binding epitopes YMCSFLFNL (aa 666-674) and SQADALKYV (aa 734-742) derived from EZH2. To confirm that these were CD8 þ T-cell responses, we compared unfractionated PBMCs with CD4-or CD8-depleted responder cell populations (FACS analysis confirmed depletion of greater than 99% of the appropriate T-cell subset) (Patient 002 shown in Figure 3B). The magnitude of the responses to the EZH2 peptides was generally greater than those to BMI-1.
Weaker T-cell responses to BMI-1 and EZH2 are also seen in normal donors We also investigated T-cell reactivity to BMI-1 and EZH2 in normal laboratory donors by measuring IFN-g production from PBMC cultures expressing BMI-1 and EZH2 proteins following nucleofection (as above). Background levels of IFN-g production to the control GFP-expressing plasmid were again low and nucleofection efficiency was consistently 1 -2%. Two out of the eight normal samples investigated demonstrated reactivity to BMI-1, and four to EZH2. Interferon-g production was generally 10 -20 times lower than with samples from patients with cancer ( Figure 2C). In the ELISPOT assay, two out of 10 normal donors showed a response to BMI-1-predicted peptides. One individual demonstrated reactivity to the HLA-A201-binding peptide TLQDIVYKL (aa 74 -82), and the other to the HLA-B2702/05-binding peptides VRYLETSKY (aa 44 -52) and KRYLRCPAA (aa 161 -169; Donor 001 shown in Figure 3C). EZH2 reactivity was observed in two out of seven donors. Both showed a positive response to the HLA-A201-binding epitopes YMCSFLFNL (aa 666 -674) and SQADALK-YV (aa 734 -742), which depletion experiments confirmed to be CD8 þ T-cell reactivity (Donor 001 shown in Figure 3D), and one of these individuals also showed weaker reactivity against two  putative HLA-B4403-restricted peptides EELFVDYRY (aa 725 -733) and . The magnitude of the responses obtained from normal donors for both BMI-1 and EZH2 was less than that for HCC patients. A cytokine secretion assay based on IFN-g was employed for the detection and enrichment of peripheral blood T cells from normal donors that responded to the EZH2-predicted epitope YMCSFLFNL (aa 666 -674). Eight HLA-A0201-positive laboratory donors were tested and significant enrichment of EZH2-specific CD8 þ T cells was seen in five of these ( Figure 4). We then used MHC class I tetramers incorporating the HLA-A0201-restricted EZH2 epitope YMCSFLFNL (aa 666 -674) to further demonstrate systemic T-cell reactivity to EZH2. Peripheral blood mononuclear cells isolated from seven HLA-A0201-positive healthy laboratory donors were stained directly ex vivo with the tetramer, and between 0.05 and 0.75% (median 0.15%) CD8 þ T cells stained. In four out of six of the normal PBMC samples, the percentage of tetramer-positive CD8 þ T cells could be increased by carrying out a 14-day in vitro incubation with the YMCSFLFNL (aa 666 -674) peptide (range 0.03 -1.04%; median 0.585%) ( Figure 5A and B).

Depletion of CD25 þ T cells allows greater in vitro expansion of EZH2-specific T cells
We hypothesized that EZH2-specific T cells might be suppressed in vivo by regulatory T cells. In order to test this, CD25 þ T-cell depletion was carried out on six HLA-A0201-positive normal PBMC samples (99% of CD4 þ CD25 þ T cells were depleted) and subsequent in vitro restimulation with the YMCSFLFNL (aa 666 -674) peptide carried out for 14 days. In two out of six of the samples, a marked expansion of YMCSFLFNL (aa 666 -674)tetramer positive cells was detected. The data for one of these are shown in Figure 5, where it can be seen that 0.67% of cells within the CD8 þ T-cell fraction were tetramer positive without CD25 depletion as opposed to 2.04% with CD25 depletion.
EZH2-specific T-cell clones are cytotoxic against endogenously processed antigen and tumour cell lines EZH2-specific CTL clones specific for the HLA-A0201-restricted peptide epitopes YMCSFLFNL (aa 666 -674) and SQADALKYV (aa 734 -742) were isolated from both HLA-A0201-positive HCC patients and normal donors by peptide restimulation in vitro. Antigen-specific cytotoxicity was observed against HLA-A0201matched fibroblasts infected with a recombinant adenovirus encoding EZH2 (RAdEZH2) ( Figure 6A) and the breast carcinoma cell line MCF7, which naturally expresses EZH2 ( Figure 6B).
Greater levels of lysis were observed following incubation of the tumour cells with IFN-g and T-cell reactivity was either greatly reduced or effectively blocked following incubation of the target cells with the anti-HLA-A0201 monoclonal antibody BB7.2. Low levels of killing were observed against the HLA-A0201-positive LNCaP prostate cancer cell line and HepG2 hepatoblastoma cells (data not shown).

DISCUSSION
We employed SEREX technology using a cDNA library to identify new TAAs associated with liver cancer. The SEREX approach was chosen because the method identifies antigens that are of immunological relevance, as their detection requires patients to have generated antibodies against them. SEREX has previously detected tumour antigens with epitopes for cellular immune responses, the best-known example being NY-ESO-1 (Gnjatic et al, 2002). In our study, screening of an HCC cDNA library with serum from patients with nonvirally induced HCC revealed different antigens, with a particularly strong response from all patient samples to the PcG protein BMI-1 (Table 2; Supplementary Material). This strong reactivity, together with the fact that the PcG proteins are known to be vitally involved in transcriptional control and carcinogenesis in humans, and may therefore be less susceptible to the development of 'immune escape' variants, made BMI-1 an attractive candidate for validation as a target for cancer immunotherapy. Recent studies showing that another polycomb protein, EZH2, which is essential for BMI-1 recruitment to the polycomb body in the nucleus (Hernandez-Munoz et al, 2005), is also highly expressed in many cancers and led us to extend our studies to include both BMI-1 and EZH2. As expression of these two proteins is often mutually exclusive, a sensible approach may be an attempt to target both.
Consistent with recent studies demonstrating overexpression of PcG genes in cancer (Vonlanthen et al, 2001;Sellers and Loda, 2002;Kleer et al, 2003;Raaphorst et al, 2003;Breuer et al, 2004;Kim et al, 2004;Shi et al, 2005), we found that both BMI-1 and EZH2 are strongly expressed by a variety of tumours (Table 3; Supplementary Material; Figure 1). This broad upregulation in cancer makes them attractive immunotherapeutic targets for malignant disease. Low level expression of both proteins was also observed in some normal tissues, which is consistent with the reported role of BMI-1 and EZH2 in self-renewal and proliferative activity of stem cells and their progenitors (Bracken et al, 2003;Molofsky et al, 2003). The low-level restricted expression on normal tissues suggests that effective immune responses against  BMI-1 and EZH2 may be compromised by self-tolerance and that stimulating immune reactivity to these proteins may be accompanied by autoimmune side effects. However, the majority of cancer-specific immunotherapies have employed self-antigens and autoimmunity has not proven to be a significant problem in most clinical studies (Gilboa, 2004). Because CTL responses are important for tumour destruction, the demonstration of CTL responses against PcG proteins would strengthen their potential as tumour antigens. Initially, we looked for IFN-g secretion from PBMCs that had been nucleofected with plasmids encoding either the BMI-1 or EZH2 proteins. A system optimised for the transfection of dendritic cells was employed so that monocytes and macrophages would be preferentially nucleofected. Fluorescence-activated cell sorter (FACS) analysis revealed that T cells and B cells were also expressing the proteins and that transfection efficiencies were around 1 -2%. Earlier experiments using EBV antigens had established that these conditions provided effective antigen processing and presentation and reliable antigenspecific IFN-g responses using low numbers of cells (Figure 7). Patients with HCC, prostate cancer, colorectal cancer and breast cancer were studied, and with the exception of breast cancer, T-cell reactivity to BMI-1 and/or EZH2 was observed in all groups. Predicted CD8 þ T-cell epitopes derived from BMI-1 and EZH2 also elicited T-cell responses as assessed by IFN-g release confirming the presence of CD8 responses against these proteins in patients with cancer. These data also confirmed that these epitopes identified by predictive algorithms are likely to represent naturally presented epitopes and may therefore be significant for the design of peptide-based vaccines.
The SEREX screening demonstrated reactivity to cDNA clones representing BMI-1 in serum obtained from most of the noncancer patients, and T-cell assays also detected BMI-1-and EZH2-specific T-cell responses in normal individuals using either nucleofection of PBMCs or ELISPOT assays with T-cell epitopes. Enrichment of IFN-g secreting CD8 þ T cells specific for YMCSFLFNL (aa 666 -674) was possible in five out of eight normal donors ( Figure 4B) and systemic T-cell responses to EZH2 were demonstrated in HLA-A0201-positive normal donors by YMCSFLFNL (aa 666 -674)-tetramer staining. Significantly, with the majority of these samples, the percentage of CD8 þ T cells, which stained positive, could be increased by carrying out an in vitro restimulation step with the YMCSFLFNL (aa 666 -674) peptide. The fact that these EZH2-specific T cells could be expanded in vitro is encouraging because it opens up the possibility of either adoptive T-cell therapy or expansion in vivo by vaccination.
It is perhaps not surprising that responses to BMI-1 and EZH2 were detected in normal donors, as there are likely to be autoreactive T cells of low avidity present in the mature T-cell population for the majority of self-antigens. This has been demonstrated for other TAAs (Marincola et al, 1996). In fact, it is the vulnerability of the immune system to activate anti-self Effector:target ratio 10:1 Interferon-g production by a PBMC culture from a normal donor following nucleofection with plasmid DNA expressing the E3C protein from EBV protein. Peripheral blood mononuclear cells were nucleofected with 5 mg of plasmid DNA encoding E3C or the empty vector and IFN-g production was measured after 48 h incubation.
immunity that is the basis for developing useful cancer-specific immunotherapies and it is of note that, although the frequency of detected responses to PcG proteins was similar between normal and cancer patients, the normal donors consistently showed a weaker immune response. It is intriguing to speculate that immune responses to PcG proteins could be involved in the control of normal haematopoeisis and/or regulating stem cell renewal.
The study revealed differences in responses between BMI-1 and EZH2 with both cancer patients and normal donors. Interferon-g production following nucleofection was more readily detected in response to BMI-1 in the patients with colorectal and prostate cancer, whereas the HCC patients tended to have stronger responses to EZH2. This dominant response to EZH2 in the HCC group was also apparent with the ELISPOT assays, where the magnitude of the responses to the EZH2 peptides was generally greater than those to BMI-1. EZH2 T-cell reactivity was more common than BMI-1 reactivity among normal donors also, suggesting that BMI-1 may be less immunogenic than EZH2. BMI-1 is more widely expressed on stem cell populations (Lessard and Sauvageau, 2003;Raaphorst, 2003) and so represents an important regulatory antigen associated with a more 'primitive' cell type. It could be that the immune system is, therefore, more tolerant to BMI-1 than EZH2. The strong response to BMI-1 among some of the patients with colorectal and prostate cancer may be due to the pattern of aberrant expression of the polycomb proteins by these tumours. Overexpression of BMI-1 may result in a break in tolerance and reactivation of autoreactive T cells by exposure to high concentrations of antigen.
EZH2-specific T-cell clones were obtained by in vitro peptide restimulation using the HLA-A0201-restricted CD8 þ T-cell epitope YMCSFLFNL (aa 666 -674). These clones were able to recognize and kill target cells overexpressing EZH2 following infection with a recombinant adenovirus, and also the HLA-A0201matched breast tumour cell line MCF7. Blocking experiments with the BB7.2 monoclonal antibody against HLA-A0201 confirmed that this recognition and cytotoxicity was MHC class I-restricted. It was encouraging that T cells cloned in response to an EZH2-derived peptide could recognize endogenously processed and presented antigen, suggesting that EZH2-specific T cells would be able to recognize and kill tumour cells. Functional EZH2-specific T cells have previously been isolated from patients with prostate cancer using HLA-A2402-restricted peptides. These effectors were shown to be cytotoxic against the C1R-A24 subline of C1R lymphoma and also the prostate cancer cell line LNCaP, which had been stably transfected with the HLA-A2402 gene (Ogata et al, 2004).
It has been suggested that SEREX identifies self-antigens that elicit naturally occurring CD4 þ CD25 þ regulatory T cells (T REG ) which maintain and regulate immunological homeostasis (Nishikawa et al, 2005) and our study supports this hypothesis. In experiments using the EZH2-derived HLA-A0201-restricted peptide YMCSFLFNL (aa 666-674), we were able to show that depletion of CD25 þ T cells promotes the expansion of YMCSFLFNL (aa 666 -674)-tetramer positive cells in normal PBMC cultures undergoing peptide restimulation in vitro. This suggests that T REG do play a role in suppressing at least CD8 þ T-cell reactivity to EZH2. In tumours, where numbers of T REG may be greatly increased, this immunosuppressive effect may be highly significant.
We have demonstrated that the PcG proteins BMI-1 and EZH2 represent promising target antigens for cancer immunotherapy. They are expressed on a wide variety of tumours where they appear essential to the oncogenic process and show restricted expression in normal tissue. They are naturally immunogenic and EZH2-specific T cells can be reactivated and expanded in vitro by a combination of antigen stimulation and T REG depletion. Most importantly, these T cells can recognize endogenously processed antigen and are cytotoxic against tumour cell lines naturally expressing EZH2, suggesting that PcG proteins are a novel group of TAA with potential as antigens for immunotherapy in a wide range of cancers.