T cells reactive with HLA-A*0201 peptides from the histone demethylase JARID1B are found in the circulation of breast cancer patients.

The nuclear protein PLU-1/JARID1B/KDM5 is widely expressed in breast cancers while showing highly restricted expression in normal adult tissues. To investigate whether JARID1B is a potential target antigen for immunotherapy of breast cancer, we have analyzed the responses of CD8(+) T cells to JARID1B HLA-A*0201 peptides in vitro and used peptide multimers to detect the presence of JARID1B reactive T cells in the circulation of breast cancer patients. Peptides were selected using two web-based algorithms: criteria for inclusion being a high score in both prediction algorithms, and nonhomology with retinoblastoma binding protein-2 (RBP2/JARID1A/KDM5A). A 65-peptide panel was selected and assayed for binding strength by competition assay to obtain the IC(50). The immunogenicity in vitro of these peptides was assessed by T cell stimulation experiments, using autologous dendritic cells as APCs in the first rounds followed by autologous lymphoblasts. Fourteen of the peptides assayed produced cultures having >2% of the CD8(+) cells being IFN-γ(+) after 3-6 rounds of stimulation. An HLA-A*0201 cell line could activate the specific T cells if pulsed with peptide, but endogenous peptide levels were insufficient for activation. Nevertheless, multimer staining of circulating T cells from breast cancer patients showed a significantly higher percentage of multimer positive CD8(+) T cells, as compared to healthy adults for two of three JARID1B epitopes tested. One of these, peptide 73 (QLYALPCVL), was analyzed for memory phenotype, and found to have a significantly higher proportion of central memory T cells than the control group, demonstrating a previous exposure to the peptide.

Mammosphere culture of metastatic breast cancer cells enriches for tumorigenic breast cancer cells.

INTRODUCTION: The identification of potential breast cancer stem cells is of importance as the characteristics of stem cells suggest that they are resistant to conventional forms of therapy. Several techniques have been proposed to isolate or enrich for tumorigenic breast cancer stem cells, including (a) culture of cells in non-adherent non-differentiating conditions to form mammospheres and (b) sorting of the cells by their surface phenotype (expression of CD24 and CD44). METHODS: We have cultured metastatic cells found in pleural effusions from breast cancer patients in non-adherent conditions without serum to form mammospheres. Dissociated cells from these mammospheres were used to determine the tumorigenicity of these cultures. Expression of CD24 and CD44 on uncultured cells and mammospheres derived from the pleural effusions was documented. RESULTS: We found that the majority (20/27) of the pleural effusions tested contained cells capable of forming mammospheres of varying sizes that could be passaged. After dissociation and plating with serum onto adherent dishes, the cells can differentiate, as determined by the increased expression of cytokeratins and MUC1. Analysis of surface expression of CD24 and CD44 on uncultured cells from 21 of the samples showed that the cells from some samples separated into two populations, but some did not. The proportion of cells that could be considered CD44+/CD24low/- was highly variable and did not appear to correlate with the ability to form the larger mammospheres. Of eight pleural effusion mammospheres tested in severe combined immunodeficiency disease (SCID) mice, four were found to induce tumours when only 5,000 or fewer cells were injected, whereas the same number of uncultured cells did not form tumours. The ability to induce tumours appeared to correlate with the ability to produce the larger mammospheres. Uncultured cells from a highly tumorigenic sample (PE14) were uniformly negative for surface expression of both CD24 and CD44. CONCLUSION: This paper shows, for the first time, that mammosphere culture of pleural effusions enriches for cells capable of inducing tumours in SCID mice. The data suggest that mammosphere culture of these metastatic cells could provide a highly appropriate model for studying the sensitivity of the tumorigenic 'stem' cells to therapeutic agents and for further characterisation of the tumour-inducing subpopulation of breast cancer cells.

Sialylated core 1 based O-linked glycans enhance the growth rate of mammary carcinoma cells in MUC1 transgenic mice.

The MUC1 mucin, found on the luminal surface of simple epithelial cells is upregulated and aberrantly glycosylated in many carcinomas particularly breast and ovarian. MUC1 expressed by normal mammary epithelial cells, carries core 2 glycans but in breast carcinomas the simple core 1 based glycans are added. The binding of the monoclonal antibody SM3 to its peptide epitope in the tandem repeat of MUC1 is blocked by the branched core 2 glycans found on MUC1 expressed by normal cells. Thus SM3 does not bind to MUC1 expressed by normal mammary epithelial cells but reacts with more than 90% of breast carcinomas, suggesting that the loss of at least some core 2 glycans is a very common event in breast carcinogenesis. To determine if the change in glycosylation observed in breast carcinomas confers an advantage to cancer cells, murine mammary carcinoma cell lines were developed that express MUC1 carrying core 2 or core 1 linked glycans. The in vivo growth rate in wild-type and nude mice were identical regardless of the O-glycosylation patterns. However, the tumors that grew out of wild-type mice lost most of their MUC1 expression. In contrast, in MUC1 transgenic mice, where expression of MUC1 was retained by the tumor, a striking difference in growth rate was observed. In these mice, cells expressing core 1 glycans grew significantly faster than cells expressing core 2 glycans. These data suggest that MUC1 transgenic mice are more tolerant to core 1 expressing tumors than to tumors expressing core 2.

Retargeting of human T cells to tumor-associated MUC1: the evolution of a chimeric antigen receptor.

MUC1 is a highly attractive immunotherapeutic target owing to increased expression, altered glycosylation, and loss of polarity in >80% of human cancers. To exploit this, we have constructed a panel of chimeric Ag receptors (CAR) that bind selectively to tumor-associated MUC1. Two parameters proved crucial in optimizing the CAR ectodomain. First, we observed that the binding of CAR-grafted T cells to anchored MUC1 is subject to steric hindrance, independent of glycosylation status. This was overcome by insertion of the flexible and elongated hinge found in immunoglobulins of the IgD isotype. Second, CAR function was highly dependent upon strong binding capacity across a broad range of tumor-associated MUC1 glycoforms. This was realized by using an Ab-derived single-chain variable fragment (scFv) cloned from the HMFG2 hybridoma. To optimize CAR signaling, tripartite endodomains were constructed. Ultimately, this iterative design process yielded a potent receptor termed HOX that contains a fused CD28/OX40/CD3zeta endodomain. HOX-expressing T cells proliferate vigorously upon repeated encounter with soluble or membrane-associated MUC1, mediate production of proinflammatory cytokines (IFN-gamma and IL-17), and elicit brisk killing of MUC1(+) tumor cells. To test function in vivo, a tumor xenograft model was derived using MDA-MB-435 cells engineered to coexpress MUC1 and luciferase. Mice bearing an established tumor were treated i.p. with a single dose of engineered T cells. Compared with control mice, this treatment resulted in a significant delay in tumor growth as measured by serial bioluminescence imaging. Together, these data demonstrate for the first time that the near-ubiquitous MUC1 tumor Ag can be targeted using CAR-grafted T cells.

PLU-1 nuclear protein, which is upregulated in breast cancer, shows restricted expression in normal human adult tissues: a new cancer/testis antigen?

The PLU-1 gene is expressed at the level of message in breast cancers and breast cancer cell lines and shows restricted expression in normal adult tissues with the exception of testis. The predicted protein sequence contains several domains, including the PLU domain, which is shared by other proteins involved in transcription and/or development. We have developed a polyclonal antiserum to a C-terminal fragment of the PLU-1 protein, which shows little homology to other family members. Immunohistochemical analysis with the antiserum alpha-PLU-1C confirmed the nuclear localisation of PLU-1. alpha-PLU-1C also reacted with the mouse homologue of PLU-1 (mPlu-1) but not with the closest family member, RBP2. Using Western blot analysis, PLU-1 was shown to be well expressed in breast cancers and breast cancer cell lines, while it was not detected in a range of normal adult tissues. Our results suggest that the PLU-1 protein may belong to the class of testis/cancer antigens.