Immunologic Research 2006;36/1–3:73–82

Human Tumor Antigens, Immunosurveillance, and Cancer Vaccines

Olivera J. Finn Department of Immunology, University of Pittsburgh School of Medicine

Abstract Cancer is a serious health problem as well as a scientific challenge. A lot has been learned about the process of transformation of a normal cell into a tumor cell by studying genes and proteins that regulate this process either in cis or in trans. However, whether these molecular mechanisms succeed in fulfilling their potential to give a clinically evident disease depends in great measure on the host response to those molecular changes. The work of my laboratory aims to provide evidence in animal models as well as in cancer patients that immune system can control cancer growth and that this important function can be improved through vaccination with welldefined tumor antigens.

Introduction The last two decades have seen tremendous progress in tumor immunology owing to an explosion of new knowledge about the immune system in general, and to the development of new techniques, critical reagents, and relevant animal models for observing and measuring antitumor immune responses. What gave the field additional energy in the early 1990s was the newly acquired ability to explore the human immune system for evi-

Olivera J. Finn E1040 Biomedical Science Tower University of Pittsburgh School of Medicine Pittsburgh, PA, 15261 E-mail: [email protected]

© 2006 Humana Press Inc. 0257–277X/ (Online)1559-0755/06/ 36/1–3:73–82/$30.00

Key Words MUC1 Cyclin B1 Premalignant lesions Cancer stem cells T cells Dendritic cells

dence that tumors are recognized by multiple effector mechanisms and can be specifically destroyed (1). Experiments in mouse models had identified cytotoxic T cells as primary antitumor effector cells. Once interleukin-2 was discovered as the T cell growth factor, it became possible to expand in vitro cytotoxic T cells (CTL) from cancer patients and show that they are able to recognize and kill human tumors (2). This recognition was MHC restricted, suggesting a whole unexplored universe of tumor-specific antigenic peptides,

73

thus yielding many more targets on tumors than had previously been identified using antibodies against strictly cell-surface molecules. We entered this field in 1988 with a publication of our discovery that patients suffering from pancreatic cancer had in the tumor-draining lymph nodes cytotoxic T cells capable of killing pancreatic tumor cells (3). This prompted the last 17 yr of investigation into the areas of tumor antigen discovery and tumor–immune system interactions. Insight into the successes and failures of tumor immunosurveillance, we believe, will inform on ways to elicit better tumor-specific immune responses with cancer vaccines based on our tumor antigens and better understanding of tumor immunity (4). This article will review our past achievements, current undertakings, and future goals, as well as some formidable barriers in our way. More detailed information can be found in our published papers that comprise the reference section. Human Tumor Antigens Tumors express many molecules that can be recognized by the immune system and they can be divided into two large categories, those that are expressed only on tumors and not on normal cells (tumor-specific antigens) and those that are expressed on both tumors and normal cells, with some important differences (tumor-associated antigens). Among the tumor-specific antigens are products of genes mutated by chemical or physical carcinogens, random DNA mutations, or non-random mutations in cancer-related genes such as ras, bcr-abl, p53, etc. These are very attractive targets of immunotherapy because of the expected exquisite specificity of such therapy that should destroy tumors and not harm normal cells. Immune response against these antigens is, indeed, possible to generate, but in animal models it has been shown that

74

tumor cells do not express sufficient number of these peptides in MHC class I molecules to be good targets for CTL. Other tumor-specific antigens are encoded by oncogenic viruses, such as antigens derived from the human papiloma virus (HPV) and expressed on cervical cancers, or those derived from hepatitis viruses B and C, and expressed on liver cancers. While it has not yet been fully understood why in some individuals the immune response fails to clear the virus and reject cancers caused by these viruses, tremendous success has been achieved using vaccines against HBV and HPV to prevent initial infection and in that way prevent cancer. Interestingly, concerted efforts to discover tumor antigens for the majority of human tumors of unknown etiology have resulted in the identification of primarily tumor-associated antigens, normal molecules that are expressed by the tumor in an apparently immunogenic form (5). This includes many differentiation and lineage-specific molecules, such as carcinoembryonic antigen (CEA), alpha-fetorpotein, gangliosides, tyrosinase, prostate specific antigen (PSA), and epithelial mucin MUC1. Another important group of antigens are derived from overexpressed transformation-related proteins, such as p53, HER-2/neu, and cyclin B1. MUC1 and cyclin B1 were defined as tumor antigens in our laboratory using tumorspecific cytotoxic T cells as indicators. In the case of MUC1, we isolated and expanded in vitro, in a T cell/tumor cell co-culture, tumorspecific T cells from lymph nodes draining sites of primary tumors and harvested at the time of tumor removal (3). This method was in use in the late 1980s and early 1990s and yielded many tumor-associated molecules whose ultimate utility was questioned because the existing immune response found in the patients was apparently not able to control the tumor. We discovered that cyclin B1 can also

Finn

be a tumor antigen by isolating MHC class I–bound peptides from a tumor cell, fractionating them in various ways, and loading peptide fractions onto dendritic cells (DC) derived from healthy immunocompetent individuals (6,7). These were used to prime naïve autologous T cells to generate tumor-specific CTL. This method tested immunogenicity of tumor peptides in the absence of immunomodulatory effects of the tumor on the immune system. At least in our experience, both MUC1, detected through the immune responses in cancer patients, and cyclin B1, detected by its immunogenicity to T cells from healthy individuals, are equally potent in priming immune responses. What we have determined, especially in the case of MUC1, is that the type of the immune response it elicits depends on the precise form of the molecule and how it is presented to the immune system. This information is important because well-designed vaccines can manipulate both its form and the context of its presentation. Good and Bad Interactions of MUC1 Tumor Antigen with the Immune System MUC1 is a heavily glycosylated transmembrane protein expressed primarily on polarized ductal or surface epithelia in many organs. These cells when transformed give rise to adenocarcinomas of the breast, ovary, lung, prostate, colon, pancreas, etc. On these tumors, MUC1 is highly overexpressed and hypoglycosylated, and MUC1-specific antibodies and T cells raised to this “tumor” form of MUC1 specifically target tumors and not normal MUC1+ cells (for review see ref. 8). With some exceptions that will be described below, healthy individuals do not have immune responses to MUC1, while patients with MUC1+ tumors have a weak response characterized by low frequency of CTL and low titer of IgM antibodies. MUC1-specific

Immune Control of Cancer

helper T cells are difficult to detect, providing the explanation for the lack of antibody switching to IgG or other isotypes. We have studied both normal and tumor form of MUC1 for its ability to be processed and presented by dendritic cells and activate helper T cells. We found that the fully glycosylated form of MUC1 can be taken up by DC but it is not trafficked into antigen-processing compartments and not cleaved into antigenic peptides (9). The tumor form that lacks sugars completely, or is only sparsely glycosylated with short sugars leaving large stretches of the polypeptide core unprotected, can be processed into peptides and glycopeptides, both presented on the DC surface in MHC class II molecules (10,11). Both peptide-specific and glycopeptide-specific T cells can be primed with these DCs (12). From our studies that compared immune responses induced by vaccination with peptides versus glycopeptides, we have determined that peptide-specific helper T cells in MUC1 transgenic mice are not functional (13) and that to induce a strong helper T cell response that will support expansion of MUC1-specific CTL and isotype switching in MUC1-specific B cells, MUC1 glycopeptides should be used as immunogens (Vlad et al., in preparation). We have shown that this is not due to preferential induction of regulatory cells but most likely to the deletion of the high-affinity T cells. We have recently generated peptide-specific TCR transgenic mice (Turner et al., in preparation), and are close to having glycopeptide-specific TCR mice, and will be able to study regulation of these cells in the presence and absence of MUC1. While the data described above are important when it comes to selection of the more immunogenic form of this antigen to include in a cancer vaccine, they do not fully explain why immune responses in the patient are ineffective. This is especially puzzling in

75

Fig. 1. Expression of MUC1 tumor antigen on colon cancer as well as premalignant adenomatous polyps. Tissue sections were stained with an anti-MUC1 antibody that recognizes the hypoglycosylated tumor form of MUC1.

light of our observations that patients with very early stage tumors or premalignant lesions have more robust immune responses that could be slowing down their disease. Fig. 1 shows MUC1 expression on colon cancer and on adenomatous polyps known to be precursors of colon cancer. Both of these tissue sections were stained with an antibody against the tumor form of MUC1. Unlike patients with colon cancer whose antiMUC1 immunity is weak, some patients with polyps generate strong immune responses, suggesting that the immune system is involved in the immune surveillance of the premalignant changes in the colon (unpublished). We are accumulating data on a large number of patients to confirm our preliminary observations that the presence of the anti-MUC1 immune response is associated with the decreased risk of polyp recurrence. Further evidence that we have recently obtained about the role of anti-MUC1 immunity in successful tumor immunosurveillance

76

comes from studying a large population of ovarian cancer patients and controls (14). Our collaborators, epidemiologists, had shown that certain conditions such as mastitis, pelvic surgery, bone fracture, and IUD use are associated with the lowered lifetime risk of ovarian cancer. All these conditions are accompanied by low-level infection and inflammation and involve tissues that express MUC1. Thus, they all have the potential to present MUC1 to the immune system under immunogenic conditions and elicit an immune response. When we examined sera of this cohort of cancer cases and controls, we found that these conditions did induce antiMUC1 immunity and that there was a highly statistically significant inverse correlation between the presence of this immune response and ovarian cancer risk. In addition to these “good” interactions of MUC1 with the immune system that lead to effective and protective immune responses, which we are trying to model with our cancer vaccines, we have also discovered an inter-

Finn

Fig. 2. Chemotaxis and maturation of dendritic cells (DC) through interactions with MUC1 at the tumor site. Immature DC are attracted to the MUC1+ tumors via a chemotactic gradient established by MUC1 shed by tumor cells. Immature DC bind MUC1, and this results in the upregulation of many cell surface markers characteristic of mature DC. They also start producing inflammatory cytokines IL-6 and TNF-α and promote activation of naïve T cells into Th2 effector cells that make IL-5 and IL-13.

esting way, depicted in Fig. 2, that this molecule can contribute to the subversion of the immune system by tumors. We were initially very encouraged to discover that the polypeptide core exposed on tumor MUC1 is chemotactic to immature dendritic cells. Our in vitro and in vivo experiments showed that MUC1+ tumors attract large infiltrates of these cells that would be expected to pick up MUC1 and other tumor antigens for presentation to T cells in the draining lymph node. We furthermore found that contact of the DC with the MUC1 on tumor cells led to their maturation, supporting their role as effective antigen-presenting cells. The surprise came when we examined the type of T cells that were primed by these DCs. Instead of expected type 1 helper T cells that could stimulate effective antitumor immunity, these DCs elicited type 2 T cells that are known to promote tumor

Immune Control of Cancer

growth (15). Thus, MUC1 that can serve as a potent immunogen to elicit either spontaneous or vaccine-induced effective antitumor immune responses, if encountered by the immune system on a fully developed tumor, serves to deviate the immune response to its ineffective mode. Cell Cycle Regulator, Cyclin B1, as a Tumor Antigen When we first identified peptides derived from cyclin B1 protein as targets of HLA-A2 restricted tumor-specific CTL, we assumed that increased proliferation of tumor cells was leading to an increased expression of this molecule and therefore increased representation of its peptides in the MHC class I molecules. We later determined that cyclin B1 was indeed hugely overexpressed in many tumor

77

Fig. 3. Cyclin B1 overexpressing lung tumor cells metastasize from the primary tumor to the draining lymph node where they appear to be contained at least temporarily by an organized wall of lymphocytes.

cells, but this had nothing to do with the fact that they were cycling but rather with their transformed phenotype. Our work to date has shown that cyclin B1 is overexpressed in many tumors and aberrantly localized in the cytoplasm rather than the nucleus. This overexpressed protein is ubiquitinated and degraded through the proteasome, leading to the immunogenic overabundance of its peptides in class I molecules on tumor cells. As we saw with MUC1, patients with cyclin B1 overexpressing tumors had both T cells and antibodies specific for this molecule (7,16). These responses were similarly of low level, but unlike MUC1, cyclin B1 was also able to induce helper cells that allowed antibody isotype switching to IgG or IgA. The presence of cyclin B1–overexpressing cells in primary tumors has been correlated previously with poor prognosis in lung cancer. We determined that the likely reason was that overexpression of this molecule is either

78

directly responsible or a specific marker of cells with a strong metastatic potential. Fig. 3 shows heterogeneity of expression of cyclin B1 in the primary tumor, but homogeneously high-level expression in tumor cells that have metastasized to the lymph node. We have recapitulated this in vitro with cells transfected with a cyclin B1 expression vector, showing that cells expressing higher levels of cyclin B1 are more invasive when tested in invasion assays across membranes coated with Matrigel (Yu et al., unpublished). In attempting to understand why cyclin B1 is aberrantly expressed in tumor cells, we made a discovery that allowed us to establish relevant animal models for testing the potential of anticyclin B1 immunity to control tumor growth. We determined that cyclin B1 overexpression was a result of inactivation of p53 function (17) and subsequent unscheduled overexpression of the transcription factor E2F1 (Fig. 4) that binds to the cyclin B1 promoter and activates its tran-

Finn

Fig. 4. Cyclin B1 and E2F1 are overexpressed in cells with defective p53 function. The Western blot was performed with anti-cyclin B1 and anti-E2F1 antibodies on cell lysates derived from two cell clones from the same tumor cell line, one of which was wild type for p53 and the other had a deletion in both p53 alleles.

scription (Suzuki et al., in preparation). We examined tumors that spontaneously arise in p53–/– mice and found that they all overexpress cyclin B1. We established tumor cell lines from these tumors and used them in our immunization and tumor challenge experiments. We showed that mice are not tolerant to mouse cyclin B1 and that immune responses elicited through various vaccination protocols can protect animals from tumor challenge (Fig. 5). Most important, we have been able to show protection of p53–/– mice and long delay in their tumor development after vaccination with cyclin B1 (Yu et al., in preparation). Attempts to Translate Laboratory Successes to the Clinic Encouraged by the immunogenicity of various forms of MUC1 and cyclin B1 in vitro and in animal models, we are also involved in testing their potential to induce or boost immunity in patients. To date we have completed four phase 1 trials of the 100 amino acids long MUC1 peptide with various adjuvants or loaded on DCs (18,19). Most trials have been in pancreatic cancer patients and vaccines have been administered following surgical

Immune Control of Cancer

removal of primary tumors. We have seen immune responses only rarely, but we have learned a lot about various ways that tumors suppress the immune system (20,21). New MUC1 trials are now being designed with two important modifications. The MUC1 peptide antigen will be replaced with the MUC1 glycopeptide, based on our animal studies that show higher affinity responses to this form in MUC1 transgenic mice. Second, early rather than late stage patients will be enrolled, based on the lack of toxicity of this vaccine and on our data showing that late stage patients are severely immunocompromised. Cyclin B1 vaccine trials will also be designed for early-stage disease. We have shown in a small study that patients with stages 1B and IIA lung cancer appear to be protected longer from disease recurrence if they have anti-cyclin B1 antibody at the time of surgery (Fig. 6), but that there is no correlation with outcome in late-stage disease. This provides support for vaccinating only patients with early disease (22). Work in Progress There are several projects currently underway in the laboratory that engender a lot of excitement. One has to do with our efforts to isolate cancer stem cells from breast, ovarian, lung, and other adenocarcinomas, and obtain their tumor antigen profile. We are primarily interested in seeing if the two tumor antigens we know most about, MUC1 and cyclin B1, are expressed by stem/progenitor cells in the same tumor-associated forms found on mature tumor cells. This is very important for two reasons. For a vaccine to be effective, the elicited immunity needs to be able to target not only the mature cells but also the stem cells that are responsible for tumor recurrence. Similarly, if and when we arrive at the point where cancer vaccines can be given in the prophylactic setting

79

Fig. 5. Vaccination of mice with cyclin B1 protects from cyclin B1+ tumors. Cyclin B1 recombinant protein was loaded onto bone marrow–derived dendritic cells and mice vaccinated three times, two weeks apart. Ten days after the last vaccine, mice were challenged with tumor cells and tumor growth monitored over a period of 6 wk.

Fig. 6. Presence of anti-cyclin B1 antibody delays recurrence of early stage lung cancer. Of seven patients with antibody (Ab+), only one patient recurred (time to recurrence, 8 months) while of nine antibody negative (Ab–) patients, 6 recurred (time to recurrence, 6-22 mo).

to individuals at high risk for cancer, it is important to know that those vaccines will be eliciting immunity capable of destroying the cancer from the very early stages of transformation. The second reason to show that immune responses can target tumor antigens on stem cells is that this

80

may turn out to be the only therapy capable of destroying cancer stem cells otherwise resistant to chemotherapy and to small molecules that target various signaling pathways, because stem cells exist in a resting state and possess a very active multiple drug resistance gene. Our work to date has shown that MUC1 is expressed in its hypoglycosylated (tumor) form on breast cancer stem cells (Engelman et al., in preparation). The second set of projects aims to derive several mouse models of spontaneous tumors that express MUC1 and cyclin B1. For that, we have been making double and triple transgenic mice by crossing MUC1 transgenic mice with recently derived mice with latent K-ras mutation that can be activated by regulated expression of Cre recombinase. Activation of the K-ras mutation initiates transformation and the resulting tumors, like their human counterparts, are MUC1+. We have also been able to show that many are also cyclin B1+. This allows us to study antigen-specific immune responses that can be generated as part of the normal immunosurveillance of these tumors and the specific immune mechanisms that are

Finn

of critical importance for tumor growth. Furthermore, it will allow us to see how the natural disease progression changes when the mice are either immunosuppressed or prevaccinated with specific tumor antigens. The third set of projects explores the influence of the tumor microenvironment on cancer progression and how that may change as the cancer develops from stem to progenitor to mature tumor cells. A lot of this work is being done in our new model of human inflammatory bowel disease (IBD) and IBD-related colon cancer (Beatty et al., submitted). Finally, we are very interested in the initial antitumor immune responses that may signal the presence of malignancy before any other biomarker of the disease can be picked up (23). To that end, we are working with large cohorts of patients at risk for developing lung, ovarian, or breast cancers evaluating the diagnostic, prognostic, and therapeutic value of the earliest immune responses against cancer (24) (Vella, unpublished). Conclusions The importance of the immune system in cancer prevention and control has been grossly underestimated mainly due to the

inability until recently to experimentally test this in relevant animal models. Similarly, the great potential of immunotherapy and immunoprevention has been grossly underestimated by the flawed translation of preclinical experience into the clinical arena, primarily due to inappropriate rules made by regulatory agencies. Science can always get better, but if it is not supported by scientifically sound regulations, it will remain academic. While being an academic scientist is a noble profession, a biomedical scientist hopes to also contribute to solving some of the major health problems in the world. We have our sights set on cancer. Acknowledgments The current members of my laboratory who are contributing to this work are Pamela Beatty, Laura Vella, Lixin Zhang, Sean Ryan, Michael Turner, Sylvia Wong, Kira Gantt, Xiaochuan Chen, Katja Engelman, Jia Xue, Andrew Lepisto, and John McKolanis. I am indebted to all my former lab trainees and my collaborators whose work may or may not be referenced here but was critically important for shaping all the ideas and the direction of the laboratory. The work has been funded by NCI.

References 1. Finn OJ, Lotze MT: A decade in the life of tumor immunology. Clin Cancer Res 2001;7:759s–760s. 2. Finn OJ: Tumor-rejection antigens recognized by T lymphocytes. Curr Opin Immunol 1993;5:701–708. 3. Barnd DL, Lan MS, Metzgar RS, Finn OJ: Specific, major histocompatibility complex-unrestricted recognition of tumor-associated mucins by human cytotoxic T cells. Proc Natl Acad Sci USA 1989;86:7159– 7163. 4. Finn OJ: Cancer vaccines: between the idea and the reality. Nat Rev Immunol 2003;3:630–641. 5. Graziano DF, Finn OJ: Tumor antigens and tumor antigen discovery. Cancer Treat Res 2004;123:89–111.

Immune Control of Cancer

6. Kao H, Amoscato AA, Ciborowski P, Finn OJ: A new strategy for tumor antigen discovery based on in vitro priming of naive T cells with dendritic cells. Clin Cancer Res 2001;7:773s–780s. 7. Kao H, Marto JA, Hoffmann TK, et al: Identification of cyclin B1 as a shared human epithelial tumor-associated antigen recognized by T cells. J Exp Med 2001;194: 1313–1323. 8. Vlad AM, Kettel JC, Alajez NM, Carlos CA, Finn OJ: MUC1 immunobiology: from discovery to clinical applications. Adv Immunol 2004;82:249–293. 9. Hiltbold EM, Vlad AM, Ciborowski P, Watkins SC, Finn OJ: The mechanism of unresponsiveness to circulating

81

10.

11.

12.

13.

14.

15.

82

tumor antigen MUC1 is a block in intracellular sorting and processing by dendritic cells. J Immunol 2000; 165:3730–3741. Hiltbold EM, Ciborowski P, Finn OJ: Naturally processed class II epitope from the tumor antigen MUC1 primes human CD4+ T cells. Cancer Res 1998;58: 5066–5070. Hiltbold EM, Alter MD, Ciborowski P, Finn OJ: Presentation of MUC1 tumor antigen by class I MHC and CTL function correlate with the glycosylation state of the protein taken Up by dendritic cells. Cell Immunol 1999;194:143–149. Vlad AM, Muller S, Cudic M, et al: Complex carbohydrates are not removed during processing of glycoproteins by dendritic cells: processing of tumor antigen MUC1 glycopeptides for presentation to major histocompatibility complex class II-restricted T cells. J Exp Med 2002;196:1435–1446. Soares MM, Mehta V, Finn OJ: Three different vaccines based on the 140-amino acid MUC1 peptide with seven tandemly repeated tumor-specific epitopes elicit distinct immune effector mechanisms in wild-type versus MUC1-transgenic mice with different potential for tumor rejection. J Immunol 2001;166:6555–6563. Cramer DW, Titus-Ernstoff L, McKolanis JR, et al: Conditions associated with antibodies against the tumorassociated antigen MUC1 and their relationship to risk for ovarian cancer. Cancer Epidemiol Biomarkers Prev 2005;14:1125–1131. Carlos CA, Dong HF, Howard OM, Oppenheim JJ, Hanisch FJ, Finn OJ: Human tumor antigen MUC1 is chemotactic for immature dendritic cells and elicits maturation but does not promote Th1 type immunity. J Immunol 2005;175:1628–1635.

16. Suzuki H, Graziano DF, McKolanis J, Finn OJ: T celldependent antibody responses against aberrantly expressed cyclin B1 protein in patients with cancer and premalignant disease. Clin Cancer Res 2005;11:1521–1526. 17. Yu M, Zhan Q, Finn OJ: Immune recognition of cyclin B1 as a tumor antigen is a result of its overexpression in human tumors that is caused by non-functional p53. Mol Immunol 2002;38:981–987. 18. Goydos JS, Elder E, Whiteside TL, Finn OJ, Lotze MT: A phase I trial of a synthetic mucin peptide vaccine. Induction of specific immune reactivity in patients with adenocarcinoma. J Surg Res 1996;63:298–304. 19. Ramanathan RK, Lee KM, McKolanis J, et al: Phase I study of a MUC1 vaccine composed of different doses of MUC1 peptide with SB-AS2 adjuvant in resected and locally advanced pancreatic cancer. Cancer Immunol Immunother 2005;54:254–264. 20. Schmielau J, Finn OJ: Activated granulocytes and granulocyte-derived hydrogen peroxide are the underlying mechanism of suppression of t-cell function in advanced cancer patients. Cancer Res 2001;61:4756–4760. 21. Schmielau J, Nalesnik MA, Finn OJ: Suppressed T-cell receptor zeta chain expression and cytokine production in pancreatic cancer patients. Clin Cancer Res 2001; 7:933s–939s. 22. Finn OJ: Premalignant lesions as targets for cancer vaccines. J Exp Med 2003;198:1623–1626. 23. Finn OJ: Immune response as a biomarker for cancer detection and a lot more. N Engl J Med 2005;353: 1288–1290. 24. Egloff AM, Weissfeld J, Land SR, Finn OJ: Evaluation of anticyclin b1 serum antibody as a diagnostic and prognostic biomarker for lung cancer. Ann N Y Acad Sci 2005;1062:29–40.

Finn