My foray into tumor immunology research 20 years ago was stimulated by the successes of my group in transplantation immunology research that I believed at the time would give us a technical and a conceptual direction. Moving in 1982 from Stanford University where I had done postdoctoral work on human HLA class II molecules to Duke University where research in HLA, especially as it related to transplantation, was heavily emphasized, I set up my laboratory to analyze human T cell-mediated graft rejection. The goals were to isolate T cells from needle biopsies of transplanted kidneys being monitored for rejection, expand them in vitro, and determine their function and Ag specificity. These were ambitious goals, because up until then growing human T cells in vitro was not possible and it had been only a short while since the T cell growth factor IL-1 (IL-2) had been discovered (1). A group of talented and hard-working graduate students turned 1 mm3 of kidney biopsy into a source of millions of graft-infiltrating T cells that could be interrogated for their Ag specificity and function (2, 3, 4, 5). As new students joined the laboratory and looked for new challenges, initiating studies on cancer-specific T cells by using what we had learned from our transplantation work seemed appropriate.

Unlike graft rejection, where HLA Ags were the predicted T cell targets, there were no candidate Ags for tumor-specific T cells and new techniques had to be devised for their in vitro stimulation and expansion. Moreover, the expanded T cells were to serve first and foremost as reagents to identify human tumor Ags. Thus, my capable and determined students and I joined the first wave of immunologists addressing human T cell recognition of cancer.

This was a very exciting time in my laboratory and in other tumor immunology groups as we successfully established tumor-specific T cell lines and used them promptly and imaginatively to define their targets, allowing us the first look into the nature of the long elusive molecules on human tumors that could be recognized by human T cells. Our work resulted in the identification and publication in 1989 of the human mucin MUC1 as one such Ag (6) with tumor-associated epitopes recognized by CTLs (7, 8).

By 1995, at least a dozen human tumor Ags had been identified (9). Based on previous work in mice and on what was known at the time about the underlying genetic and epigenetic causes of malignant transformation, we all expected that the tumor Ags revealed through the activity of tumor-specific T cells would be products of oncogenic viruses or fragments of randomly mutated cellular genes or cellular oncogenes. In other words, we expected to isolate unique tumor-specific Ags. Instead, one Ag after another that was identified was a peptide or multiple peptides derived from proteins expressed by many different tumors or all tumors of the same tissue type and that could also be found in corresponding normal cells. These became known as shared tumor-associated Ags or self/tumor Ags.

The tumor Ag we identified, MUC1, clearly fell into this category and we and others spent many years trying to understand what made this molecule the target of the immune system when expressed on tumor cells but not when expressed on normal cells. Before we identified it as a tumor Ag by human T cells, others had isolated mouse mAbs specific for human epithelial adenocarcinomas and several of those reacted with different epitopes on MUC1. These Abs could be divided into two groups: those that recognized MUC1 on both normal and tumor cells and those that recognized MUC1 only on tumor cells. On normal epithelial cells the Abs detected a low level of transmembrane MUC1 expression polarized to the apical cell surface and a low level of secreted MUC1 in the lumen of the ducts. In contrast, in tumors derived from epithelial cells such as breast, pancreatic, and colon cancers, very high level of MUC1 expression was seen all over the cell surface and high levels of secreted MUC1 could be seen in the tumor bed. So the first difference that the immune system appeared to see was the difference in the quantity of this Ag.

Abs that recognized only “normal” or only “tumor” MUC1, however, also suggested qualitative differences between the two forms of the molecule. Sequencing the MUC1 gene from normal and tumor cells showed no tumor-specific mutations and thus attention was focused on the protein. Work from our laboratory and others showed that the difference resided in the extracellular domain of MUC1 known as the VNTR (variable number of tandem repeats) region. MUC1 is a glycoprotein heavily glycosylated with O-linked carbohydrates on serine and threonine residues in the 20-aa-long tandem repeats. The carbohydrates are long, branched, and terminate in sialic acid, giving the normal molecule its mucinous character. For a number of reasons, some well understood and others less so, tumors glycosylate MUC1 incompletely and the carbohydrates that occupy the glycosylation sites in the tandem repeats are abnormal: they consist of only one sugar, GalNAc (known as Tn Ag), or two sugars, GalNAc-Gal (known as T Ag). This abnormal glycosylation generates new protein and carbohydrate epitopes recognized by tumor-specific Abs and allows processing of the tandem repeat region into peptides and glycopeptides recognized by T cells.

The difference in the levels of MUC1 expression between normal cells and tumor cells and the tumor-specific posttranslational modifications recognized by Abs and T cells suggest that, when made by tumor cells, this molecule is not perceived by the immune system as self but rather as abnormal self. Consequently, instead of continuing to refer to it as a self/tumor Ag it is more appropriate to consider it as an abnormal self/tumor Ag.

This is a very important distinction intended to change how we think about antitumor immune responses. If we continue to consider tumor-associated Ags as self-Ags, we will continue to assume that they are subject to self-tolerance and to believe that this is the reason why immune responses against cancer are in general weak and ineffective. Furthermore, our efforts to overcome this anticipated self-tolerance so as to generate strong and effective antitumor immunity, when successful, will continue to raise the specter of autoimmunity that currently prevents more appropriate clinical application of these Ags for cancer immunotherapy. On the other hand, if we understand the immune responses against these Ags as an immune reaction to the abnormal forms of former self-molecules that in tumors acquire epitopes that are foreign to the immune system and elicit and boost immunity to those epitopes, we are likely to be successful in generating strong antitumor immunity while maintaining self-tolerance against the normal forms of these Ags expressed by normal cells.

We have tested this experimentally with MUC1, and others have done the same with similar abnormal self/tumor Ags. By focusing only on the tandem repeat region of MUC1 that is abnormally glycosylated in tumor cells and using synthetic peptides and glycopeptides that mimic the tumor-associated epitopes, we have shown in many publications over many years that tumor MUC1-specific T cells can be generated in vitro from healthy individuals as well as from cancer patients (10, 11) and that in MUC1 transgenic animal models these immune responses cause tumor rejection or tumor prevention without causing autoimmunity (12). We have also insisted on testing the ability of this Ag to elicit or boost tumor-specific immunity in cancer patients, but because MUC1 has been considered a self/tumor Ag, the outcome of these trials has been significantly negatively impacted by having to test it in individuals with late stage disease (13, 14).

Looking at the ever-growing list of abnormal self/tumor Ags and a lot of work that has been done in many laboratories, including ours, to understand them as targets of antitumor immune responses, I see a new paradigm emerging in tumor immunology that may have a more general application to our understanding of immunosurveillance of not only cancer but also of viral and bacterial infections. This paradigm proposes that: 1) abnormal self is immunogenic; 2) immune responses against abnormal self are important for successful tumor immunosurveillance; and 3) based on the evidence that will be discussed below, immune memory against self-Ags abnormally expressed during dangerous but nonmalignant events such as infections and tissue inflammation is generated early in life and may be boosted through other events throughout life.

First experimental support for this idea came from observations by us and others that abnormal expression and glycosylation of MUC1, initially defined as tumor specific, can also be detected on ductal epithelial cells during lactation or in mastitis (15). Women with history of mastitis have Abs and T cells specific for abnormal MUC1. We also recently showed that inflammatory bowel disease leads to abnormal MUC1 expression in the colon (16). In collaboration with Dr. Daniel Cramer (Harvard University, Cambridge, MA), we tested the hypothesis that immune responses against abnormal MUC1 generated during events such as mastitis, pelvic surgery, mumps virus infection of the salivary gland ducts, and other events predicted to result in abnormal MUC1 expression would elicit anti-MUC1 immunity and immune memory that throughout life will be part of successful immunosurveillance of MUC1-expressing tumors. In a large retrospective case control study we showed that women who had experienced two or more of these events were three times as likely to have Abs against abnormal MUC1 than women who did not and that their lifetime risk of ovarian cancer was proportionately and significantly reduced (17).

Seven years ago we identified another tumor Ag, cyclin B1 (CB1), which we could also firmly place in the abnormal self/tumor Ag category (18, 19). CB1 is a cell cycle regulatory protein that is produced transiently in normal cells in very small amounts to promote G2-M transition, after which it is rapidly degraded. In many human tumors CB1 is constitutively expressed in high levels in the cytoplasm, leading to constitutive presentation of CB1 peptides in MHC molecules on tumor cells. T cells specific for CB1 peptides recognize only tumor cells and not normal proliferating cells. In the case of CB1, the abnormal self is characterized by constitutive overexpression while the normal self is only transiently expressed. Patients with CB1-overexpressing tumors have CB1-specific Abs and CB1-specific memory T cells (18, 20). More importantly, healthy individuals also have Abs and memory T cells specific for CB1, and the question we are interested in answering is what has elicited these immune responses and are they part of normal immunosurveillance against tumors. Two papers published recently in The Journal of Virology show that cells infected with the chicken pox virus, varicella zoster virus (VZV) (21), or with human CMV (22) cause abnormally high expression of CB1 in the cytoplasm resembling that found in tumor cells. This would suggest that exposure to viruses, such as those causing early childhood diseases, would establish virus-specific immune memory but also simultaneously immune memory for abnormal self-molecules expressed by the infected cells. This memory response may be the first line of defense against future infections with viruses that may not share viral Ags but induce the same abnormalities in the proteins of the host cell. Moreover, this immune memory may support better priming of immune responses against new viruses and ensure that the right type of immunity is generated.

When abnormalities in cellular proteins arise as a result of malignant transformation, immune memory for abnormal self would be expected to react against the tumor by recognizing these molecules as tumor Ags. In this context, it is relevant to review findings published by Ludewig et all in 2004 (23). The authors profiled Ab responses in mice following infection with either vaccinia virus or lymphocytic choriomeningitis virus. They found that in addition to IgG Abs specific for several viral proteins, both infections induced Abs to host cell proteins such as Golgi and endoplasmic reticulum proteins, DNA binding proteins, and cytoskeletal proteins. Importantly, 83% of vaccinia virus infection-related Abs and 79% of lymphocytic choriomeningitis virus infection-related Abs were against proteins previously described as human tumor-associated Ags. In ongoing experiments in my laboratory we are challenging mice that have recovered from viral infections with tumors that we know express known abnormal self/tumor Ags. Our preliminary results show that these mice control tumor growth significantly better than uninfected control mice (L. Vella et al., manuscript in preparation).

Thus, the new paradigm in tumor immunology sees the immune memory for abnormal self as an important contributor to cancer immunosurveillance. Immune responses against abnormal self do not cause autoimmunity because they are not a result of broken tolerance to self and do not target normal self. Vaccines based on these abnormal self/tumor Ags could be used to further strengthen this acquired immune memory for more effective cancer immunosurveillance.

The most attractive corollary of this new paradigm is that it brings up the possibility of a universal, all-purpose vaccine based on a set of abnormal self Ags that are shared by infected cells, cancer cells, and inflamed tissues (Fig. 1). Some of these are already known through the tumor Ag discovery effort and others would be of interest to define. Such a universal vaccine would replace pathogen-specific vaccines because it would prepare the immune system to recognize and destroy infected, inflamed, or transformed cells by recognizing specific changes in normal cellular molecules rather than specific causative agents of those changes.

I am personally energized by this possibility and have fallen even deeper in love with tumor immunology for providing me with yet another window from which to peer into the endless intricacies of the immune system. Having grown up in a socialist country, the former Yugoslavia, I am used to making 5-year plans. My next 5-year plan (my petoletka) is to use the power of proteomic approaches to reveal the identity of many abnormal self-molecules predicted in Fig. 1F2 and to try combinations of them as vaccines against cancer, against pathogens, and against chronic inflammatory diseases. As I have for the last 25 years, I will rely on the exceptional talents of my trainees and their courage to try something very new.

I presented only a few examples from 25 years of research performed in my laboratory by many talented young scientists. They have had a profound influence on the directions I have taken in my research and are responsible for the pleasure that I experience coming to the lab every day. They are: M. Carrie Miceli, Donna L. Barnd, Derek A. Persons, Nancy S. Schek, Lindsey A. Kerr, Bruce D. Gitter, Susan Hand, Bruce Lee Hall, Keith Jerome, Allan Kirk, Elisabeth Schweins, Dawen Bu, J. Darrell Fontenot, Yasuo Kotera, Gabriele Pecher, Yasuyuki Kirii, Nieves Domench, Robert Henderson, Julie Magarian-Blander, Mark Alter, Simon Barratt-Boyes, Ira Bergman, James Snyder, Elizabeth Hiltbold, Jan Schmileau, Melina Soares, Henry Kao, Pawel Ciborowski, Daniel Graziano, Anda Vlad, Jessica Candelora, Nehad Alajez, Hiroyuki Suzuki, Casey Carlos, Min Yu, Michael Turner, Pamela Beatty, Andrew Lepisto, Kira Gantt, Xiaochuan Chen, Anna Furr, Brian Geller, Annie Willman-Silk, Katja Engelman, Sharmila Pejawar-Gaddy, Sean Ryan, Laura Vella, Lixin Zhang, and Seung Chul Heo. Special thanks to Carrie Miceli, my first graduate student, for honoring me with a wonderful introduction before my Presidential Address.

My husband Seth and my children, Sasha and Sonja, have provided support and much needed balance to the challenging life of research. The question should never be how does one manage the demanding life of a scientist and also have a family. The better question is how does one manage without a family.

I am grateful to my mentors Henry Kaplan and Ron Levy, at Stanford University, Bernard Amos and Richard Metzgar at Duke University, and Ron Herberman and Art Levine at the University of Pittsburgh, and my colleagues, tumor immunologists, who have been generous with their time, advice and friendship: Ellen Vitetta, Mac Cheever, Jacques Banchereau, Esteban Celis, Ralph Reisfeld, Soldano Ferrone, Jim Allison and Ralph Steinman. My professional and personal life has been buoyed by a life-long friendship with three outstanding immunologists, my soul mates Paola Ricciardi-Castagnoli, Yehudit Bergman, and Donna Paulnock.

The author has no financial conflict of interest.

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.

Smith, K. A., P. E. Baker, S. Gillis, F. W. Ruscetti.
. Functional and molecular characteristics of T-cell growth factor.
Mol. Immunol.
Miceli, M. C., O. J. Finn.
. T cell receptor β-chain selection in human allograft rejection.
J. Immunol.
Kirk, A. D., M. A. Ibrahim, R. R. Bollinger, D. V. Dawson, O. J. Finn.
. Renal allograft-infiltrating lymphocytes. A prospective analysis of in vitro growth characteristics and clinical relevance.
Hand, S. L., B. L. Hall, O. J. Finn.
. T cell receptor V β gene usage in HLA-DR1-reactive human T cell populations. The predominance of V β 8.
Hall, B. L., S. L. Hand, M. D. Alter, A. D. Kirk, O. J. Finn.
. Variables affecting the T cell receptor V β repertoire heterogeneity of T cells infiltrating human renal allografts.
Transpl. Immunol.
Barnd, D. L., M. S. Lan, R. S. Metzgar, O. J. Finn.
. Specific, major histocompatibility complex-unrestricted recognition of tumor-associated mucins by human cytotoxic T cells.
Proc. Natl. Acad. Sci. USA
Jerome, K. R., D. L. Barnd, K. M. Bendt, C. M. Boyer, J. Taylor-Papadimitriou, I. F. McKenzie, R. C. Bast, Jr, O. J. Finn.
. Cytotoxic T-lymphocytes derived from patients with breast adenocarcinoma recognize an epitope present on the protein core of a mucin molecule preferentially expressed by malignant cells.
Cancer Res.
Magarian-Blander, J., P. Ciborowski, S. Hsia, S. C. Watkins, O. J. Finn.
. Intercellular and intracellular events following the MHC-unrestricted TCR recognition of a tumor-specific peptide epitope on the epithelial antigen MUC1.
J. Immunol.
Boon, T., T. F. Gajewski, P. G. Coulie.
. From defined human tumor antigens to effective immunization?.
Immunol. Today
Jerome, K. R., N. Domenech, O. J. Finn.
. Tumor-specific cytotoxic T cell clones from patients with breast and pancreatic adenocarcinoma recognize EBV-immortalized B cells transfected with polymorphic epithelial mucin complementary DNA.
J. Immunol.
Domenech, N., R. A. Henderson, O. J. Finn.
. Identification of an HLA-A11-restricted epitope from the tandem repeat domain of the epithelial tumor antigen mucin.
J. Immunol.
Soares, M. M., V. Mehta, O. J. Finn.
. 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.
Ramanathan, R. K., K. M. Lee, J. McKolanis, E. Hitbold, W. Schraut, A. J. Moser, E. Warnick, T. Whiteside, J. Osborne, H. Kim, R. Day, M. Troetschel, O. J. Finn.
. 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.
Goydos, J. S., E. Elder, T. L. Whiteside, O. J. Finn, M. T. Lotze.
. A phase I trial of a synthetic mucin peptide vaccine. Induction of specific immune reactivity in patients with adenocarcinoma.
J. Surg. Res.
Jerome, K. R., A. D. Kirk, G. Pecher, W. W. Ferguson, O. J. Finn.
. A survivor of breast cancer with immunity to MUC-1 mucin, and lactational mastitis.
Cancer Immunol. Immunother.
Beatty, P. L., S. E. Plevy, A. R. Sepulveda, O. J. Finn.
. Cutting edge: transgenic expression of human MUC1 in IL-10−/− mice accelerates inflammatory bowel disease and progression to colon cancer.
J. Immunol.
Cramer, D. W., L. Titus-Ernstoff, J. R. McKolanis, W. R. Welch, A. F. Vitonis, R. S. Berkowitz, O. J. Finn.
. Conditions associated with antibodies against the tumor-associated antigen MUC1 and their relationship to risk for ovarian cancer.
Cancer Epidemiol. Biomarkers Prev.
Kao, H., J. A. Marto, T. K. Hoffmann, J. Shabanowitz, S. D. Finkelstein, T. L. Whiteside, D. F. Hunt, O. J. Finn.
. Identification of cyclin B1 as a shared human epithelial tumor-associated antigen recognized by T cells.
J. Exp. Med.
Yu, M., Q. Zhan, O. J. Finn.
. 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.
Suzuki, H., D. F. Graziano, J. McKolanis, O. J. Finn.
. T cell-dependent antibody responses against aberrantly expressed cyclin B1 protein in patients with cancer and premalignant disease.
Clin. Cancer Res.
Leisenfelder, S. A., J. F. Moffat.
. Varicella-zoster virus infection of human foreskin fibroblast cells results in atypical cyclin expression and cyclin-dependent kinase activity.
J. Virol.
Sanchez, V., D. H. Spector.
. Cyclin-dependent kinase activity is required for efficient expression and posttranslational modification of human cytomegalovirus proteins and for production of extracellular particles.
J. Virol.
Ludewig, B., P. Krebs, H. Metters, J. Tatzel, O. Tureci, U. Sahin.
. Molecular characterization of virus-induced autoantibody responses.
J. Exp. Med.