Many lines of evidence show that cancer is not a single disease. However, there are important principles that apply to many, if not all, cancers. There is substantial evidence that cancers in mice and humans are the result of multiple sequential mutations. As a result, certain molecules in cancer cells are mutant, up- or downregulated, or no longer expressed. In addition, most if not all cancers show epigenetic changes in gene expression. An estimated 15% of the worldwide cancer incidence is attributed to infections,^14,15 but chemical and physical carcinogens are involved in the induction of most human cancers in industrialized countries.¹⁶ Many of these carcinogens are mutagens.^17,18,19 While most mutations in cancer cells seem to be acquired (somatic mutations), increasing numbers of germline mutations are being discovered that make individuals prone to develop cancer.²⁰ A cancer may require as many as 10 or more mutations to develop full malignancy. The first stage of cancer development is called tumor initiation,²¹ which is generally assumed to be irreversible due to somatic mutations or germline mutations in various oncogene, tumor suppressor gene, or deoxyribonucleic acid (DNA) repair pathways. Initiated cells do not form tumors. However, initiated cells clonally expand to premalignant lesions evolving over many years, often decades. This second protracted stage is driven by tumor promotion (ie, exposure to promoting conditions or chemicals^22,23,24 [see Cancer and Inflammation]). The most advanced stage of these premalignant lesions is referred to as intraepithelial neoplasia also carcinoma in situ, often abbreviated as, for example, CIN for cervical intraepithelial neoplasia, VIN for vulvar intraepithelial neoplasia, or PIN for prostatic sites. The premalignant process ends with invasion, the appearance of the first cancer cells. Remarkably, however, there is now clear molecular evidence for premalignant cells spreading to distant sites where these cells remain premalignant unless promoted to become malignant.^25,26,27

A cancer cell that survives treatment may cause recurrence of the entire malignancy. Such cells are referred to as cancer stem cells or also tumor-propagating cells.^28,29,30 The elimination of these cancer stem cells is critical to prevent relapse. The sweeping assertion that these cells represent an extremely small percentage of the cancer cells in a tumor is not supported by rigorous experimentation or clinical experience. Some of the misconception comes from the observed rarity of only a few human tumor cells able to adapt to growth in a foreign (mouse) milieu.³⁰ Furthermore, flow cytometric analysis of cancer cells may detect a rare population that results from intraclonal and nonheritable heterogeneity in cancer cell populations.^31,32

Tumor progression originally meant the change from a benign neoplasm to cancer^33,34 but is now usually used to describe the third phase of the multistep process. Invasive growth of a lesion usually ends with a highly aggressive, widely metastatic cancer that ultimately kills the host.^35,36 There is compelling evidence that most cancers are clonal in origin and that in cancer progression, new subpopulations of cells arise continuously due to Darwinian selection of genetic variants that have a growth advantage, escape homeostatic controls, or resist destruction by defense mechanisms or treatment.³⁷ During this evolution, sequential mutations result in changes in rate of growth, morphology, hormone dependence, enzyme and cytokine production, and expression of surface antigens. Importantly, by the time cancer is first detected in a patient, it measures usually at least 1 cm in average diameter, contains ˜10⁹ cancer cells, and has already undergone about 30 generations.¹³ Thus, most of the diversity of a cancer has already occurred at time of detection with only 10 generations left before death of the individual unless treatment intervenes.

The term primary indicates the tumor from which cancer cells emigrate to secondary sites (ie, metastatic growth in tumor-draining [sentinel] lymph nodes or more distant organs). Experimentalists use the term spontaneous metastases to describe metastases that occur without experimental manipulations; artificial metastases are caused by cancer cells injected into systemic or portal veins of a tumorfree mouse to cause lung or liver metastases, respectively. Cancer cells can disseminate without further cell division (microdissemination) or they divide only minimally causing micrometastases, conditions that can only be recognized by microscopy and immunohistochemistry but are potential sources of relapse, the central problem of cancer therapy. Similarly, residual microscopic foci of cancer cells may remain at sites of incompletely excised cancer and cause local recurrence. Efforts are ongoing to develop sensitive markers and assays for determining the need for additional therapy or determining the effectiveness of a therapy before relapse is detected clinically.

A critical determinant for any cancer therapy is the number of the targeted cancer cells that have proliferative potential (“cancer stem cells”). That number determines the likelihood of recurrence/relapse after most cancer cells have been destroyed by treatment. Even microscopic parts of a cancer left behind by a surgeon often lead to recurrence. Nevertheless, the size of tumor-stem cell population is not the only factor determining the likelihood of relapse as other cancer cells, even when dead or lethally irradiated, can increase the tumorigenicity of the remaining cancer cells by orders of magnitude.¹⁷⁸

Most human tumors are not detected until they are 0.5 to 1 cm in diameter, have a volume of ˜500 mm³, and contain ˜10⁹ cancer cells.¹³ In leukemia, malignant cells generally do not form tumors, but the patient has also ˜10⁹ cancer cells when the disease is clinically detected. It is not appropriate to adjust for the difference in host size when comparing cancer in man and mice. Not the size of the species but the size of the cancer cell population determines the chance of relapse because the latter correlates with the number of therapy-resistant variants causing relapse after therapy. Skipper, who pioneered combination chemotherapy of childhood leukemia in the mouse model of L1210 leukemia, targeted 10⁹ cancer stem cells as the starting population. This was one major reason why the principles he established in an animal model were clinically relevant and led to the cure of most childhood leukemias. Cellular heterogeneity within a tumor becomes much more relevant when the tumor accumulates 1 billion (10⁹) cancer cells, equivalent to a tumor with a diameter of 1 cm.^183,184,185 At this point, the accumulation of so many cancer cell variants makes it improbable that all cancer cells are susceptible to a single chemotherapeutic agent or specific T cell. Cancer cell variants can be considered analogous to drug-resistant bacteria or viruses (ie, the nature of the problem is fundamentally the same in both cases). Unfortunately, in most experimental studies on immunotherapy, almost fivefold smaller populations of cancer cells are being treated⁴ (Fig. 47.1).

Duration of growth of a cancer greatly influences experimental results. Summarizing three decades of studies on immunity to cancer, Woglom concluded in 1929 that immunotherapy is futile against an established tumor, and “nothing may accordingly be hoped for at present in respect to a successful therapy from this direction.”¹⁸⁶ Eighty-two years later, the outlook is not quite so grim, but unfortunately, the vast majority of research in animal models is still concentrating at treating malignant cell populations grown for an average of only 5 days after cancer cell inoculation⁴ (see Fig. 47.1). The word “established” is not a scientific term, yet is frequently used to describe neoplastic lesions caused by recently implanted cancer cells,¹⁸⁶ conveying the message that the malignancy being treated does not differ from what would be found in a human cancer patient. Just the opposite is true; transplanted tumors must grow for at least 2 weeks before they are histologically indistinguishable from autochthonous murine or human cancers.¹⁰⁷ Most human cancers have resided in the patient for months if not years before being detected and treated, whether primary, metastatic, dormant, or relapsing. An additional problem of experimental models is that growth of many serially transplanted “standard” cancer lines is so rapid that death may occur so early that treatment has to be started before solid tumors have established a microenvironment even vaguely comparable to that of an autochthonous tumor.

Tumors are masses and have weight and volumes best approximated by the formula of an ellipsoid (V = πwh/6 or V = 0.5236 lwh or V-½ lwh); length 1, width w, and height h are the orthogonal diameters in the three perpendicular
axes. As was carefully documented, the third dimension, height (or depth), has an inordinately large effect on errors in volume¹⁸⁷ but can usually be determined accurately for subcutaneous tumors.^187,188,189 Describing tumors as areas is incorrect and illogical despite widely practiced misuse. By definition, areas have no volume nor contain a single cancer cell. Area measurements as commonly used by radiologists frequently disagree with volumetric measurements.^187,188,190 Depth of a tumor can greatly vary and determines malignancy and prognosis (eg, in melanoma).

There are important variables when extrapolating the number of cancer cells from a given tumor volume; for example, the human or murine adenocarcinoma of the pancreas usually consists predominantly (˜90%) of nonmalignant stromal cells yet is highly aggressive, whereas in most tumors stromal and cancer cells are more balanced in numbers. Furthermore, while it takes weeks for a tumor to disappear completely even after all cancer cells have been eradicated, it is usually impossible to determine from volume measurements how many viable cancer cells are left. Certain cancers release hormones that have systemic effects such as insulin; even microscopic growth can have major systemic effects.¹⁹¹ While many experimental models utilize subcutaneous tumors, some cancers, particularly human tumor xenografts, may require transplantation to sites where it is more difficult to assess growth objectively.¹⁹² Cancer cells have been transduced with luciferase to give a signal proportional to the tumor size when substrate is injected; this allows whole body scans with a luminometer, though sensitivity decreases for fewer than 1,000 cancer cells, and there is absorption by overlying tissues.¹⁹³

The effects of the immune system on cancer can be read out in different ways. Certainly, cure of the cancer is the objective. There should be clear prolongation of survival and absence of relapse, but proving that all cancer cells have been eradicated is often difficult. Unfortunately, for most cancers, we lack assays sensitive enough to detect and quantify remaining dormant cancer cells. This information would indicate whether or not a patient requires further treatment. Only in a few types of cancers^194,195,196 does a negative diagnostic polymerase chain reaction analysis of the tissue reservoir from where a cancer may relapse suggest—though not prove— complete eradication. Experimental evidence for dormant cancer cells may come from provoking relapse by treating the animal with antibodies neutralizing factors or cells suspected of causing tumor dormancy. Experimentation should include waiting for minimally 30 days after the tumor has disappeared completely (it is best to wait several months or longer). A common unacceptable practice is using the word eradication or cure when there is no follow-up after the cancer becomes undetectable. Relapse of cancers may occur within days, weeks, months, or even years after complete disappearance, and is one of the most important problems of cancer therapy. Nevertheless, experimentalists commonly describe treatments as effective even when followed by rapid relapse.¹⁹⁷ Eradication means tearing a tree out with its roots so it cannot regrow, and the term is synonymous with cure. Thus, using the term eradication is only appropriate if the host does not harbor dormant cancer cells, comparable to “sterilizing immunity” in infections.^{198,199,200,201}

Dependent on the extent and type of destruction as well as when it occurred, tumors may be found to have smaller or larger volumes after treatment compared to controls. It is important to determine whether the rate of tumor growth has been altered or the rate of growth remained unaltered but the onset of growth has changed. Altered rates of growth require an ongoing process, whereas altered onset of outgrowth is usually caused by an event that happened at times of inoculation. Slower growth rates (growth inhibition) should be distinguished from shrinkage (regression) of a cancer. A steady size of a treated tumor compared to controls is referred to as growth arrest, equilibrium, or progression-free survival in cancer patients, an important goal when cure cannot be achieved.^169,202

It is completely inappropriate yet quite customary to study immunologic therapies specific for a self-antigen on a human cancer in mouse models (human tumor xenograft) when the murine host does not express the same target. These irrelevant models often yield impressive results and mislead the reader that the targeted molecule is tumor-specific when in the real situation it is not.^203,204 If someone subscribes to the widespread highly questionable perception that many, or even most, useful tumor antigens are self-antigens, then this investigator should also use appropriate experimentation. Either a mouse model must be used in which normal tissues express the target molecules closely resembling expression patterns in man or there should be a clear warning to the reader. For example, anti-carcinoembryonic antigen (CEA) immune responses caused toxicity in mice that expressed the target antigen also in normal tissues, as humans do,²⁰⁵ predicting the severe toxicity later observed in a clinical study.²⁰⁶ Side effects often become apparent only when the treatment is intense enough to provide clinical efficacy.²⁰⁷ It is therefore questionable to dismiss results in proper animal models.^208,209

Experimentalists have to be able to translate their findings to clinicians and vice versa. There is no single human cancer, let alone single animal model, that can serve as appropriate model for all human cancers.²⁸ Organ site as well as histologic type of a cancer may greatly influence the results. Thus the complexity of cancer makes it extremely important that the experimental model used to study cancer immunity be relevant to the question asked; a single model, if carefully chosen, may be appropriate to answer a specific question. Nevertheless, it is essential that we uncover the broader principles underlying cancer-host interactions. For example, the extensive clinical and experimental research on immunotherapy of melanoma has failed to answer the
central question: why are we struggling to make similar advances in the immunotherapy of more common cancers such as breast, colon, prostate, and lung cancers? Weiss argued in 1980 that the failures of clinical immunotherapy were due to using irrelevant laboratory tumor models for extrapolating results for clinical application.²¹⁰ One of the major reasons for translating results from animal models to clinical cancer immunotherapy being frustratingly evasive is the disregard of using truly established tumors for experimental therapy (see Fig. 47.1).⁴ Current experimentation in cancer immunology mostly uses young hosts to study the effects of the immune system on small, recently implanted inocula of cancer cells derived from tumors once induced in inbred mice (with homozygous genetic loci) and then passed in vivo for decades from animal to animal, whereas human cancers are already well established when first detected, have a very high probability of heterozygous loci, have never been transplanted, and develop in mostly older individuals.

Spontaneous. The term spontaneous cancer is defined as cancers arising “in the absence of any experimental manipulation.”²¹¹ Spontaneous murine cancers develop without any known exposure to carcinogens or genetic (often transgenic) manipulations introducing oncogenes or eliminating tumor suppressor genes. For example, spontaneous tumors can be observed in many mouse strains with advancing age. All genes have a spontaneous rate of mutation, and genetic instability, though often a consequence, may not be a requirement for tumor development.^212,213 As would be expected from the sporadic occurrence of spontaneous mutations, the occurrence of spontaneous tumors is random and unpredictable.

Carcinogen-Induced (Physical, Chemical, or Viral). Many of the physical and chemical carcinogens involved in the induction of cancers are mutagens.^16,17 Since repeated application of coal tar was used to induce the first chemically induced cancers,²¹⁴ a very large number of chemicals such as polycyclic hydrocarbons and nitroso compounds has been identified with remarkably potent cancer-inducing activity.²¹⁵ Cancers have been induced in many tissues and organs and several animal species. Ultraviolet light (UV) has potent skin cancer-inducing activity in man and mice,^216,217 and many well-defined models of UV-induced tumors are now available.^218,219 The potential of ionizing radiation to cause cancer in humans was recognized soon after Roentgen’s discovery of x-rays in 1895.^220,221 But study of radiation carcinogenesis in animals mostly occurred after World War II when large-scale tumor-induction studies were carried out in many species over the succeeding three decades in response to the threat of irradiation from nuclear reactors or bombs.^221,222,223 Finally, many models of viral cancer induction have been developed after Rous²²⁴ showed that viruses can also induce cancer in animals.

Transgene-Induced. Experimental cancers are produced artificially by inserting oncogenes into the germline of mice or by manipulating the mouse genome to allow excision of tumor suppressor genes. Genes used for these manipulations are driven by tissue- or cell lineage-specific promoters that are either constitutively active or inducible locally or systemically. While these mice become prone to develop cancer, the tumors they develop are autochthonous but by no means “spontaneous” despite widespread misuse of terminology.

Autochthonous tumors originate in the place where they were found (autochthonous, in Greek, means indigenous). The antonym is transplanted tumors. Autochthonous tumors can be spontaneous, carcinogen-induced, or transgeneinduced. Even when induced in the same experiment by the same mode of induction, autochthonous tumors will differ genetically, biologically, and antigenically from one another, because additional but individually differing genetic changes are required for each tumor to develop. Thus autochthonous tumors lack the uniformity of well-defined transplantable models. An advantage of autochthonous over transplanted tumors, however, is that the host’s immune system has been neither artificially primed nor altered by an inoculum.

Unlike autochthonous tumors developing after exposure to physical, chemical, or viral carcinogens, transgeneinduced autochthonous tumors have the disadvantage that transgene expression in the thymus during development usually causes systemic tolerance to the transgenic proteins. Nevertheless, several excellent transgenic cancer models have been developed.^225,226 Some models are based on immunologic findings first made in patients.¹² For example, the R24C mutation in the cyclin-dependent kinase 4, first identified as tumor-specific antigen by T cells of a melanoma patient, causes familial melanoma when in the human germline.²²⁷ Introducing this mutation in the germline makes mice highly susceptible to develop melanoma²²⁵; tumor development, however, requires additional carcinogenic insults followed by prolonged chemical promotion.

Certain oncogenes such as SV40T are powerful because they inactivate several important suppressor pathways and may therefore require fewer additional mutations. In some transgenic cancer models, the transforming genetic event can be controlled, sometimes reversibly, by topical or systemic application of an inducer such as tamoxifen or tetracycline.¹⁹³ Alternatively, systemic or topical application of Cre-recombinase may excise a “floxed” blocking element or attenuator of gene expression.²²⁸ These temporal controls appear to be helpful for answering important questions on tolerance, because expression at birth causes neonatal tolerance to a highly antigenic oncogene such as SV40 T antigen. An approach that more closely mimics the sporadic nature of human cancer relies on a spontaneous mutational event activating an introduced floxed oncogene. The sporadic nature of this event unfortunately also makes time and site of tumor development less predictable.²²⁸

Most current experimental work in tumor immunology uses transplantable tumors. However, there are substantial differences among transplantable tumors, and it is
important that the chosen cancer is appropriate for the particular question being asked. It is important to know that 24 hours after inoculation, the majority of the injected cancer cells usually have died,¹⁷⁸ leaving only a shallow outer rim of viable cancer cells at the oxygenated margin.¹⁰⁷ Histology or in vivo imaging reveals pronounced edema at the injection site. In fact, much of the “tumor growth” early after inoculation is due to tissue swelling caused by the inflammatory reaction. Cancer cells may well be invading surrounding normal tissues and be vascularized within 48 to 72 hours after transfer, thereby meeting standard criteria for malignancy. But experienced pathologists will immediately spot the abnormal inflammatory reaction to the transplantation injury and the necrotic center of the early inoculum. Thus, common terms such as “three-day established cancer” or “three-day established metastases” are highly misleading. Such necrosis would not be found in an autochthonous cancer of similar size. The inflammatory reaction progressively decreases with time, and at 14 days, the necrotic material has usually disappeared. The tumor then contains ˜10⁹ cancer cells, measures ˜1 cm diameter, and is histologically indistinguishable from primary autochthonous tumors. Ideally, therapy uses tumors 2 weeks or more after inoculation (see Fig. 47.1).

Even though the terms tumor transplantation and tumor injection are old and commonly used, they are quite misleading, as they usually refer to inoculation of tumor fragments or a suspension of cancer cells. A common error is the belief that tumors can be injected or “transplanted” like a vascularized organ such as kidney or heart. Instead, much of the inoculum—whether a cancer cell suspension or fragments of a tumor—dies initially and needs about 2 weeks before it becomes histologically indistinguishable for autochthonous tumors.

Tumors can be induced in animals by injecting cancer cell suspensions prepared from cultures or by mincing tumors into 1-mm³ pieces and injecting the tumor fragments using a 12-gauge trocar. Cancer cells in fragments are 10- to 100-fold more tumorigenic than stroma-free cancer cell suspensions.^47,108 It had long been known that certain cancers would only grow in mice when transplanted as tumor fragments. This was erroneously thought to be due to more cancer cells being inoculated with fragments.²²⁹ Later analyses revealed that fragments contained fewer cancer cells than injected cell suspensions, yet produced a higher take or larger tumors earlier.^47,108 Another erroneous explanation was that the stroma of tumor fragments provided a physical barrier preventing cancer cells from migrating to draining lymph nodes and priming a protective T-cell response.^108,230 More likely, more cancer cells remain viable when embedded in tumor stroma (by preventing anoikis¹⁶⁰) and therefore release less antigen than suspended cancer cells, most of which die. In any case, as long as cancer cells express sufficient levels of antigen, professional antigen-presenting cells in the tumor stroma pick up the antigen and travel to the draining lymph nodes where they present the antigen to naïve T cells.²³¹

It is important to know that whether suspensions of cancer cells or tumor fragments are being used, a threshold number of cells or fragments must be inoculated for tumors to develop. The threshold in T cell-deficient mice may be due to innate immunity or nonimmunologic mechanisms.^178,232 The increase in threshold in immunocompetent mice is probably due to adaptive immunity.

Cell Lines from Tumors Serially Passed in Mice. When choosing tumors that have been serially passed in animals for transplantation, it is critical to be aware of how such tumors were altered by the serial transplantation. As was discovered decades ago, even a single in vivo passage of a cancer can select for heritable cancer variants.^{45,219,233,234,235,236,237} Even half a century ago, investigators already noted that serial transplantations of these cancers “inevitably result in progression toward more rapid growth rate, loss of functional and histological differentiation, loss of responsiveness to extraneous stimuli”^238(p.522) and a diminution of strain specificity, a problem shared by B16.²³⁹ These serially transplanted cancers, such as B16, can easily be transferred in mice using cancer cell suspensions rather than tumor fragments. As a result of hundreds of passages in mice,²³⁹ these cancer cells have become resistant to anoikis caused by lack of stroma by acquiring alternate signaling pathways that replace the prosurvival signals of ECM.²⁴⁰ Many of these tumors grow and kill so fast that they must be treated early before they are truly established. Certainly, many of these tumors no longer resemble primary mouse or human tumors that virtually always grow at much slower rates and have been established for months or years before being treated.

B16 melanoma was derived in C57BL/6J mice in 1954 and is the parent of many available B16 sublines.²³⁹ B16 had been transplanted serially through 328 mice for 13 years before it became available as a standard cancer cell line. Nevertheless, B16 still is arguably the “Escherichia coli” of tumor immunology with almost 1,000 entries in PubMed in 2011. The A/J-derived Sa1 originated in 1947 in an A albino mouse and was serially transplanted in A/J mice for 1,017 generations for 19 years. Similarly, the A/J-derived neuroblastoma 1300 derived in 1940 was serially transplanted for decades from mouse to mouse. The Lewis lung carcinoma, isolated by Lewis,²⁴¹ arose spontaneously in 1951 at the Wistar Institute in a black C57 mouse (not a C57BL/6 mouse), was serially transplanted extensively, and is still being used in numerous studies as a model for exploring the immune responses to lung cancers and their metastases. When genetically inbred mouse strains became available as sources for murine cancers over half a century ago, serial transplantation was necessary for maintaining a tumor. Dependable long-term cryopreservation did not become available until the late 1960s and early 1970s.^242,243,244 Even after dependable cryopreservation became available, investigators still continued to use serial transplantations to propagate newly derived cancers such as the BALB/c-derived CT26 colon cancer,²⁴⁵ the C57BL/6-derived MC38 colon cancer,²⁴⁶ and the BALB/cCr-derived RENCA renal cancer.²⁴⁷ Many cell lines are renamed sublines of old parental cell lines. Thus, it is often overlooked that RMA, RMA-S, MBL-2, and EL-4 tumors are derived from the same single tumor line,²⁴⁸ most likely
E.L.4 induced by 9:10-dimethylbenzanthracene in 1945 by Gorer²⁴⁹ and serially transplanted for many decades from mouse to mouse. Importantly, referring to EL-4 as a C57BL/6 tumor is incorrect as this strain did not exist in 1945. Similarly, Neuro2a is a subline derived in 1969 from the A/J-derived neuroblastoma 1300 derived in 1940. Unfortunately, organizations providing these lines and/or publications often fail to cite the fact of long-term in vivo passage before cryopreservation.

Cell Lines from Autochthonous Mouse Tumors. Autochthonous cancers never need to be serially transplanted. Standard lines can be created by adapting the autochthonous cancer cells to culture and/or by freezing fragments or cells in liquid nitrogen.²¹⁸ To prove that antigens expressed by the malignant cancer are truly tumor-specific, nonmalignant cells, tissue and DNA from the host of tumor origin must be available.^218,219 This approach is commonly used by researchers studying the antigenicity and genetics of human cancers. Carefully controlled experimental cancers exist.^218,219

Human Tumor Xenografts. For certain purposes, human tumor xenografts are useful models. Human cancer cell lines grown in vitro can cause tumors when injected into T-, B-, and/or NK-deficient mice. It is important to remember that the stroma of such cancers is entirely of mouse origin and the cancer cell-stromal loops are defective because of multiple mismatches in ligand-receptor signaling.²⁵⁰ This problem can be partially alleviated by “humanizing” the recipient mouse. Humanizing usually refers to expressing certain human molecules such as receptors/cytokines in mice that lack murine T, B, and/or NK cells, and/or transferring human mesenchymal and/or hematopoietic cells to such mice. Interestingly, human T- and B-lymphocytes and fibroblasts contained in human tumor fragments may coengraft thereby “humanizing” the mouse¹⁹² (eg, when injecting cell clusters or aggregates from human fresh ovarian cancers into nonobese diabetic-severe combined immunodeficiency IL2γR null mice). Progressive growth of these xenografts leads with great regularity to ascites formation, and pleural metastasis closely simulating classical tumor progression observed in patients with ovarian cancer.¹⁹² Despite continuous improvement, mice can never be completely humanized. Only very few cancer cells of the xenografts may be able to progress in the chimeric milieu.³⁰ Nevertheless, once a tumor grows, its sensitivity to potential therapeutic agents might reveal the sensitivity of the original cancer growing in the patient.

In the late 1800s,^260,261 it was discovered that in some instances tumors developing spontaneously in experimental animals could be transplanted into other animals of the same species and in this way could be propagated continuously. This finding provided an important experimental tool for cancer research.^40,262 Immediately, scientists began to investigate the possibility of immunizing against such transplantable cancers. Rodents exposed to a small nonlethal challenge of certain tumors became immune to subsequent challenge with large transplants of the same tumor that regularly killed nonimmunized recipients. Also, complete removal of the transplanted tumors, after initial growth, immunized animals against that tumor. These early results seemed to suggest that immunization against cancer was possible. Furthermore, there were certain other spontaneous tumors that were not readily transplantable, and this was taken as evidence for “natural resistance” or “natural immunity” to the cancers. Many years later, it became clear that no such conclusions could be drawn from these early studies, because outbred, or incompletely inbred, rats or mice had been used. The problem became apparent when it was realized that the immunization with tumor would also immunize the host against normal tissue of the donor and that normal tissues of the donor could also immunize the host against the tumor.¹⁸⁶ These experiments brought the idea of tumorspecific antigens into disrepute but also started the search for antigens that caused rejection of normal transplanted tissue. This research eventually led to the discovery of the MHC and to the development of inbred mouse strains.^8,263,264 Once inbred mouse strains became available, it was found that cancers transplanted within an inbred mouse strain usually grew so well that the existence of tumor-specific antigens seemed very unlikely. In fact, transplantability of tumors in syngeneic animals became (and still is) a diagnostic criterion for the malignant phenotype of an experimental tumor. This criterion was especially useful because many rodent
tumors are cancers of nonepithelial origin (sarcomas) and also because a clear histologic demonstration of local invasive growth can be especially difficult in such cancers.

After discovery of the MHC and the development of inbred mouse strains, the gloom over tumor immunology ended with the discovery that inbred mice could be immunized against syngeneic sarcomas induced by the chemical carcinogen methylcholanthrene (MCA). The first demonstration of induced immunity to a transplantable MCA-induced sarcoma was by Gross in 1943²⁶⁵; however, it was not until the 1950s that more complete experiments provided unequivocal evidence for “tumor-specific” rejection of transplanted cancers.^{229,266,267,268,269,270} In particular, the experiments of Prehn and Main²⁶⁸ in 1957 made it likely that the rejection antigens on the MCA-induced sarcomas were functionally tumor-specific, because transplantation assays could not detect these antigens in normal tissues of the mice used. However, irrefutable evidence for the existence of tumorspecific antigens in autochthonous unmanipulated cancers only came after a 50-year-long search^{3,5,218,219,236,267,268,269,270,271,272,273,274,275,276,277,278,279,280,281,282} when in 1995, it was proven that tumor-specific antigens on cancer cells were encoded by somatic cancer-specific mutations absent in the normal cells of the host of tumor origin.^10,11,12

Even though it is clear that cancer-specific antigens are encoded by somatic cancer-specific mutations, the terms “tumor-specific” or “cancer-specific” are frequently used inappropriately, sometimes even used in conjunction as “relatively” tumor-specific. The discussion is far from being semantic. In reality, an antigen either is or is not tumor-specific. Germline controls are absolutely critical for proving mutations are somatic and tumor-specific.^{283,284,285,286} Unfortunately, germline controls are missing from virtually all tumors used for experimental work today. Nevertheless, mouse tumors with proper autologous controls have also been used and are available for distribution.^218,219 Notorious problems are genetic polymorphisms. For example, even after 20 backcross generations when a strain is arbitrarily pronounced “inbred” (because it is then more than 99.9% genetically identical), about 373 polymorphic proteinencoding loci remain allogeneic.²⁸⁷ Any one of these loci could encode a pseudotumor-specific antigen when a tumor is transplanted into a mouse misperceived to be fully syngeneic. Mice respond preferentially to nonself- or mutant-self-antigens whether caused by genetic polymorphism or tumor-specific somatic mutation.

T cell-mediated destruction of cancer cells requires the interaction of T-cell receptor (TCR), peptide, and MHC molecules. In this “three body problem,” two affinities simultaneously determine the interaction^288,289: the peptide to the MHC and that of the TCR to the peptide-MHC complex. Even when the complex cell-cell interaction is reduced to the three molecules interacting in vitro, biochemical analysis is still too complex for analyzing physiologic interactions. TCR affinity to peptide-MHC is therefore usually measured by plasmon resonance in the presence of saturating nonphysiologic amounts of the peptide, and TCR affinities measured this way range between (K_d) 1 to 100 µM.^288,289 This is a narrow range considering that affinities of peptides to the MHC range from 1 to more than 20,000 nM.²⁹⁰ This difference points at the peptide-MHC affinity probably being the greatest variable and emphasizes the paramount importance of choosing target peptides with highest possible affinity to the presenting MHC. Proper selection is particularly important as the amount of peptide produced by the cancer cell may be relatively small and always must compete for binding with all other peptides naturally present in the cancer or in the cells cross-presenting the antigen. Affinity of a peptide to a given MHC molecule is best measured empirically and many empirical affinities are already available (eg, http://tools.immuneepitope.org or www.syfpeithi.de).^290,291,292 However, even when a peptide binds with high affinity to MHC, it will only be expressed on the cell surface when it is naturally processed and present in sufficient amount.^293,294

“Reverse immunology” is the attempt to predict T-cell epitopes within a given amino acid sequence. Traditionally, “reverse immunology” has focused on finding optimal T-cell antigens, properly referred to as peptide epitopes, on infectious agents for generating vaccines. More recently, self- or mutant proteins recognized by T cells or antibodies from cancer patients have been analyzed in the search for peptide epitopes that may be effective targets of T-cell immunity. Computerized algorithms (eg, http://tools.immuneepitope.org^295,296 and www.syfpeithi.de²⁹⁷) have been developed to predict the affinity of peptide-MHC binding, appropriate proteasomal cleavage, and transport by the transporter associated with antigen processing (TAP).^{290,291,292,297} Though these algorithms are useful, it appears that these tools cannot replace empirical biochemical measurements. Many of the antigens proposed by reverse immunology have affinities to MHC insufficient to serve as effective targets when tested in appropriate animal models. Together, high peptide-MHC affinity is essential but not sufficient to predict that a given peptide serves as an efficient target.

Complete sequencings of cancer cell genomes from individual patients is becoming increasingly affordable and has revealed up to many thousands of mutations per cancer cell.^298,299,300 However < 1% of these mutations cause amino acid substitutions. For example, of the 33,345 nucleic acid base substitutions found in a human melanoma, most of the mutations were intergenic, intronic, noncoding, silent, or truncating. Only 187 caused amino acid substitutions in coding genes,³⁰⁰ and only a few of these 187 coding substitutions are expected to lead to mutant peptides that serve as effective targets. To be effective, the mutant peptide 1) must bind with a high affinity (IC₅₀ in the nM range) to the particular MHC molecules of that individual and 2) be naturally processed, 3) escape destruction by proteasome cleavage, and 4) be present in sufficient amounts. Only a few of the myriad of tumor-specific mutations identified in human or murine cancers give rise to antigens that fulfill these requirements. However, the frequency of unique tumor-specific targets in
a cancer cell may be larger than predicted when considering only mutations in coding sequences, because mutations in intron sequences can be translated and also encode tumorspecific antigens.¹¹ These mutations are over 50-fold more frequent than those in protein-encoding loci.

In the search for the best antigens to eradicate the cancer, T cells infiltrating tumors (TILs) have been recovered, expanded, and used therapeutically. However, TILs are removed from cancers that have not been destroyed. It is, therefore, a widespread misperception that the TILs were necessarily fighting cancer growth. Three questions become obvious: 1) Are strong antigens retained because the host has been tolerized during very early stage of cancer growth? 2) Can T cells specific for the strongest antigens be recovered from the TILs? 3) Can T cells specific for weak antigens promote rather than inhibit tumor growth?

It is possible that 1) potent rejection antigens are retained because they tolerize the host very early during cancer development,^228,301 2) a few T cells have not been tolerized and can be propagated for therapy, and 3) tolerization is reversible, and competent effector cells can be obtained from TILs. Clinically, the patients who fare best have T cells in their peripheral blood that are specific for antigens encoded by somatic tumor-specific mutations.^12,302,303 Whether the same T cells were infiltrating these cancers as TILs is unknown. In any case, it is likely that only a few (if any) of an endless array of antigens recognized on cancer cells by antibodies or tumor-infiltrating T cells may have significance as targets or diagnostic markers.

All cancers in man and mouse that have been analyzed carefully express bona fide tumor-specific antigens that could be targeted by T cells. Because tumor-specific antigens arise from mutations,^10,11,12 they are usually unique. Each patient’s cancer seems to have a unique set of mutations, and unique antigens can provoke powerful immune responses. In mice, immunization with one tumor protects only against the same tumor (Fig. 47.2).^5,274 Shared cancer-specific antigens do exist when the same mutations occur in several cancers, but by comparison, such antigens are relatively rare. Most cancer-specific mutations affect intracellular proteins that may be recognized by T cells as mutant peptide-MHC complex on the surface of viable cancer cells. Very few cancer-specific mutations affect surface proteins, such as the mutant epidermal growth factor receptor (EGFR), a shared tumor-specific antigen on glioma cells.^304,305

A major misconception is that unique cancer-specific antigens are caused by random mutations and are incidental to the oncogenic process. For example, a group of experts
stated in 2009^306(p.5326) that unique antigens come “from random mutations presumed to be present in all patients,” implying mutations encoding unique tumor antigens have no functional significance in the malignant process. Nothing is further from the truth. Of course, all mutations caused by chemical and physical carcinogens are random, but virtually all of them are lost during cancer evolution except those that promote the neoplastic process. When it was first shown in 1995 that cancers harbor cancer-specific antigens caused by somatic tumor-specific mutations,^10,11,12 every one of the three antigens represented a mutation in a tumor suppressor locus.^{307,308,309,310,311} For example, the mutation in the cycline-dependent kinase 4 reduces the binding to its inhibitor and tumor-suppressor protein p16INK4a,¹² and the same mutation is found in the germline in cases of familial melanoma.²²⁷ The fact that selection for a tumorspecific antigen commonly represents a mutation in genes with functional significance in the malignancy of that cancer (oncogenes or tumor suppressor genes) has been shown in many other studies.^{219,312,313,314,315} Remarkably, T cells against unique tumor antigens identify mostly novel, functionally important tumor-specific mutations and oncogenic proteins that would not have been easily detected by other technologies.^{10,11,12,219,307,308,309,316} Importantly, several of these unique tumor-specific antigens are excellent targets because they cannot be lost by immune selection. The reason is that some of these mutant proteins are not only oncogenic but are also needed to provide an essential household function no longer provided by the second allele due to Knudson-type loss³¹⁷ or mutational inactivation.^{302,307,308,309}

Another misperception surrounding unique tumor-specific antigens is that they have remained unexploited clinically, because truly personalized therapy would be required. Again, nothing is further from the truth. One of the most effective immunologic treatments of cancer today is the adoptive cell transfer of autologous tumor-infiltrating lymphocytes from patients with metastatic melanoma: response rates are in the range of 50% to 70% of the patients, and a few of these patients are cured.³¹⁸ This truly personalized therapy involves reinfusion of the patients’ own lymphocytes isolated from the patients’ own melanomas and expanded in vitro. Because the TILs response is dominated by T cells to unique tumor-specific antigens,³⁰³ it is likely that the success of the reinfused T cells depends on their reactivity to unique tumor-specific antigens. Similarly, clinical studies show that therapeutic vaccinations with autologous cancer cells are likely to be much more effective,^319,320 confirming decades of experimental work.^{229,234,268,269,271,321,322} Self-antigens may serve as useful and effective targets in a very few instances (eg, CD20 and CD19), but industry supported by government panels and organizations focuses almost exclusively on self-antigens,^306,323,324 so far with very modest results.³²⁵ Finally, it is a misperception that immunizing each patient to the unique antigen of the individual’s tumor is impractical because of cost and should not be pursued. This is incorrect compared with other highly individualized treatments such as those used in renal transplantation. Thus, individualizing cancer therapy could be affordable as other strategies used in the clinics.

Ideally, antigens targeted on cancers would be expressed exclusively on malignant cells but be shared by cancers of the same type or at least subtype. Several cancer-specific mutations have been identified that are shared between cancers,^{9,304,326,327} but very few of them have so far been found to be an effective immunologic target. The main reason probably is that, aside from specificity, additional requirements for effective recognition by T cells are 1) high-affinity binding to the patients’ MHC molecules and 2) expression of the protein at sufficient amounts to compete with other peptide for presentation. Unique tumor-specific antigens must fulfill the same requirements but, because these unique mutations are more abundant, chances are much higher to yield an effective antigen. A few examples of shared tumor-specific antigens are discussed in the following.

Mutant Epidermal Growth Factor Receptor. See further under Immunotherapy.

Fusion Proteins. Fusion proteins found in cancer cells are the result of internal deletions (see mutant receptor EGFRvIII under Immunotherapy) or chromosomal translocations.^9,326 New antigenic determinants can result from the juxtaposition of previously distant amino acid sequences, resulting in a mutant peptide sequence at the breakpoint and possibly a change in conformational structure. The same chromosomal breakpoints consistently recur in different individuals with the same cancer, therefore result in shared tumor-specific antigens. Fusion proteins encoded by these translocations are usually essential for cancer maintenance, making them ideal targets for pharmacologic and immunologic intervention because tumor cells may not easily escape therapy by losing expression of the fusion proteins.³²⁸ Pharmacologic approaches are exemplified by the drug imatinib targeted to the fusion protein of the 9;22 translocation in chronic myelogenous leukemia.^9,294,329 The antikinase drug imatinib does not selectively inhibit the catalytic activity of the BCRABL fusion protein, and this results in toxicity to normal cells in the patient. The intracellular BCR/ABL fusion proteins can be recognized specifically by antibody but only in fixed cells.³³⁰ BCR/ABL fusion peptides can also be recognized by human CD4+ T cells in the context of MHC class II³³¹ and serve as targets for CD8+ human cytolytic T cells.³³² Why do we then still lack effective immunologic therapies for targeting this and other fusion proteins? As explained previously, the predicted epitopes may 1) not be generated naturally by cells,²⁹³ 2) lack sufficient affinity to the particular MHC of that patient, or 3) not be produced in sufficient amounts.

Mutant RAS. Point mutations in oncogenes can also be shared by several cancers and could encode useful antigens. For example, a valine for glycine substitution at position 12 of RAS is one of the most common mutations in human cancers and can be recognized by human CD4+ T cells.³³³ The region of the mutant RAS protein from which
the peptide was derived is identical for all the three members of the RAS proto-oncogene family, namely, H-RAS, K-RAS, and N-RAS, which have different prevalence in different cancers. In addition, about 90% of pancreatic adenocarcinomas, a very aggressive human cancer, have one of three to four different single amino acid substitutions in codon 12 of the cellular Kirsten RAS gene.³³⁴ Despite remarkable clinical and experimental efforts,^{335,336,337,338,339,340} targeting mutant RAS-derived epitopes has so far met with little if any success, the likely reasons being the same as those mentioned under the section titled Fusion Proteins. Immunizing cancer-prone mice harboring “initiated” mutant ras-expressing cells with mutant ras peptide resulted in mutant ras-specific CD4+ T cells, antibodies to the mutant ras protein, and more tumors and faster tumor development.³⁴¹ Possibly, the antibody, produced in response to antigen, activated myeloid cells via FcR-γ to become tumor-promoting, and/or antibody carrying TGF-β downregulated CD8+ effector responses.^342,343

Growth Factors and Their Receptors. See Immunotherapy.

Survivin. Although survivin is overexpressed in many cancer cells, it is a problematic antigen particularly as a vaccine because it is widely expressed on lymphocytes causing fratricide.³⁵¹ Not surprisingly, this antigen has been ineffective in vaccine trials.

p53. Research using p53 as target for immunotherapy is exemplary of the problems and persistent misconceptions of research on tumor-specific (unique or shared) and selfantigens. Mutations in the p53 suppressor gene are among the most common found in human and experimental cancers.^353,354 These mutations tend to cluster in evolutionarily conserved regions of the gene, but the exact locations are highly diverse in individual cancers.³⁵⁵ Targeting these mutations would require an individualized therapy considered by many impractical. Therefore, researchers have attempted to exploit the fact that mutant p53 is usually overexpressed in cancers. Thus, major efforts have been made to target conserved, nonmutated regions of the p53 protein hoping for “relative tumor specificity” based on the usual belittlement for low-level expression of this protein by every normal cell including lymphocytes. As a result, T cells expressing highaffinity anti-p53 TCRs commit fratricide unless the anti-p53 TCRs are transduced into T cells from p53 knockout mice.
However adoptive transfer of these T cells causes lethal hematopoietic ablation in normal mice.³⁵⁰ Mortality can be prevented by reconstitution of the recipients with bone marrow from p53 knockout mice,³⁵⁰ an option unavailable in humans. Not surprisingly, there is no evidence for clinical efficacy of targeting wild-type p53 in humans.

Prostate-Specific Membrane Antigen. Prostate-specific membrane antigen (PSMA) is a type II transmembrane glycoprotein more correctly referred to as glutamate carboxypeptidase II. PSMA is an inappropriate misnomer because it is anything but specific for prostate; it is also expressed by duodenal mucosal cells, proximal renal tubule cells, and a subpopulation of neuroendocrine cells in colonic crypts,³⁵⁶ and has important functions in the brain.³⁵⁷ Vaccination of mice with human PSMA induces antitumor effects without causing toxicity but because the mouse model lacks expression of human PSMA on normal tissues, the findings in this inappropriate model are meaningless.³⁵⁸ Similarly, a “fully humanized” anti-PSMA antibody had potent antitumor xenograft activity in mice that lacked expression of this antigen on normal cells.³⁵⁹ Therefore, this antibody may have severe, possibly lethal, toxicity in patients.

Some antigens expressed on tumor cells are also expressed during at least some stage of differentiation on nonmalignant cells of the lineage from which the tumor developed. Differentiation antigens may therefore help to determine the organ or cell type of origin (lineage) of a cancer.³⁶⁰ For example, B-cell tumors express surface Ig, and T-cell leukemias can be separated into helper and suppressor cell leukemias using T-cell subset-specific monoclonal antibodies to surface and intracellular antigens including transcription factors. For metastatic cancer of unknown origin, differentiation antigens can be important indicators of the histologic and organ site of the primary tumor.³⁶¹ Careful diagnostic delineation of different subtypes of cancer is important because different tumor subtypes may have different prognoses and may be susceptible to different therapies. However, the use of differentiation markers for histologic or cytologic tumor classification has pitfalls.³⁶¹ Lineage-specific antigens represent a very diverse group of proteins: glycoproteins (including mucins) and glycolipids (carbohydrate, peptide, glycopeptide, or glycolipid epitopes). Several of these antigens are being explored as potential immunotherapeutic targets. However, the normal cell types that express these antigens will determine toxicity and usefulness of any antibody or agent for which these are targets.

Cluster of Differentiation 20 and Cluster of Differentiation 19. CD20 is a signature B-cell differentiation antigen targeted by a genetically engineered monoclonal antibody that is relatively effective in the treatment of B-cell lymphoma. Lineage restriction of this marker limits the cytocidal effects to long-term depletion of normal B cells. This depletion is usually well tolerated because patients can be protected by intravenous administration of IgG. While antibodies to CD20 antibodies are therapeutically effective, relapse is common. CD19, but not CD20, appears to be expressed on the more immature malignant cells causing the relapse. Therefore, anti-CD19 Fv has been used in the killing of these cancer cells, either as a fusion protein with an anti-CD3 Fv to engage T cells¹⁹⁵ or as chimeric antibody receptor inserted into the patient’s T cells.¹⁹⁶ These treatments may prevent relapse.

Melanocyte-Specific Differentiation Antigens. Several differentiation antigens (such as tyrosinase, the related brown locus protein or tyrosinase-related protein 1 (Trp-1), gp100, and Melan A/MART-1^362,363) appear to be restricted to melanocytes, and all of them are being explored as immunotherapeutic targets in melanoma.^364,365,366 Immune recognition of the melanocyte differentiation antigens can lead to rejection of a tumor challenge but this self-antigen-specific immunotherapy not only targets the tumor cells but also normal cells expressing the shared antigens,^365,367 resulting in the depigmentation of normal skin (vitiligo) and possibly other, more serious toxicities (see following discussion). For example, mice immunized with syngeneic Trp-1 in various adjuvant settings fail to produce a T- or B-cell response.³⁶⁸ Xenogeneic or altered Trp-1, however, induces responses that cause vitiligo and protect against challenge with melanoma cells.³⁶⁸ Passive transfer of Trp-1-specific antibodies from these mice causes vitiligo and protects against metastatic spread of melanoma cells in mice when given at the time of seeding of the malignant cells.³⁶⁷ Similar antibodies may have analogous beneficial effects in human melanoma patients,³⁶⁹ but evidence is lacking for such antibodies eliminating bulky human melanoma. Trp-1-deficient mice respond to Trp-1 as a foreign antigen and Trp-1-specific CD4+ T cells from these mice can eradicate large B16 melanoma 10 to 14 days after inoculation,^370,371 cause autoimmune vitiligo, and damage the retina.³⁷⁰ Designating these T cells as “tumor-specific” when they are clearly not is incorrect and misleading.³⁷⁰ Trp-1 is expressed in all neurocrest-derived pigmented cells, not only those of the skin but also those of the eye (uvea), inner ear, and brain (substantia nigra, forebrain, and midbrain).^{372,373,374,375} Other targets include gangliosides GD2 and GD3 that are also not only overexpressed in melanoma but are also found in other cells of neurocrest origin and in other tissues.^376,377

Prostate-Specific Antigen. Prostate-specific antigen (PSA), also called kallikrein-3, is a chymotrypsin-like protease that digests semenogelin I and II to release motile sperm.³⁷⁸ Elevation of PSA above the normal range occurs in inflammation (prostatitis) and benign hypertrophy of the prostate as well as in prostate cancer. Using PSA levels for early detection of prostate cancer is controversial. The U.S. Preventive Services Task Force no longer recommends this test in healthy man.³⁷⁹ However, detection of any PSA following complete surgical removal of the prostate indicates residual tumor cells and/or recurrence.³⁸⁰

Prostate-specific phosphatase is selectively expressed and secreted by the epithelial cells of the prostate gland.^378,381 Sipuleucel-T, a vaccine targeting prostate-specific phosphatase
and approved by the U.S. Food and Drug Administration for the treatment of castration-resistant prostate cancer, results in only a very modest improvement in overall survival.³²³ Furthermore, a recent analysis of internal documents that became available after the U.S. Food and Drug Administration approval questions the adequacy of the trial and correctness of the conclusions drawn.³⁸²

Epithelial Cell Adhesion Molecule. Normal and malignant cells of epithelial origin express the transmembrane glycoprotein epithelial cell adhesion molecule (Epcam).³⁸³ Anti-Epcam antibody 17-1A³⁸⁴ showed promising results in patients with colon cancer in several early studies,³⁸⁵ whereas subsequent randomized clinical trials consistently failed to show any benefit.³⁸⁶ A high-affinity engineered anti-Epcam antibody caused cases of acute pancreatitis and was discontinued.³⁸⁷ Other anti-Epcam antibodies in the form of bi- and trifunctional constructs may be more effective, but finding efficacy while keeping lethal toxicity under control presents a major hurdle.

Aberrant glycosylation and the overexpression of certain carbohydrate moieties is a consistent feature of cancers,^388,389 and tumor-associated oligosaccharides are actively investigated as targets for immunotherapy.

Mucins. Mucins are normally heavily glycosylated glycoproteins (ie, containing complex O-glycans) that protect the luminal mucous epithelial surfaces. Mucins of cancer cells often show decreased expression of the complex O-glycans and increased expression of short oligosaccharides, the TF, sialyl-Tn, and Tn antigens³⁹⁰ (see following discussion). Thus, human adenocarcinomas of the pancreas, breast, and colon express mucins³⁹¹ that can be recognized on cancer cells by MHC-unrestricted cytolytic T cells. These T cells apparently react specifically with repeated epitopes on the protein core of the mucin molecules,³⁹² exposed because of deficient glycosylation in the malignant cells.³⁹³ The epitopes are expressed at high density, do not require processing, have a stable conformation, and can directly bind to certain TCRs without being presented by MHC molecules.³⁹⁴ Hypoglycosylated MUC1 is expressed on about two-thirds of newly diagnosed cancers. However, multiple types of vaccines using the nonglycosylated tandem repeat peptides or tumor-associated saccharide antigens conjugated to carriers failed to immunize effectively.^395,396 This finding is consistent with multiple lines of evidence that humans and MUC1-transgenic mice are tolerant to the unglycosylated long MUC1 peptide. Numerous approaches have been developed capable of overcoming this tolerance such as using short and long synthetic glycopeptides or plant-expressed MUC1.^{397,398,399,400} Many different mucin genes have been identified⁴⁰¹ and antigens have been defined, but evidence for clinical efficacy with controllable toxicity is lacking.

T and Tn Antigens. Tn antigen on human erythocytes, first described by Moreau et al. in 1957,⁴⁰² is the cause of a hemolytic autoimmune disorder, Tn syndrome.⁴⁰³ Tn (“T antigen nouvelle”) is a Ser/Thr-O-linked N -acetylgalactosamine monosaccharide distinct from the disaccharide T (or TF) antigen galactose-β1-3-N-acetylgalactose O-linked to a Ser or Thr (Galβ1-3GalNAcα1-Ser/Thr) described earlier by Huebner, Thomsen, and Friedenreich.⁴⁰⁴ Human cancers very frequently express TF and Tn antigens.^405,406 Recently it was shown that the Tn syndrome is caused by somatic mutations in the chaperone COSMC.⁴⁰⁷ While TF antigen is an oncofetal antigen highly expressed in the embryo and fetus,³⁹⁰ there is no evidence that Tn is also an oncofetal antigen.⁴⁰⁴ Mutational deletion of the chaperone COSMC leads to expression of the Tn antigen on virtually every embryonic cell and leads to early embryonic death.⁴⁰⁸ Most adults naturally have anti-Tn as well as anti-TF antibodies due to antigenic stimulation by Tn and TF antigens expressed on the bacterial flora.^409,410 Tn antigen is also expressed on human immunodeficiency virus-1 and pathogenic parasites. Thus, evidence is lacking for any expression of Tn antigen on adult or embryonic human or murine cells except by patients suffering from Tn syndrome and cancers that may or may not show a tumor-specific deletion of COSMC. There is no evidence that Tn is an effective therapeutic target. This may be due to anti-Tn antibodies being usually IgM or IgA of very low affinity, though recently developed IgG antibodies reduced slightly the growth rate of human cancer cells in vitro and in vivo.⁴¹¹

Glycopeptide Antigens Resulting from Tumor-Specific COSMC Mutations. The remarkable characteristic of this new class of antigens is the exquisite tumor specificity of the antigen^412,413,414 even though it is encoded by normal genes. Appearance of the antigen, however, depends on a tumor-specific somatic mutation that destroys the chaperone COSMC that is essential for any functioning of the core 1β (1-3) galactosyl-transferase, T-synthase. COSMC protects the newly synthesized T synthase from aggregation and subsequent endoplasmic reticulum-associated degradation.⁴⁰⁴ T-synthase is essential for extending O-linked glycosylation beyond a single O-linked N-acetylgalactosamine (ie, Tn antigen). Importantly, the antigen recognized by the high-affinity, tumor-specific antibody does not bind Tn alone but contains the Tn hapten (ie, the single O-linked N– acetylgalactosamine on a threonine or serine). X-ray crystallography shows that the antibody completely envelops the carbohydrate moiety while interacting with the unique sequence of the peptide moiety in a shallow groove.⁴¹³ Because Tn does not exist on normal embryonic or adult human or murine cells, there is no central or peripheral tolerance to these antigens allowing high-affinity destructive immune reactions to occur. This also explains the severity of the autoimmune disease associated with the Tn syndrome caused by somatic (not germline) mutations of the COSMC gene.⁴⁰⁷ Importantly, somatic tumor-specific mutations disabling the X chromosome encoded COSMC gene (for which one copy is naturally silenced) seem to be more common than originally assumed and occur in a wide variety of spontaneous or virally induced human and murine cancers.⁴¹⁵ Nevertheless, it must be shown that high-affinity receptors when used as CARs on T cells or linked to anti-CD3 are effective against tumors but have little if any toxicity to
normal tissues. This may be expected from the fact that the epitopes depend on somatic tumor-specific mutations.

Over 50 years ago, human cancer cells were found to express antigens that serologically cross-react with normal embryonic tissue.⁴¹⁶ Since then, it has been postulated repeatedly that certain normal genes that are completely silent in all nonmalignant cells may be activated exclusively in malignant cells.⁴¹⁷ Alternatively, it has been postulated that cancer cells may express proteins (or immature forms of a protein) that are only expressed in fetal but not in nonmalignant adult cells.⁴¹⁸ There certainly is profound hope for finding a universal tumor rejection antigen that can be used for vaccination, prevention of cancer, and antitumor therapy. However, repeated claims of selective activation of normal genes or selective expression of immature forms of these proteins in cancer cells leading to tumor-specific antigens have not been substantiated. Expression of the same antigen by at least one normal cell type in the adult was later often discovered.^{419,420,421,422}

Cancer-Germline/Cancer-Testis Antigens. Changes in gene methylation lead to the expression of a large number of genes encoding a group of antigens referred to “cancer-testis,” “cancer-germline,” or “cancer-spermatogonal” antigens. Many cancer-testis antigens have been described in humans such as MAGE, BAGE, GAGE, LAGE/NY-ESO-1, SAGE, HAGE, and BORIS.^{422,423,424,425,426,427,428,429,430} All cancer-testis antigens are expressed at high levels in spermatocytes in the testis. And most, if not all, are also expressed in the thymus.^430,431,432 Many, but not all, cancer-testis antigens are encoded on the X chromosome.⁴³³ Most of cancer-testis antigens can be recognized by autologous cytotoxic T-lymphocyte (CTLs), but there is no evidence for these antigens being generally effective targets for cancer cell destruction in human.^434,435 P1A was the first cancer-testis antigen to be identified in the mouse mast cell tumor line P815.^436,437 P1A can be induced in a broad range of tumors of diverse histologic origins with the demethylating agent 5-aza-2′-deoxycytidine.⁴³⁸ Various normal tissues including thymus and premeiotic spermatocytes also express P1A.^437,439 While active immunization has only modest protective effects against cancer cell inoculation,⁴⁴⁰ adoptively transferred T cells (monoclonal or polyclonal) targeting only the P1A antigen shrink large tumors (>1 cm in diameter) for several weeks followed by relapse caused by epitope-loss variants.⁴⁴¹ For eradication, cancer cells must first be transfected to express costimulatory molecules, which probably leads to the induction of T cells to recognize other tumor antigens.⁴⁴² Remarkably, autoimmunity has not been reported to be associated with a response to this antigen. In humans, the HLA-A2-restricted NY-ESO1 peptide has very poor affinity to its presenting MHC class I molecule, but high-affinity TCRs to this antigen have been generated and their efficacy in therapy is being explored.⁴⁴³ While many cancer-testis antigens have been described and studied extensively for two decades, their usefulness as targets eradicating human cancers still remains to be shown. Despite these uncertainties, occasional successes have been observed. The hypothetical explanation: the relatively few T cells induced against the cancer-testis antigen served to elicit additional antitumor T-cell clones directed against other antigens that are actually responsible for the tumor regression by a poorly understood phenomenon referred to as epitope spreading.^434,444

Numerous cancer-testis antigens have been detected using sera of cancer patients for identifying antigens by recombinant expression cloning (SEREX). Importantly, sera from mice infected with cytopathic or noncytopathic viruses or injected with tumor cell lysates also show an autoantibody response of broad specificity, and intriguingly the majority of the identified autoantigens have been previously described as autoantigens in humans.^445,446 This suggests that human SEREX antigens may have to be regarded as afterglows of infection-associated immunopathology and/or tissue damage. This is consistent with the assumption that the human adult IgG autoantibody repertoire is the result of lifelong encounters with bacterial and viral agents and tissue damage. Together, usefulness of the SEREX-defined selfantigens remains to be demonstrated.

Carcinoembryonic Antigen. CEA is a 200-kDa membraneassociated glycoprotein that is expressed not only in fetal but also in adult nonmalignant tissues such as normal colonic mucosa, lung, and lactating breast tissue.^447,448 It is released into surrounding fluids. At one time, it was hoped that CEA could be used as a marker for early diagnosis of gastrointestinal and other malignancies⁴⁴⁹; however, elevated serum levels of CEA are also found in the absence of malignancy (eg, in smokers and in inflammatory bowel diseases such as ulcerative colitis). Though serum levels of CEA are not useful for detecting early cancer, the level of CEA in the blood can be used to monitor the effects of therapy to indicate whether a cancer has been successfully eradicated or has recurred.⁴⁵⁰ Using CEA as immunotherapeutic target is highly problematic because of severe toxicity. Mice expressing the target antigen in normal tissues closely resembling expression in humans showed severe toxicity from anti-CEA T-cell responses,²⁰⁵ a result that predicted the severe adverse effects later observed in patients treated with anti-CEA CARs.²⁰⁶

Alpha-Fetoprotein. Alpha-fetoprotein (AFP) was the first defined oncofetal protein.⁴⁵¹ Though produced by fetal liver and yolk sac cells, it is also present in small amounts in cells and the serum of normal adults. The amount of this protein is elevated in some patients with cancer of the liver or testis and also in some patients with various nonmalignant liver diseases. Therefore, similar to CEA, using AFP as a marker for the early diagnosis of cancer is of questionable use.⁴⁵² Nevertheless, assays of AFP can detect primary liver cancer at a time when the cancer is treatable, and AFP assays are also used for monitoring patients after therapy.⁴⁵² The use of AFP as target for active or passive immunotherapy is complicated by lack of specificity.^417,453

Clonal antigens are expressed only on the clone of cells from which the cancer originated.⁴⁵⁴ Except for idiotypes of surface Ig-positive B-cell and T-cell malignancies, there
are presently no candidates for other clonal antigens. Immunization of animals against the idiotype induces an idiotype-specific transplantation immunity against the growth of cancers (myeloma, lymphoma, and leukemia) expressing the idiotype.^455,456 Furthermore, idiotype-specific antibodies prevented the growth of surface idiotype-positive murine malignant B cells in mice and guinea pigs in vivo and in vitro,^{457,458,459,460} and occasionally induced the cancers to go into a long-lasting dormant state.⁴⁶¹ Finally, idiotypespecific monoclonal antibodies have caused several partial remissions and one complete remission in patients with B-cell lymphoma,⁴⁶² and targeting the idiotype on B-cell malignancies continues to be explored.^463,464,465 It is necessary to generate different monoclonal antibodies or idiotypic vaccines for each individual cancer to be treated, but the advantage of using such clonal antigens over less restricted tumor-associated antigens is that eliminating the few normal cells bearing the same antigen would probably not adversely affect the patient.

Clinical support comes from the appearance of highly antigenic EBV-associated lymphomas in immunosuppressed transplant recipients that are virtually unknown to occur in immunocompetent humans.⁵⁰² In addition, renal transplant patients experience a reduced risk of developing UV-induced skin cancer after immunosuppressive medications are stopped.⁵⁰³ It is tempting to suggest that many of these
cancers developing in transplant patients are actually “regressor” tumors that have arisen only because host defenses were damaged.

In normal adult mice, Moloney sarcoma virus⁴⁹⁶ induces tumors larger than 2 cm in diameter that then regress while tumors continue to grow and kill immunodeficient adult or newborn mice.^497,498,499 The first experimental evidence for UV-induced regressors came from Kripke in 1974,²⁷⁵ showing that UV irradiation made mice immunodeficient and induced cancers that were often “regressors,” meaning that they were rejected with any testable size inoculum by naïve syngeneic immunocompetent mice. Tumor transplants will grow for about a week and then disappear, though small numbers of the same tumor cells will grow and kill athymic nude mice. The same regressors grew after inoculation of few cancer cells in UV-irradiated mice. So, it may be highly relevant that some carcinogens are immunosuppressive.^500,501 Repeated exposures of mice to UV induce persistent immune suppression,^502,503,504 leading to the development of highly immunogenic regressor tumors. In contrast, the single injection of MCA induces tumors in 100% of mice and only a short-lived state of immune suppression,⁵⁰¹ thus allowing a fully immunocompetent host to select for less immunogenic variants. The concept that immunocompetence of a host influences the immunogenicity of the developing tumor has been tested experimentally by Roberts and Daynes⁴⁸⁰ comparing the MCA-induced tumors occurring in immunodeficient UV-irradiated mice with cancers induced in normal mice. Indeed, cancers induced with MCA in mice immunosuppressed with UV were frequently regressor tumors, whereas none of tumors induced with MCA in immunocompetent mice were regressors.⁴⁸⁰ Decades later, this concept was confirmed with tumors induced with MCA in nude, severe combined immunodeficiency, or Rag-/- mice.^487,505,506

Immunocompetent hosts can select for cancer variants. Thus, when regressors are transplanted into normal immunocompetent hosts, heritable progressor variants can escape. Some of these variants show antigen loss,^{218,219,507,508,509} but many others retain their antigenicity and grow faster than the parental tumor when their growth is compared in T, B, or NK cell-deficient hosts,^{43,45,218,307,505,508,510} suggesting that mechanisms other than antigenicity must participate determining the differences in growth behavior between regressors and progressors.