Immune selective pressure and HLA class I antigen defects in malignant lesions

Abstract

The revived cancer immune surveillance theory has emphasized the active role the immune system plays in eliminating tumor cells and in facilitating the emergence of their immunoresistant variants. MHC class I molecule abnormalities represent, at least in part, the molecular phenotype of these escape variants, given the crucial role of MHC class I molecules in eliciting tumor antigen-specific T cell responses, the high frequency of HLA class I antigen abnormalities in malignant lesions and their association with a poor clinical course of the disease. Here, we present evidence for this possibility and review the potential mechanisms by which T cell selective pressure participates in the generation of tumor cells with MHC class I molecule defects. Furthermore, we discuss the strategies to counteract tumor cell immune evasion.

Introduction

It has been more than three decades since the first few reports of HLA class I antigen defects in malignant cells were published [1–4]. In the following years, a number of studies (for review, see ref. [5]) to define the molecular mechanisms underlying these defects, their functional significance and their clinical relevance have been stimulated by at least three reasons. They include (1) the major emphasis on T cell-mediated cytotoxicity as a major mechanism to control tumor growth [6–8], (2) the role of HLA class I antigens in the interactions of tumor cells with HLA class I antigen-restricted, tumor antigen (TA)-specific cytotoxic T lymphocytes (CTLs) [9, 10], and (3) the expectation of a negative impact of HLA class I antigen defects in malignant lesions on the course of the disease and on the outcome of T cell-based immunotherapy [5]. As a result, more than 700 (PubMed search with “MHC class I” and “tumor” as keywords) papers on these topics have been published. These studies have shed light on the frequency of HLA class I antigen defects in various types of tumors, on their underlying molecular mechanisms and on their impact on interactions of malignant cells with immune cells in vitro [5]. Although a statistically significant association of HLA class I antigen defects with the clinical course of the disease has been demonstrated in at least some tumor types, the biological and clinical significance of this tumor phenotype remains under debate [11]. The debate has focused on two issues. First, whether HLA class I antigen abnormalities, which are frequently found in malignant cells, are simply the by-product of genomic instability or reflect clonal selection as a result of HLA class I antigen-restricted, TA-specific CTL-mediated immune selection. Second, whether HLA class I antigen defects in malignant lesions play a role in the clinical course of the disease and have an impact on the outcome of T cell-based immunotherapy.

A significant body of new information about the role of immune surveillance in the control of tumor growth in the host has been recently published [12]. The revival of the cancer immune surveillance theory has led to a reevaluation of the role of immune selective pressure in the generation of malignant lesions with HLA class I antigen defects. In this article, we review this topic, since this information may contribute to the optimization of the design of immunotherapeutic strategies for the treatment of cancer. Specifically, we first summarize the recent results, which, although conflicting, have revived the immune surveillance theory and its potential clinical relevance. Second, we describe the types of HLA class I antigen defects that have been identified in malignant cells, as well as some emerging in vitro and in vivo evidence for the role of T cell selective pressure in the generation of malignant lesions with HLA class I antigen defects. Third, we discuss how the host immune system adapts to the generation of malignant lesions with HLA class I antigen defects. Lastly, we propose strategies one could apply to prevent the generation of malignant lesions with HLA class I antigen defects and/or to counteract their potential effects on the clinical course of malignant diseases and on the outcome of T cell-based immunotherapy.

Cancer immune surveillance theory

The original cancer immune surveillance theory was proposed by Thomas and Burnet in the 1960s [13, 14]. It was based on the requirement for malignant transformation of cells of somatic mutations in genes that potentially create new antigen epitopes on the surface of tumor cells. The latter provide T cells with markers for the recognition of tumors and with targets for their elimination. A corollary of this theory is that dysfunction of the host immune system is associated with an increased tumor incidence. However, this theory fell out of favor when Stutman [15, 16] demonstrated that nude mice, which do not have a functional thymus and lack thymus-dependent T cells because of a spontaneous mutation, develop neither spontaneous nor chemically, e.g. methylcholanthrene, MCA-induced tumors faster or more frequently than their control littermates. In humans, tumor incidence was reported to increase only in certain types of malignancies in individuals with spontaneous or drug-induced immunological dysfunctions [17]. In the last decade, the cancer immune surveillance theory has been revived by the higher tumor incidence in IFNGR1^−/− [18], IFN-γ^−/− [19], perforin (pfp)^−/− [19–22], RAG-2^−/− [23], severe combined immunodeficiency (SCID) [24], STAT1^−/− [18], TCRJα281^−/− [24], TCRβ^−/− [25] and TCRδ^−/− [25] mice than in control mice following injection of MCA. More recently, the tumor immune surveillance theory was also demonstrated in a MCA-induced mouse tumor model with impaired function of NKG2D, an activating receptor involved in both innate and adaptive immune responses [26]. The discrepancy between these results and those obtained in nude mice has been explained by the existence in the latter mice of residual T and natural killer (NK) cells, although their functionality has not been convincingly shown. Notably, in the IFNGR1 knock-out and RAG-2 knock-out experiments, Schreiber and his collaborators [23] have also demonstrated that tumors derived from RAG-2^−/− mice could be less efficiently transplanted into the wild-type mice compared with those derived from the wild-type mice. Schreiber and his colleagues [27] took this finding as the proof of selective elimination of less-immunogenic variants by T cells and proposed the theory of three Es (elimination, equilibrium, escape) of cancer immunoediting. This view was challenged by Qin and Blankenstein [28] who failed to reach the same conclusion in a similar experimental setting, although the latter investigators [29] do not question the protective role of IFN-γ/IFNGR against MCA-induced tumors, and the increased frequency of spontaneous tumors in IFN-γ^−/− mice. The mechanisms underlying the role of the IFN-γ/IFNGR system in controlling tumor growth are thought to be both immunologic and nonimmunologic. Among the former ones is tumor initiation facilitated by pathogen-induced chronic inflammation in the absence of IFN-γ [30] besides its well-known immuno-stimulatory properties. The nonimmunologic mechanisms include the effect of IFN-γ on nonhematopoietic tumor stroma cells and the induction of angiostasis as well as of a foreign body reaction, all of which contribute to the suppression of tumor growth in experimentally transplanted or MCA-induced tumor models [29, 31].

The molecular phenotype of the less-immunogenic tumor variants was shown by Schreiber and his colleagues [23] to be associated with down-regulation of TAP1, one of the antigen processing machinery (APM) components responsible for loading peptides onto MHC class I molecules and induction of T cell responses. This finding supports the hypothesis that T cell-mediated immune selective pressure facilitates the emergence and outgrowth of tumor variants with impaired MHC class I molecule expression and/or function. This possibility is supported by the high frequency of HLA class I antigen abnormalities in malignant lesions as well as by evidence derived from in vitro and in vivo experiments performed with mouse and human malignant cells.

HLA class I antigen defects in malignant cells

Immunohistochemical staining of a large number of surgically removed malignant lesions with HLA class I antigen-specific monoclonal antibodies has convincingly documented an association between malignant transformation of cells and HLA class I antigen defects in most, if not all types of solid tumors, although with marked differences among various types of tumors [5, 32]. These abnormalities include loss or down-regulation of all or certain HLA class I allospecificities on cells and malfunctions of APM which lead to defective expression of HLA class I–peptide complexes [5, 32]. The molecular mechanisms underlying these abnormalities range from mutations in the genes encoding HLA class I antigen subunits, such as HLA class I heavy chains and/or β₂-microglobulin (β₂m), to down-regulation of APM components, such as proteasome subunits, TAP subunits and tapasin, which are responsible for generating and loading peptides onto HLA class I molecules [5, 32]. Given the crucial role of HLA class I antigens in presenting TA-derived peptides to T cells and in eliciting TA-specific T cell responses, HLA class I antigen abnormalities in malignant lesions provide a mechanism for tumor growth in an immunocompetent host and for the association between defects in HLA class I antigen expression and/or function in primary malignant lesions and poor prognosis of the disease in some tumor types. The only exceptions are represented by malignancies in which the major role in the control of tumor growth is believed to be played by NK cells and not by HLA class I antigen-restricted, TA-specific CTLs [33–35].

Mechanisms underlying generation of tumor cell populations with HLA class I antigen defects

Experiments conducted in vitro and in vivo have shown that at least two events play a role in the generation of malignant lesions with HLA class I antigen defects. One is represented by a mutation that causes a defect in the expression and/or function of HLA class I antigens in a tumor cell in a tumor cell population. The other one is represented by immune selective pressure, which favors the outgrowth of malignant cells with HLA class I antigen defects so that they become the predominant population in a tumor cell population. At a time when T cell-based immunotherapy is increasingly applied to treat cancer, immune selective pressure can come not only from the host spontaneous immune response evolved along with tumor development and disease progression, but also from the immune response elicited by active immunotherapy and/or from the introduction by adoptive transfer of immune components into the treated patient. It is not known at present if and to which extent these two types of immune selective pressure are mechanistically different. Below we summarize evidence in vitro and in vivo for the two-event immune selection hypothesis.

In vitro evidence

Repeated exposure in vitro of the irradiated HLA-A2-positive melanoma cell line SK-MEL-29.1 to HLA-A2-restricted, melanoma associated antigen (MAA)-specific CTLs has resulted in the isolation of the clones SK-MEL-29.1.22 and SK-MEL29.1.29, both of which had lost HLA-A2 antigen [9]. Different mutations in the HLA-A2 gene underlie the HLA-A2 antigen loss in the two-melanoma cell lines. In SK-MEL-29.1.22 cells, the 5′-flanking region, exon 1, intron 1, and a region at the 5′ end of exon 2 of the HLA-A2 gene are deleted [36]. The breakpoint of the HLA-A2 gene, which is recombined with a DNA fragment of unknown origin, was localized between two GTTCG sequence repeats at position 101 of exon 2 [36]. These repeats may provide the basis of sequence-specific misalignment in the process of DNA deletion. In SK-MEL-29.1.29 cells, loss of HLA-A2 allele as well as of HLA-B44 and HLA-Cw5 alleles, is caused by the loss of one copy of chromosome 6 [36].

In vivo evidence in animal model systems

A number of studies in mice also support immune selection as the mechanism underlying the generation of immunoresistant variants with MHC class I antigen defects. Harding’s group has shown that immune selective pressure targeting H2-restricted CTL epitope can isolate tumor cells lacking the targeted epitope by failing to express the MHC-anchored peptide from a tumor cell population [37]. Using matched panels of TAP1-positive and TAP1-negative tumor cell lines generated from a parental transformed murine fibroblast line, Harding and co-workers demonstrated that both TAP1-positive and TAP1-negative cells produce tumors in athymic mice, while only TAP1-negative cells form large and persistent tumors in the immunocompetent autologous C57BL/6 mice [37]. Moreover, inoculation of C57BL/6 mice with mixtures of TAP1-positive and TAP1-negative cells produced tumors composed exclusively of TAP1-negative cells [37]. These data suggest that the selection pressure applied by MHC class I antigen-restricted, TA-specific CTLs favors the outgrowth of cells with defective presentation of TA-derived peptides because of TAP defects. In other words, the tumor’s MHC phenotype has been “immunoedited” in the course of the disease, resulting in the survival of tumor variants with defective presentation of TA-derived peptides by MHC class I antigens.

In studies that are in progress in our laboratory, we have taken advantage of the melanoma cell line T372A with a selective HLA-A2 antigen loss because of a deletion in the 5′ flanking region of the HLA-A2 gene to study the role of selective pressure in the generation of lesions with HLA-A2 antigen defects. An autologous cell line with reconstituted HLA-A2 antigen expression was established following transfection with a wild type HLA-A2 cDNA. Administration to a SCID mouse of a mixture of the HLA-A2 antigen-negative melanoma cell line and of autologous melanoma cells with reconstituted HLA-A2 antigen expression resulted in the formation of a tumor which contained both HLA-A2 antigen-positive and HLA-A2 antigen-negative melanoma cells. However, a selective growth of HLA-A2-negative melanoma cells was observed following administration of a HLA-A2 antigen-restricted, TA-specific CTL line. Similar results have been described, although in a different antigen system, by Lozupone et al. [38]. They have shown that MAA MART-1-negative melanoma cells can be selected in a SCID mouse-transplanted melanoma tumor with heterogeneous MART-1 expression following systemic treatment with a HLA-A2 antigen-restricted, MART-1-specific CD8⁺ T cell clone.

Clinical evidence

The results obtained in the above-mentioned animal studies are paralleled by those obtained in a clinical setting. Jager et al. [39] have demonstrated total HLA class I antigen loss and selective HLA-A2 antigen loss in three of five and one of five, respectively, melanoma lesions which progressed in spite of the expression of the targeted MAAs and of the presence of a MAA-specific T cell response in patients immunized with MART-1- and tyrosinase-derived, HLA-A2-binding peptides. Similar results have been described by Khong et al. [40] who have characterized melanoma metastases that recurred after an initial dramatic clinical response in a patient immunized with gp100-, MART-1-, and tyrosinase-derived peptides with HLA-A2-binding motifs emulsified in incomplete Freund’s adjuvant. T-cell clones specific to these MAAs were present in the patient’s peripheral blood as well as in the isolated tumor-infiltrating lymphocytes (TILs). In one recurrent melanoma metastasis, multiple MAAs had been lost, but HLA class I antigen expression was retained [40]. In another recurrent metastasis, HLA class I antigens were not detectable, while MAAs continued to be expressed [40]. Lastly, Restifo et al. [41] have described total HLA class I antigen loss in recurrent melanoma metastases in five patients who experienced an initial clinical response following T cell-based immunotherapy. In a subsequent collaborative study we showed that lack of HLA class I antigen expression by the melanoma cell lines derived from the recurred metastases was caused by loss of one copy of the β ₂ m gene associated with mutations in the other copy [42]. Three of the five melanoma cell lines carry the same type of mutation in the β ₂ m gene (CT dinucleotide deletion in exon 1) [42]. These results in conjunction with those of other studies have revealed a mutational hotspot that may be associated with the immune selective pressure introduced by adoptive T cell-based immunotherapy [42]. This finding implies a relationship between the modified tumor microenvironment during immunotherapy and the type of genomic instability and/or DNA repair capacity possessed by tumor cells. This model is supported by an elevated frequency of β ₂ m gene CT dinucleotide deletion mutations identified in microsatellite instability (MSI) (+) colon carcinoma lesions [43]. The β ₂ m gene mutations may be preferentially selected by T cell selective pressure in situ, since MSI (+) tumors have been found to be infiltrated by a large number of activated CD8⁺ CTLs presumably recognizing a number of neoepitopes generated by the tumor’s genomic mutator phenotype [44].

Immune selection and microevolution of tumors

On the basis of the aforementioned experimental results and clinical data, immune selection may be viewed as conceptually equivalent to the theory of evolution proposed by Darwin more than a century ago [45]. In this theory, complex creatures evolve from more simplistic ancestors naturally over time. As random genetic mutations occur within an organism’s genetic code, the beneficial mutations are preserved because they aid survival against the hardship in their habitat, a process known as “natural selection”. These beneficial mutations are passed on to the next generation. Here, if we consider an organism to be a tumor cell, mutations in a tumor cell’s genome will generate variants; some of them can survive their host microenvironment, which is meant to suppress their growth. It is important to note that the microenvironment includes not only the intrinsic tumor suppressor system, such as p53 and pRb, both of which mediate cell cycle arrest and/or apoptosis for a damaged genome [46], but also the putative extrinsic tumor suppressor system, such as the tumor immune surveillance system, where immune system is able to detect tumor cells’ abnormal phenotype and destroy them. This model implies that only mutations advantageous to tumor cell survival will accumulate and shape a phenotype that can aid tumor escape from both the intrinsic and extrinsic tumor suppressor systems. Since HLA class I antigens, which display an array of TA-derived antigenic peptides on the cell surface, play a crucial role in the control of tumor growth by CTLs in the tumor microenvironment, HLA class I antigen abnormalities in tumor cells may be envisioned as the result of immune selection advantageous to tumor cell survival in the host and, therefore, fit the Darwin’s “natural selection” model.

Host immune system adaptation to HLA class I antigen defects in malignant lesions

If immune selective pressure plays a role in the generation of malignant lesions with HLA class I antigen defects, one might ask whether and how the host immune system adapts once the selective pressure has facilitated the outgrowth of tumor cells that have lost the molecule(s) targeted by the ongoing immune response. Can the host immune system change the target(s) of its immune response? If so, do tumor cells targeted by the second immune response develop escape mechanisms (Fig. 1)? To the best of our knowledge, no study has addressed these questions with well-controlled experimental systems and in a clinical setting. Nevertheless, suggestive evidence derived from the clinical setting argues in favor of changes in the specificity of host’s immune response against the tumor cells that have developed an escape mechanism. Coulie and his associates [47] have demonstrated that the antigens recognized by the CTLs isolated from recurrent metastatic melanoma lesions are different from those recognized by the CTLs isolated from autologous primary and metastatic lesions. More recently, Slingluff and his associates [48] in a collaborative study with us have shown a shifting of CTL specificity from the HLA-A2-MART-1_27–35 complex to the HLA-A2-Tyr_369–377D complex in two sequential melanoma metastases with distinct abnormal HLA class I antigen phenotypes. The changes in the CTL specificity in the course of the disease are consistent with the presence of multiple defects in the HLA class I-associated APM in the two melanoma metastases. We have reached similar conclusions by determining whether multiple defects in the molecules involved in the recognition of melanoma cells by HLA class I antigen-restricted, TA-specific CTLs are frequent in melanoma cells [42]. Our working hypothesis is that the presence of multiple defects in a tumor cell reflects the ability of the host immune system to change its target when a resistant tumor cell population is selected. The change in the target of the immune response facilitates the outgrowth of a tumor cell population that has developed an additional effective escape mechanism. In five melanoma cell lines derived from lesions exposed to strong T cell selective pressure, we have found that three of them have multiple defects in APM and HLA class I antigen presentation pathway. One (1259MEL) of these three cell lines possesses β₂m loss, HLA-A2 antigen loss, and HLA-B and -C antigen down-regulation [42]. In light of the ability of the host immune system to reshape tumor cell populations over time, we favor the possibility that the multiple HLA class I antigen defects found in 1259MEL cells may have developed sequentially in the order of HLA-A2 antigen loss, HLA-B and -C antigen down-regulation and β₂m loss, although the first two abnormalities can also occur in the reverse order. The β ₂ m gene mutation is postulated to be a late event, since HLA class I heavy chain abnormalities would provide 1259MEL cells with a mechanism to escape lysis by CTLs that recognize immunodominant epitopes before a complete HLA class I antigen loss phenotype is acquired. On the other hand, if total HLA class I antigen loss caused by β₂m loss occurs early, abnormalities in heavy chains would not be advantageous to the β₂m-deficient 1259MEL cells, because they are not expected to be recognized by HLA class I antigen-restricted, MAA-specific CTLs. Similar to these findings, Seliger and her colleagues [49] in a collaborative study with one of us have described selective HLA-A2 antigen loss in conjunction with a loss-of-function TAP1 mutation near the ATP-binding site in a melanoma cell line. Recently, we have shown that a selective HLA-A3 antigen loss is associated with one HLA haplotype loss and a germline-originated tapasin frame-shift mutation in a metastatic melanoma cell line with HLA class I antigen down-regulation derived from a patient with progressive disease (C. -C. Chang, manuscript in preparation).

Fig. 1.

Host immune system’s adaptation to tumor immune escape and multiple HLA class I antigen defects. When a tumor cell carrying a HLA-A2, -A3, -B5, -B7 phenotype is exposed to immune selective pressure imposed by HLA-A3-restricted, TA-specific CTLs, its variants harboring a single defect in the expression of the immunodominant HLA-A3-TA peptide complexes will not be destroyed and survive. These single defects, as illustrated here, may include a selective HLA-A3 allospecificity loss, which abolishes HLA-A3-TA peptide complex expression, b APM component downregulation, which causes a general change in the presented TA peptide repertoire, c TAP1 loss, which reduces the amount of presented TA peptides on all allospecificities except HLA-A2, which can present TAP-independent peptides, and d immunoproteasome component (LMP2, LMP7 and LMP10) expression, which changes the type of peptides processed and presented on certain allospecificities. In response to these single HLA class I antigen defects, the host immune specificity changes and then targets the second immunodominant CTL epitopes, namely the HLA-A2–TA peptide complexes. When exposed to selective pressure imposed by HLA-A2-restricted, TA-specific CTLs, the tumor variants that harbor an additional defect in HLA-A2–TA peptide complex expression will emerge and become the predominant populations. The second defects added to the previous ones may include a′ and b′ HLA-A2 allospecificity loss, which abolishes HLA-A2–TA peptide complex expression, c′ β₂m loss, which results in total HLA class I antigen loss, and d′ APM component downregulation, which changes the peptide repertoire previously processed by the immunoproteasome

It is noteworthy that a cause–effect relationship between multiple rounds of immune selection and the appearance of multiple HLA class I antigen abnormalities has not been proved yet. Nevertheless, if correct, our view about the role of immune selection in the generation of a malignant cell phenotype implies that a tumor will grow only when it has developed enough escape mechanisms to avoid the range of immune responses a patient’s immune system is able to mount.

Strategies to counteract immune selection

There is increasing consensus among tumor immunologists and oncologists that defects in HLA class I antigen expression and/or function in malignant cells as well as other immune escape mechanisms utilized by them represent a major obstacle to the successful treatment of malignant diseases with T cell-based immunotherapy as well as with other types immunotherapy [50]. This limitation is compounded by the likelihood that immune evasion mechanisms, or at least some of them, are likely to be accelerated in their appearance and/or enhanced in their frequency by the selective pressure imposed by the application of immunotherapy. As a result, in recent years the development of strategies to counteract or to prevent immune escape mechanisms utilized by tumor cells has been emphasized. Encouraging results have been obtained in animal model systems by combining T cell-based immunotherapy with other effector mechanisms that do not require functional HLA class I antigen expression by tumor cells to control tumor growth. The latter include cytokines [51], antibody-based immunotherapy [52] and chemotherapy [53]. However, the latter approaches, like T cell-based immunotherapy, suffer from the limitation of generating immunoresistant or chemoresistant variants, when they are applied individually. Therefore, there is the concern that the long-term application of these combination therapies may result in the isolation of tumor cells that can escape the multiple destructive mechanisms because of the accumulation of escape mechanisms. Whether this limitation can be overcome by targeting with antibodies molecules that are essential for tumor cell survival and/or growth remains to be determined, although some encouraging results have been obtained. In this regard, administration of Herceptin^®, a humanized anti-HER2/neu monoclonal antibody, in patients with HER2/neu-positive breast carcinoma (approximately 30% of all breast cancer cases) has demonstrated an effect on tumor regression and survival prolongation in stage IV diseases and more recently in earlier stages of the disease [54, 55].

On the other hand, less attractive are approaches that rely on multiple TAs to target tumor cells with T cell-based immunotherapy. This approach is likely to be effective in counteracting immune evasion caused by mutation(s) that result(s) in the loss of the targeted epitope or TA and/or of a single HLA class I allospecificity, if the utilized TAs are presented by distinct HLA class I allospecificities. However, use of multiple TAs as targets will not be effective when mutations or dysfunctions in the APM impair the ability of tumor cells to process TAs and/or to present TA-derived peptides to CTLs and when multiple or all HLA class I allospecificity expression and/or function is abnormal.

Another group of published strategies have in common the principle to target cells that, at variance with tumor cells, do not have genomic instability and, therefore, are less likely to develop escape mechanisms. One strategy aims at destroying normal cells that are required for neoangiogenesis; their destruction blocks blood supply to tumors and eventually inhibits tumor growth. In this regard, Folkman and colleagues [56] have demonstrated that inhibition of tumor angiogenesis by endostatin, an endogenous inhibitor of endothelial cell proliferation, can limit tumor growth in mice. This approach has also been applied to patients [57, 58]. Similarly, Schreiber and his associates [59] have shown that destruction by CTLs of stroma cells, which express high levels of MHC class I–TA peptide complexes at the tumor site, can cause a bystander killing of TA-loss tumor variants, thus preventing their escape from the host immune responses. Whether the same concept of using CTLs to target genetically stable, tumor vasculature-associated endothelial cells can work as well remains to be determined.

An alternative approach takes advantage of the changes, which have been documented with molecular and histopathological techniques, in cells that have already occurred in early stages of tumorigenesis. The antigens that are differentially expressed between normal cells and premalignant cells provide useful targets for the selective elimination, with immunological approaches, of transformed cells at their earliest appearance at a disease stage when defects in HLA class I antigen expression and/or function are rarely detected and, therefore, unlikely to counteract the efficacy of T cell-based immunotherapy. The validity of this strategy is indicated by a series of studies by Forni and his colleagues. They [60] have demonstrated that, in a HER2/neu transgenic mouse model, application of HER2/neu-based vaccines before the time point of spontaneous tumorigenesis is effective in preventing the emergence of mammary tumors. This information provides the background and rationale to test immunoprevention in patients with hereditary tumors and in subjects with a strong family history [61].

Conclusion

Our understanding of how the immune system interacts with tumor cells at the cellular and molecular levels has tremendously increased in the last decade. This wealth of knowledge has led to the design of immunotherapeutic strategies to treat cancer that may result in clinical responses in many patients and may cause significant survival prolongation in some patients. One lesson we have learned is that, complicated at least by the tumor’s genetic instability, the effectiveness of cancer immunotherapy is in most cases transient, since clinical responses are in general followed by recurrence of lesions populated by immunoresistant tumor variants with additional mutations. The results we have summarized strongly suggest that immune selection is a mechanism underlying the generation of immunoresistant tumor variants with HLA class I antigen defects in malignant diseases. Other mechanisms, which do not involve immune selection, may include oncogene-mediated MHC class I molecule and APM component down-regulation [62–65].

It is noteworthy that immune selection is not restricted to T cells, but is likely to operate also with NK cells. Because the activity of NK cells is inhibited by MHC class I molecules [66], in situations where NK cells provide the major source of selective pressure, HLA class I antigen expression may be indeed advantageous to tumor survival. This possibility is supported by the high-level HLA class I antigen expression in uveal melanoma [33], breast carcinoma [34] and lung carcinoma [35], where NK cells are thought to control tumor progression. In addition, higher frequency of NK cell activating ligand MICA/B loss has been observed in metastatic melanoma lesions than in primary lesions [67], suggesting that MIC could also be a target of immune selection in the course of a malignant disease. Experimental data in vitro have shown that MICA loss was associated with resistance to NK cell-mediated lysis of two human melanoma cell lines isolated from recurrent metastases in spite of lack of HLA class I antigen expression [68]. Similar results have been obtained in C57BL/6 mice. The mock-transfected, but not the Rae-1-transfected, RMA lymphoma cells form tumors in syngeneic mice and depletion of NK cells in mice abolishes the control of Rae-1-transfected RMA cell growth [69]. It is also worth noting that tumor cells may shed MIC, which blocks the function of the cognate receptor NKG2D [70]. Elevated serum MIC levels have been reported to be associated with a poor prognosis in patients with prostate carcinoma [71]. Following the same scenario as with cell-mediated immunity, immune selection may also work with antibodies that recognize membrane-bound TAs, provided that humoral immune responses are strong enough to pose selective pressure and that the targeted TAs are not essential to tumor survival.

It is worth noting that aside from MHC class I molecule abnormalities, tumor cells can utilize other means of escape such as induction of T cell tolerance [72] and secretion and/or release of suppressive immune modulators [73]. When multiple escape mechanisms are possessed by malignant cells, the host immune system can eventually be rendered unresponsive to tumor cells and loses the battle. In light of the many limitations we have described that potentially hamper the host’s effort to eliminate tumors, one should be cautious when selecting a target for immunotherapy. Ideally, these targets should be molecules essential for survival and/or growth of tumor cells. With the application of alternative therapeutic strategies, these targets may be utilized to restore the function of both the intrinsic and extrinsic tumor suppressor systems in patients with malignant diseases.