A Believer’s Overview of Cancer Immunosurveillance and Immunotherapy

After a period of skepticism, overwhelming evidence for cancer immunosurveillance

It has been 60 y since the first experimental evidence was published by Prehn and Main (7) that mice could generate immunity against autochthonous carcinogen-induced tumors. The immune response was specific for each tumor, suggesting the existence of unique molecules that are recognized as tumor Ags. Several of these molecules were later identified as products of carcinogen-induced mutations in various genes, including the p53 tumor suppressor gene (8), which was later also found to be mutated in human tumors (9). Klein et al. (10) contemporaneously showed that even a progressing tumor could generate an immune response such that if it was removed, the mouse remained immune and could reject the challenge with the same tumor. Work with transplantable tumors in experimental mouse models (11), combined with rare but nevertheless remarkable observations of spontaneous tumor regressions in humans, raised the idea that surveillance of tumors was one of the highly important functions of the immune system. The main proponents were Burnet (12, 13) and Thomas (14), who independently proposed conceptually the same hypothesis that large complex organisms must possess a system that recognizes and destroys nascent tumors that likely arise frequently in tissues where cells undergo numerous proliferation cycles, each capable of giving rise to potentially carcinogenic genetic mutations.

The prevailing picture of the immune system at that time was that it evolved to distinguish self from nonself (15). In addition to tissue allografts, the nonself would include viruses, bacteria, and other pathogens. Opponents of the cancer immunosurveillance hypothesis held that tumors, being derived from self-tissues, would be invisible to the immune system. This narrow view of the immune system, which primarily applied to T and B cells and their Ag receptors, failed to explain and incorporate a number of findings that followed, such as evidence of the importance of innate immunity in initiating immune responses (16, 17), the need for costimulation (18), and the presence of TLRs (19) on APCs. If only specific recognition of nonself by an Ag receptor was important, these other molecules and cells would be superfluous. Eventually the self–nonself discrimination hypothesis was replaced by the danger hypothesis championed by Matzinger (20), which stated that to the immune system the “foreignness” of an entity is not as important as whether that entity causes damage to normal tissues. This helped rescue the cancer immunosurveillance hypothesis because clearly tumors could cause damage and that would be important to the immune system. The specific danger signals that would trigger an immune response against cancer were still to be identified.

Although the conceptual barrier was partially crossed, a technical barrier to the acceptance of the cancer immunosurveillance hypothesis remained—the need for an appropriate animal model. If the immune system were responsible for eliminating nascent tumors, then in immunocompromised animals there would be a higher incidence of spontaneous or carcinogen-induced tumors. Various methods of immunosuppression in mouse models were used, including neonatal thymectomy, steroids, and antilymphocyte serum. The results were far from conclusive. Even when a state of immune deficiency could be achieved, and even when immunocompromised mice did show increased incidence of carcinogen-promoted tumors, there was usually an alternative explanation for the observed results that could be considered as likely as cancer immunosurveillance.

The discovery of the mutant mouse without a thymus (21), the “nude” mouse, that exhibited multiple effects of its particular genetic mutation, including lack of T cells and a profound deficiency in adaptive immunity (22), promised to provide a perfect mouse model for testing the cancer immunosurveillance hypothesis. It also eventually, but fortunately only temporarily, led to its demise. The most influential experiments that appeared to disprove the hypothesis were those of Stutman (23, 24), confirmed by several other groups, that showed no difference in tumor incidence between CBA/H nude mice and wild-type CBA/H littermates treated at birth with a chemical carcinogen methylcholanthrene. Ironically, as he and others were using these experiments to reject the existence of cancer immunosurveillance, Stutman et al. (25) were reporting on a new cell type with tumor-killing capacity, the NK cell, present also in nude mice, and proposing that it might have a role in cancer surveillance. It would take at least two decades to gain a better understanding of the interplay between the adaptive immune system and the innate immune system to which NK cells belong, as well as the many cytokines and chemokines that help orchestrate antitumor immunity, before it was clear that experiments in nude mice were given too much credence considering the limited available information about its immune system.

In addition to the developments in immunology that brought better understanding of what it would take to generate immunity to cancer, development of new gene engineering technologies allowed the creation of new mouse strains that lacked diverse and very specific immune components or had deficiencies in signaling pathways important for immune effector functions. These included RAG^−/− mice deficient in T, B, and NKT cells; mice lacking an important immune cytokine IFN-γ or its receptor, or could not signal through the STAT-1 pathway used by IFN; perforin^−/− mice that lacked cytotoxic T and NK cell function; α/β T cell^−/− or γ/δ T cell^−/− mice; and IL-12^−/− mice. Each of these strains compared with their wild-type counterparts was found to be deficient in cancer immunosurveillance in one form or another (26–33). The cancer immunosurveillance hypothesis was back and quickly garnering support. Importantly, it was also being updated and modified to be consistent with all the new data.

Robert Schreiber and his group (34) proposed a new and improved version of the cancer immunosurveillance hypothesis, “tumor immunoediting.” It incorporates three different potential outcomes: tumor elimination, equilibrium with the immune system, and escape from immune control. The immune system is alerted to the presence of the tumor as it begins to exert abnormal physiological and metabolic pressure on the surrounding normal tissues (as it becomes dangerous). The innate system is activated first and its activities at the nascent tumor site cause a certain amount of tumor cell death and initiate an inflammatory environment that recruits additional innate cells to amplify inflammation and attract T and B cells. The outcome is tumor elimination, the main tenet of the basic cancer immunosurveillance hypothesis. The second tenet is that if the immune system is compromised and these orderly processes are disturbed, the tumor will escape. Data from the new mouse models provided the first look into the equilibrium phase. This occurrence is different from what was previously known as tumor dormancy, in that the tumor is not really dormant but continues to proliferate and mutate against the immune pressure (immune editing) until it finally evolves into a less antigenic tumor capable of escaping immune destruction. This ability to survive in the face of an immune attack has now been recognized as one of the major hallmarks of cancer (35).

Immunosurveillance of human tumors

Results of experiments in immunocompromised mice provided mechanistic explanations for similar observations in humans, which were at first anecdotal and later confirmed with retrospective and prospective analyses. Spontaneous regressions of growing tumors were repeatedly observed but, as in mice, there were many alternative explanations for those phenomena in addition to natural immune surveillance. The earliest experiments in humans that were intended to boost immune surveillance and affect cancer regression were those of W. Coley in the late nineteenth century who noticed that occasional serious bacterial infections in cancer patients were associated with tumor regressions. He proceeded to infect patients intentionally and saw increased numbers of cases of cancer regression, which he credited to the immune defenses against the pathogens being able to strengthen in some manner the immune defenses against the cancer. These experiments are considered the first approach to cancer therapy that intended to stimulate cancer immunosurveillance, the basic principle of modern immunotherapy.

The advent of organ transplantation that led to the development of strong immunosuppressive drugs also created an opportunity to determine if immunocompromised transplant patients would be more susceptible to developing cancer. Indeed, numerous studies showed that lifelong immunosuppression led to highly significant increases in over 30 different cancer types (35–41). The emergence of HIV and the accompanying AIDS also resulted in higher cancer incidence in the affected population (42–44).

In addition to these acquired immunodeficiencies, there are inborn primary immunodeficiencies. The ability to control infections in these populations and prolong life has allowed observations of increased incidence of cancer in these individuals as well (45). Conditions such as common variable immunodeficiency and X-linked agammaglobulinemea are associated with defective humoral immunity and an increased incidence of cancer (46, 47). Patients with hyper-IgE syndrome that carry a mutation in the STAT-3 gene have impaired B cell maturation into plasma cells, as well as deficiency in Th17 T cells and IL-17 production, which causes drastically reduced immunosurveillance of both viruses and cancer. Immunodeficiencies that result from mutations in DNA repair genes also show increased susceptibility to cancer and reduced immunosurveillance due to deficiencies in multiple immune cell functions (48). The importance of innate immunity in cancer immunosurveillance was also revealed through various primary immunodeficiencies. For example, individuals with a mutation in the GATA2 gene (49), a transcription factor responsible for differentiation of hematopoietic cells, and those with the CSF3R mutation (50) suffer among other things from neutropenia and other phagocytic disorders leading to disseminated bacterial and fungal infections and also multiple types of leukemias and lymphomas.

Identification of human tumor-specific Abs and T cells as the indisputable proof of cancer immunosurveillance

Development of hybridoma technology (51) that enabled production of monoclonal Abs launched a large effort to discover molecules on cancer cells that are different from normal cells. Mice were immunized with every type of human tumor or tumor cell line and monoclonal Abs from these immunizations were screened for the recognition of cancer cells and not their normal counterparts. These studies clearly showed that such molecules (also known as tumor Ags) existed and, furthermore, yielded potentially immunotherapeutic Abs that could be conjugated to toxins, drugs, or radioisotopes for use in cancer imaging/diagnosis or therapy. This work, however, did not bring the field any closer to confirming cancer immunosurveillance or further elucidating human tumor immunity. The question still remained whether the human immune system would also be capable of seeing these tumor-associated or tumor-specific Ags and if both Abs and T cells would be involved.

Efforts to answer these questions intensified in the mid-1980s and early 1990s, spurred by numerous advances in basic immunology, genomics, and proteomics and development of many useful technologies such as tandem mass spectrometry. Tumor-specific Abs could be isolated from cancer patients and used to screen random peptide and protein libraries or tumor gene expression libraries to identify target Ags (52). T cells from cancer patients could be grown in vitro in the newly discovered T cell growth factor, IL-2, and their tumor specificity maintained with tumor-loaded dendritic cells that had just been recognized as professional APCs, and methods to grow them from blood monocytes had just been established (53, 54). This work generated numerous tumor-specific reagents with which human tumor Ags could for the first time be identified and fully characterized.

In 1989, epithelial mucin, MUC1, was reported as the first human tumor Ag to be recognized by human CTLs that were grown from lymph nodes of patients with pancreatic cancer (55). MUC1 had previously been identified by a mouse mAb, DUPAN-2, (56) that detected its abnormal expression on all human adenocarcinomas, and by Abs HMFG-1, HMFG-2, SM3, and DF-3 against breast cancer (57, 58). The gene for this Ag was cloned soon thereafter (59, 60) allowing transfection into MUC1^– cells to confirm MUC1 as the target for tumor-specific CTLs. Using melanoma-specific CTL clones and transduction of melanoma genes into Ag-negative targets, the first melanoma tumor Ag was cloned and reported in 1991 (61). Another highly productive method for tumor Ag discovery was elution of all peptides bound to HLA class I or class II molecules from tumor cells, separating them by tandem mass spectrometry, loading them onto dendritic cells, and presenting them to tumor-specific T cell clones (62, 63). Peptides that activated the T cells were identified as candidate tumor Ags that could then be synthesized and further characterized and confirmed. Within several years many human tumor Ags were identified belonging to several different categories, and on both viral and nonviral cancers (64, 65). Some were products of mutated oncogenes, such as K-ras and H-ras, and others were nonmutated Ags differentially expressed on tumor versus normal cells. Recent technical improvements that increased the ease of gene sequencing enabled sequencing of total tumor genomes and focused attention on many random mutations that could generate new mutated peptide epitopes unique to each patient’s tumor, similar to the unique mouse tumor Ags in the early models of methylcholanthrene-induced sarcomas (66). In several instances, T cells specific for the mutated epitopes predicted by the gene mutation have been found in the patient (67). Thus, immune responses are spontaneously generated to both the nonmutated shared tumor Ags and the mutated unique tumor Ags as the tumor develops, as would be predicted by the cancer immunosurveillance and tumor-editing hypothesis.

Timing of immunosurveillance

Because most of the information about human antitumor immune responses was acquired studying immunity in cancer patients, the best understood phase of cancer immunosurveillance is escape. The conundrum that arose when tumor Ags were identified and antitumor humoral and cellular immunity confirmed was why and how the tumor escapes and would an antitumor immune response ever be a tumor rejection response (68). One way to show the protective function of antitumor immunity even in the escape phase has been to evaluate tumor-specific immunity at the time of diagnosis and its effect on the disease outcome. Many such studies were done and results showed that patients with pre-existing anticancer immunity at diagnosis have longer disease-free survival, slower tumor progression, and extended overall survival. The best known are studies that evaluated tumor-infiltrating T cells and their state of activation across different tumor types, which found that tumors that are more extensively infiltrated with activated T cells and other immune effector cells and show evidence of organized lymphoid structures within which these cells cooperate recur much later and patients experience longer survival (69–71). It was also learned, however, that with advancing stages of tumor, the infiltrating cell composition changes. The effector cells become fewer and less activated, whereas the tumor microenvironment becomes dominated by cells with regulatory and immunosuppressive activities, such as T regulatory cells, tumor-associated macrophages, and myeloid-derived suppressor cells. These cells and their soluble mediators interfere with the ability of tumor-specific effector T cells, NKT cells, and NK cells to kill tumor cells (72, 73).

There is still very little information available on when in cancer development the immune system becomes involved. The answer may be different for each tumor type or even for each tumor and its etiology or the initiating mutation. A lot of current emphasis is on using various imaging and other modern screening technologies to detect early tumors and their precursor lesions. The same type of research that has been performed on later stage tumors that yielded a very detailed picture of the tumor immune microenvironment is beginning to be done on early tumor stages and on premalignant lesions. The limited information so far shows that premalignant lesions are under immune surveillance (74–76) and that their progression to cancer is also accompanied by changes that begin to shift the balance from effector immune cells to regulatory and suppressive cells, which likely promote tumorigenesis.

Natural immunosurveillance as basis for immunotherapy

Fig. 1 illustrates the three established major outcomes of natural immunosurveillance of cancer but applied here to a premalignant lesion. An optimally functional immune system would be expected to detect very early the disorder in the normal tissue morphology and physiology and the danger it presents to its integrity, and to recruit multiple innate and adaptive immune effector mechanisms to eliminate abnormal cells and restore tissue homeostasis (77). If the race for control between the immune effector mechanisms and their regulatory and suppressive counterparts is tied, the premalignant lesion could remain in an equilibrium with the immune system without further progression. If the balance shifts in favor of the regulatory and immunosuppressive mechanisms, premalignant lesion could escape immune control and progress to metastatic cancer. The odds of one outcome versus another depend on many variables unique to each individual and some common variables such as age (78). The goal of immunotherapy is to intervene in natural immunosurveillance and to change the odds in favor of elimination or at least equilibrium. As Fig. 2 illustrates, a preventative vaccine based on tumor Ags expressed on both premalignant lesions and cancer could strengthen tumor-specific adaptive immunity and shift the balance at the site of a premalignant lesion in favor of elimination. This approach is the new frontier in immunotherapy (79). Other immunotherapies have been developed for tumors that have already escaped immune surveillance. Therapeutic vaccines and checkpoint inhibitors are designed to restore immunosurveillance. The experience with therapeutic vaccines is extensive but therapeutic efficacy has been limited to short-lived partial regressions or temporary disease stability (equilibrium) followed by tumor escape (80). This has resulted in the FDA approval of only one such vaccine, sipuleucel-T, for prostate cancer (4). So far the best therapeutic outcomes have been seen with checkpoint inhibitors such as anti–CTLA-4, anti–PD-1, and anti–PD-L1 (81), which have caused tumor regression and long-term equilibrium in a large percentage of treated patients and in some instances complete cancer elimination (82). There are almost monthly FDA approvals of one checkpoint inhibitor or another or their combinations for different cancer types. Adoptive T cell (83) and Ab (82) immunotherapy is designed for full elimination of cancer. The therapeutic effects have been impressive resulting in the FDA approval of several Abs (e.g., anti-Her2/neu trastuzumab for the treatment of breast cancer and anti-CD20 rituximab for the treatment of some leukemias and lymphomas) and the most recent approval of the first of many to come T cell therapies (anti-CD19 CAR Kymriah).

• Download figure
• Open in new tab
• Download powerpoint

FIGURE 1.

Three established outcomes of natural immunosurveillance against cancer applied to the setting of a premalignant lesion.

• Download figure
• Open in new tab
• Download powerpoint

FIGURE 2.

Immunotherapeutic interventions can change the outcome of natural immunosurveillance of a premalignant lesion or an advanced tumor.

Cancer immunosurveillance: variation on the premise

All portrayals of cancer immunosurveillance start with nascent tumors beginning to express danger signals that attract innate immunity, and with expression of new tumor Ags that elicit specific T cells and Abs. Finding T cells specific for the unique mutated tumor Ags supports this picture. Yet, the majority of known tumor Ags are shared and not mutated but tumor associated due to their abnormal expression (e.g., overexpression, differential posttranslational modification, and unscheduled expression) compared with healthy tissue. The change in expression of these molecules can be caused by many physiologic changes in the cell and its microenvironment and thus they could be transiently abnormally expressed by nonmalignant tissues under other dangerous circumstances. Many molecules that we know as tumor Ags are expressed in their abnormal “tumor” form on acutely inflamed tissues during viral infections or in the setting of chronic inflammation. These abnormal forms of self-antigens can be encountered very early in life during strong febrile infections that characterize the common childhood diseases with which the immune system has evolved and which can serve to train the immune system during its development. It has been shown that healthy individuals often have stronger immune responses against certain tumor Ags than cancer patients. Moreover, these responses appear to lower lifetime risk of cancer. Thus, immunosurveillance of cancer is part of the general immunosurveillance based on the immune memory for a family of self-molecules abnormally expressed in many diseases including cancer and marking diseased cells for immune destruction. I proposed this modification of the immunosurveillance hypothesis in 2008 in my American Association of Immunologists presidential address (84), based on experimental data derived from human studies (85, 86) that have been supplemented recently by additional studies in humans (87–89) and testing the concept in mice (90).

In addition to many recently shown effects of the microbiota on the outcome of cancer immunotherapy (91), Zitvogel et al. (92) have proposed the possibility that specific antitumor immunosurveillance is in place before a tumor develops but they offer another possible explanation for the existence of tumor Ag-specific immunity in the absence of tumor—the gut microbiome. They propose that microbial proteins might be sufficiently similar to tumor Ags and are thus capable of eliciting tumor-specific T cells and Abs that recognize future tumor cells via “antigenic mimicry.” These memory T cells primed as intraepithelial or lamina propria T cells can then be seeded to all epithelial tissues where they can provide immunosurveillance against, in this case, epithelial tumors. Microbiota at other sites could prime and seed T cells to different sites for immunosurveillance of other types of tumors.

What these two proposals have in common is the idea that immune surveillance based on shared Ags is part of the immune preparedness program, which serves as the first responder, maintains the balance in favor of antitumor effector cells, and likely promotes generation of other tumor Ag-specific T cells and Abs, including those against mutated tumor Ags, via the already known process of epitope spreading (93). If this is indeed the case, and the work is ongoing to test this, it would be possible to strengthen immunosurveillance much earlier in life with a vaccine based on a variety of these Ags, especially now that we have eliminated most childhood diseases and limited exposure to microorganisms through excessive hygiene.