This abstract was published in Cancer Immunity, a Cancer Research Institute journal that ceased publication in 2013 and is now provided online in association with Cancer Immunology Research.
As I began to prepare my presentation for this meeting, I asked my laboratory group to provide me with their simplest definition of the field of tumor immunology. Their answers were very similar. They defined tumor immunology as the study of immune tumor recognition and/or elimination (Figure 1). Continuing with this reductionist approach, I asked them to provide me with the minimum number of possible functional outcomes of immune tumor recognition and elimination. Once again, the members of my lab were in agreement. They felt that there were three, i.e., immune recognition and/or elimination of tumors does happen, cannot happen or can happen.
By does happen, we mean that tumors can be recognized and/or eliminated as a result of natural tumor-specific immune responses that develop in the host (Figure 1). This outcome is supported by the findings that (i) the immune system can indeed protect the host against the development of spontaneous and chemically induced tumors, (ii) the immunogenicity of a tumor is often imprinted on it by the immunological environment in which it develops and (iii) individuals with cancer sometimes develop spontaneous reactivity against the antigens of the tumor [reviewed in (1, 2, 3)]. By cannot happen, we mean that many different influences often render a tumor either invisible to the immune system or resistant to its cytocidal functions (Figure 1) [reviewed in (3)]. This situation can occur when the tumor is non-immunogenic, either because it never expressed any tumor antigens or lost them during development, or because it acquired defects in the capacity to present tumor antigens to cells of the immune system (4, 5, 6, 7). In addition, the immune system may not be able to recognize or eliminate a tumor because the tumor itself produces immunosuppressive moieties or induces immunosuppressive responses in the host (8, 9, 10, 11, 12, 13, 14). Finally, tumors may escape immune recognition and/or destruction simply because of their growth characteristics or anatomical location. By can happen, we are referring to the potential beneficial outcomes of immunotherapy (Figure 1) (15, 16). In this scenario, the attention of the immune system is re-directed to the presence of a growing tumor by some form of immunotherapeutic intervention - be it active or passive vaccination with molecules or cells or elimination of immunosuppressive influences either by antibody blockade or cellular depletion (17, 18, 19, 20, 21, 22).
Now this reductionist view of the field, although very much oversimplified, nicely frames the focus of the current meeting. Five talks address the statement that tumor recognition and/or elimination does happen. Two focus on NKT cells (23, 24), two with the interaction between cellular activating receptors and their ligands (25, 26) and one with a mutant mouse that expresses heightened innate immunity against tumor development (27). Interestingly, these talks address aspects of innate immunity to tumors, an area that until recently has not been heavily represented at this meeting in particular or in the field of tumor immunology in general. Three talks address the statement that tumor recognition and/or elimination cannot happen in an otherwise unmanipulated cancer-bearing individual. Two deal with regulatory T cells, their antigens and their capacity to inhibit naturally developing immune responses to tumors (28, 29), while the third deals with immunosuppression arising from signals such as CTLA-4 that inhibit T cell function (30). Finally, eighteen talks are directed at whether immunotherapy can lead to effective immune recognition and/or elimination of tumors. Importantly, these talks provide an excellent summary of the translational and clinical status of our field and also highlight the exciting work that has come out of the Cancer Vaccine Collaborative that is jointly sponsored by the Cancer Research Institute and the Ludwig Institute for Cancer Research. Four talks deal with tumor antigen discovery (31, 32, 33, 34), four with cancer vaccination studies (35, 36, 37, 38), four with new protocols of tumor vaccination (39, 40, 41, 42), two with monitoring the effectiveness of vaccination (43, 44) and four with ongoing clinical trials and attempts to define the mechanisms underlying clinical responses (45, 46, 47, 48).
As my lab and I continued to discuss the focus of this presentation, we were struck by the relevance of this reductionist model to the area of tumor immunology that we work in - the study of natural immune responses to developing primary tumors. This process was originally called "cancer immunosurveillance" but we renamed it "cancer immunoediting" about three years ago so as to better emphasize the fact that the immune system plays a dual role in the process (1, 49). Specifically, the cancer immunoediting hypothesis holds that the immune system not only protects the host against tumor development but can also promote tumor development by selecting for tumor escape variants with reduced immunogenicity. We have since gone on to refine this concept such that we now envisage cancer immunoediting as a process comprised of three phases, which we call the 3 Es of cancer immunoediting (1, 2, 3) (Figure 2). The first phase is called Elimination, which is the same as cancer immunosurveillance, in which cells and molecules of the innate and adaptive immune systems recognize and destroy developing tumors, thus protecting the host against cancer. The second phase is Equilibrium, similar to the concepts of tumor dormancy (50, 51) or viral latency, which is a protracted period in which the tumor and immune system enter into a dynamic equilibrium of tumor destruction and tumor escape. The third phase is Escape, where tumor variants that emerge from a Darwinian-type immune selection process of the equilibrium phase develop into clinically apparent tumors that grow in an unrestricted manner in the immunocompetent host.
When one directly compares the reductionist model of tumor immunology to cancer immunoediting, one sees great conceptual parallels. Specifically, the immune system does protect against tumor development, tumors that are sculpted by their interaction with immunity cannot be rejected in a naive immunocompetent host, yet immunologically-sculpted - or edited - tumors can be eliminated with appropriate immunotherapy. Let me provide you with a few examples that support each statement.
First, I will present data that support the statement that the immune system does protect against tumor development. Over the last 4-5 years, several groups including the groups of Mark Smyth, Adrian Hayday, Rolf Zinkernagel and my own group in collaboration with Lloyd Old have shown that any event that disables innate or adaptive immunity in mice renders them highly susceptible to the development of chemically-induced tumors and to the formation of spontaneous tumors of non-viral origin [reviewed in (2)]. Importantly, as shown in Figure 3, this statement is supported by experiments that employed gene-targeted or transgenic mice with genetically-encoded, developmental immunodeficiencies or wild type mice that developed normally but were then rendered immunodeficient by treatment with neutralizing or blocking monoclonal antibodies specific for distinct molecular or cellular components of the innate or adaptive immune system (52). Using these types of experimental approaches, several immune components (including immunologically important molecules such as IFN-gamma, IL-12, Perforin and TRAIL, or cells such as NK, NKT alpha beta- and gamma delta- cells) were clearly shown to be required for effective cancer immunosurveillance.
Here are a few specific examples from our own work that highlight the particularly important roles that lymphocytes and IFN-gamma play in the cancer immunosurveillance process. Shown in Figure 4 is an experiment performed by Vijay Shankaran and Hiroaki Ikeda when they were in my lab in which large groups of age- and sex-matched wild type 129 strain mice or 129 strain mice lacking either RAG2, IFNGR1 (the ligand binding subunit of the IFN-gamma receptor), STAT1 (the transcription factor responsible for inducing many IFN-gamma- and IFN-alpha/beta-dependent biologic responses), or both RAG2 and STAT1 (RkSk mice), were treated with the chemical carcinogen 3-methylcholanthrene (MCA) and tumor development was monitored over time (49). As can be seen, all of the gene-targeted mice developed 3-5 times more MCA-induced sarcomas than wild type mice. Since there were no significant differences in tumor susceptibility between any of the gene-targeted mice, we concluded that the IFN-gamma/STAT1 and lymphocyte dependent tumor suppressor mechanisms are heavily overlapping.
We also examined whether immunologically compromised mice showed an increase in spontaneous tumor formation. To test this possibility, Vijay Shankaran and Allen Bruce set aside several wild type mice, RAG2-/- mice and RkSk mice and followed the development of tumors in the aging population (49) (Figure 5). As seen in the upper panel, the vast majority (9 out of 11) of wild type mice aged 15-21 months showed no evidence of neoplastic disease. One mouse developed a benign Harderian gland cystadenoma at 19 months while another mouse aged 16 months had a premalignant intestinal tubular adenoma, but none of the wild type mice had cancer. In contrast, as shown in the middle panel, 12/12 RAG2-/- mice aged 15-16 months displayed neoplastic lesions. Six were found to harbor adenomas in the intestine that, in humans, are thought to represent precancerous lesions. More importantly, 6 of the older mice in this group (aged 15.8-16.1 months) presented with frank adenocarcinomas. Five had adenocarcinomas of the intestine while one had an adenocarcinoma in the lung. Even more impressive was the analysis shown in the bottom panel of 11 RAG2-/- x STAT1-/- (RkSk) mice housed in the same room. As seen in the bottom panel, all 11 RkSk mice developed neoplastic disease and 9/11 presented with unequivocal carcinomas. The appearance and progression of the tumors in these mice was generally more rapid than that seen in the RAG2-/- mice, with the earliest appearing at 12 months of age. It is interesting that these mice also developed a broader range of tumors compared to the aged mice that lacked Rag2 only. Six of eleven of these mice developed mammary carcinomas and these tumors tended to appear earlier than the intestinal and lung tumors that showed up as well. In addition, the number of mice with multiple tumors also increased. Thus mice that lack either lymphocytes and/or the ability to respond to IFN-gamma show a significantly increased rate of spontaneous tumor formation compared to mice with intact adaptive and innate immune systems.
During the last two years, Ruby Chan and Allen Bruce in the lab continued to enroll additional mice into this study and we now have a more complete picture with larger group sizes and longer observation periods (53). Neoplasia remains rare in our 129 strain wild type mice. Although 6/33 wild type mice eventually developed malignancies and another 5 displayed benign tumors, almost all of the disease in affected wild type mice occurred at 26-28 months, which is at the very end of the normal life spans of these mice. Moreover, the tumors appeared in random tissues/organs. In contrast, most of the immunodeficient RAG2-/- mice, IFN-gamma insensitive STAT1-/- mice and the doubly deficient RkSk mice developed cancer much earlier in their life times, with many presenting in what we would call middle age. Interestingly, spontaneous mammary gland tumors were noted in about 50% of aging mice that lack either STAT1 alone or both STAT1 and RAG2, revealing that the mammary gland tumor phenotype tracks with the STAT1 deficiency. Ruby has done a very nice job ruling out the possibility that the high incidence of mammary gland cancer in STAT1 deficient mice was due to viral infection and is currently working on the molecular basis of the disease. Taken together, this work shows that an effective cancer immunosurveillance process is indeed at work in mice.
Although it is not possible to obtain such direct experimental evidence for a cancer immunosurveillance process in humans, strong correlative clinical data has accumulated supporting that a similar process is indeed also operative in humans. Three lines of evidence support this conclusion [reviewed in (2)]. First, several studies have now shown that immunosuppressed transplant patients display a significantly higher susceptibility to the formation of a variety of different cancers of non-viral origin. Second, a positive correlation has been made between the presence and location of T cells - particularly CD8+ T cells - in a tumor and the survival of patients with a variety of different cancers. Third, cancer patients often develop spontaneous immune responses to the tumors that they carry. Thus, the combined work from many labs now strongly supports the statement that immunosurveillance happens.
Second, there is also strong data showing that tumors that are sculpted by their interaction with the immune system cannot be rejected in a naive immunocompetent host. This statement is supported by data from a follow-up study to the MCA carcinogenesis experiment just discussed in which we asked whether similar tumors generated in mice of similar genetic backgrounds that only differed by the presence or absence of an intact immune response displayed differences in their inherent immunogenicities. For this purpose we compared the in vivo growth phenotype of MCA-induced sarcomas derived from immunocompetent wild type mice versus immunodeficient RAG2-/- mice (49). As shown in Figure 6, 17/17 sarcomas derived from wild type mice (panel a) and 20/20 sarcomas from RAG2-/- mice (panel b) grew in an equivalent progressive manner when transplanted into immunodeficient RAG2-/- mice. Thus there are no inherent growth differences between tumors derived from normal and immunodeficient mice. Moreover, as seen in panel c, all 17 sarcomas from wild type mice grew progressively when transplanted into naive wild type mice. In contrast, 40% of the sarcomas that originated in immunodeficient RAG2-/- mice failed to establish progressively growing tumors in naive, syngeneic wild type mice even when injected at high cell numbers (panel d). Thus, chemically-induced tumor cells from immunodeficient RAG2-/- mice are, as a group, more immunogenic than tumors formed in immunocompetent mice. This result showed that the immune system eliminates the most immunogenic cells in a developing tumor and sometimes leaves behind tumor cell variants of reduced immunogenicity that escape immune recognition and/or elimination in the naive host.
Third, let me provide an example where effective immunotherapy can lead to immune destruction of an edited tumor. For these studies Dan Kaplan used an MCA-induced fibrosarcoma cell line derived from an IFNGR1-/- mouse called RAD.gR.28 as a model of an edited tumor that was unable to respond to IFN-gamma. When these sarcoma cells were transplanted into immunocompetent mice, they grew in a highly aggressive manner (54) (Figure 7). Allen Bruce in the lab has shown that as few as 10-100 IFN-gamma insensitive RAD.gR.28 cells can form a progressively growing tumor in naive wild type mice. Enforced expression of a cytoplasmically truncated form of IFNGR1 (IFNGR1deltaIC) in RAD.gR.28 cells neither reconstituted IFN-gamma receptor signaling, nor altered the aggressive growth behavior of the tumor. In contrast, complementation with full length, wild type IFNGR1 reconstituted IFN-gamma receptor signaling and tumor immunogenicity such that the tumor cell line was rejected when transplanted into naive syngeneic mice. In subsequent experiments, Vijay Shankaran and Allen Bruce explored what IFN-gamma was doing to tumor cells to alter their immunogenicity (49, 55) (Figure 8). As shown previously, IFN-gamma insensitive RAD.gR.28 tumor cells were highly tumorigenic and poorly immunogenic and thus formed progressively growing tumors when transplanted into naive syngeneic recipients. However, when the tumor cells were engineered for high expression of the MHC Class I components H2-Db or Tap1, they were rendered highly immunogenic and were rejected even when mice were challenged with 3 x 107 cells. The rejection was specific and MHC restricted since the same tumor transduced with H2-Kb was not rejected. Thus appropriate immunotherapy induced rejection of an otherwise non-rejectable tumor.
The current research program in my lab is now heavily based on the 3 E model of cancer immunoediting. The efforts of Hiroaki Ikeda, Gavin Dunn and Ravindra Uppaluri in the lab are aimed at understanding the elimination phase and are focused on answering the following questions: What are the key components of innate and adaptive immunity that protect the host against tumor development and how do they function? Catherine Koebel and Kathleen Sheehan in the lab are studying the equilibrium phase and are seeking answers to the questions: Is the immune system indeed capable of maintaining cancer in a dormant state? Does editing occur in the equilibrium phase? What are the molecular targets that are edited and which immune components function as the "editors"? Finally, Jack Bui, Ruby Chan and Mark Diamond in the lab are working to better understand the escape phase by determining the answers to the questions: What are the key alterations that a tumor must undergo to escape immune control? How does this happen? And, can these changes be used to define the extent to which a tumor has been edited so as to identify those patients who would most benefit from immunotherapy?
Let me finish by making a few statements about the implications of this work. First, I think the most significant clinical implication of the cancer immunoediting hypothesis is that most, if not all, tumors that develop in immunocompetent hosts have undergone immunological sculpting. Thus we need to find a way to determine the extent to which a tumor has been edited. Moreover, any immunotherapy regimen needs to take into account that the tumor has already found a way to circumvent immune recognition and elimination. Finally, I propose that if we can indeed obtain hard evidence supporting the existence of the equilibrium phase of cancer immunoediting, then defining the molecular mechanisms that mediate equilibrium may help to identify an additional desirable outcome of cancer immunotherapy. Of course cures are best but - if cancer cant be eliminated, then perhaps it might be possible to therapeutically maintain it in an induced and durable equilibrium phase. In this manner we may well be able to change the clinical course of a growing tumor - from one that cannot be recognized and/or rejected by immunity to a tumor that can be a target for immunotherapy such that immunity does control or eliminate the tumor and extend the life of the cancer patient (Figure 9).
The author is grateful to the many individuals from his lab who participated in these studies over the past 12 years. He is also fortunate to have collaborated with a number of colleagues from his own and other institutions. He is particularly appreciative of the long and productive collaboration and interaction he has had with Dr. Lloyd Old and his colleagues at the New York Branch of the Ludwig Institute for Cancer Research. The work performed in the authors lab discussed in this review was supported by grants from the National Cancer Institute, the Ludwig Institute for Cancer Research and the Cancer Research Institute.
- Copyright © 2005 by Robert D. Schreiber