Rejection versus escape: the tumor MHC dilemma

Abstract

Most tumor cells derive from MHC-I-positive normal counterparts and remain positive at early stages of tumor development. T lymphocytes can infiltrate tumor tissue, recognize and destroy MHC class I (MHC-I)-positive cancer cells (“permissive” phase I). Later, MHC-I-negative tumor cell variants resistant to T-cell killing emerge. During this process, tumors first acquire a heterogeneous MHC-I expression pattern and finally become uniformly MHC-I-negative. This stage (phase II) represents a “non-permissive” encapsulated structure with tumor nodes surrounded by fibrous tissue containing different elements including leukocytes, macrophages, fibroblasts, etc. Molecular mechanisms responsible for total or partial MHC-I downregulation play a crucial role in determining and predicting the antigen-presenting capacity of cancer cells. MHC-I downregulation caused by reversible (“soft”) lesions can be upregulated by TH1-type cytokines released into the tumor microenvironment in response to different types of immunotherapy. In contrast, when the molecular mechanism of the tumor MHC-I loss is irreversible (“hard”) due to a genetic defect in the gene/s coding for MHC-I heavy chains (chromosome 6) or beta-2-microglobulin (B2M) (chromosome 15), malignant cells are unable to upregulate MHC-I, remain undetectable by cytotoxic T-cells, and continue to grow and metastasize. Based on the tumor MHC-I molecular analysis, it might be possible to define MHC-I phenotypes present in cancer patients in order to distinguish between non-responders, partial/short-term responders, and likely durable responders. This highlights the need for designing strategies to enhance tumor MHC-I expression that would allow CTL-mediated tumor rejection.

Introduction

The history of tumor immunology began with the transplantation of chemically induced tumors into newly developed inbred strains of mice and rats. These experiments formed the basis for the MHC discovery by Gorer in 1935 [1]. Foley in 1953 [2], Baldwin in 1955 [3] and Prehn and Main in 1957 [4] described the antigenic properties of methylcholanthrene-induced tumors and reported that it is possible to protect inbred mice and rats against growth of chemically induced sarcomas by pre-immunizing with the same autologous tumor. Later, Klein et al. showed that this protection can be achieved when the tumor challenge is done in the same mouse, which ruled out alloimmune interactions [5]. In 1970, Basombrio reported the absence of tumor cross-reactivity in the protection, suggesting the existence of individual unique tumor-associated transplantation antigens in chemically induced sarcomas [6]. The discovery of human tumor antigens recognized by T lymphocytes was a major breakthrough [7], which highlighted the role of antigen presentation of small peptides via MHC molecules [8]. Currently, it is widely accepted that T lymphocytes play a major role in the recognition and destruction of solid tumors.

For decades, a variety of procedures to boost T lymphocyte responses have been used in cancer treatment, including IL-2 [9] and IFN-α [10] in the treatment of melanoma, BCG in bladder cancer [11] or autologous transfer of T lymphocytes in different types of malignancy [12]. In the mid-1990 s, tumor peptides alone or dendritic cells loaded with peptides were also used to augment anti-tumor T-cell responses [13] and were defined as “therapeutic vaccination.” Recently, cancer immunotherapy has been revisited and currently has a strong impact in clinical oncology. The use of monoclonal antibodies that target molecules involved in the regulation of tumor recognition by T lymphocytes and in T-cell cytotoxicity directed against metastatic cells has provided a new tool for increasing anti-tumor immune responses [14, 15]. This new approach confirmed that cytotoxic CD8+ T lymphocytes play a major role in the destruction of solid tumor masses when tumor cells have proper antigen presentation capability. However, all these therapies aimed at boosting anti-tumor T-cell responses in the tumor microenvironment have a common obstacle, namely different strategies used by tumor cells to escape immune recognition.

T lymphocytes can infiltrate human solid tumors [16–18], and the intensity and type of the infiltration has been strongly correlated with the clinical prognosis in cancer patients [19]. It is well documented that tumors can escape T-cell-mediated immune responses by losing or downregulating MHC class I (MHC-I) molecules (HLA class I (HLA-I) in humans) [20–26]. This phenomenon has been observed in a variety of human tumors derived from HLA-I-positive epithelia. On the other hand, low HLA-I expression makes cancer cells particularly susceptible to NK cell-mediated killing due to the lack of ligands for the killer-cell inhibitory receptors (“missing-self” hypothesis) [27]. However, as has been demonstrated in different types of cancer, solid tumors are rarely infiltrated by CD56+ NK cells compared to CD8+ T-cells, while the loss of MHC-I is randomly associated with NK cell infiltration. This inability of NK cells to migrate closer to malignant cells and inactivation of NK cell effector functions could be caused by inhibitory factors produced by immunosuppressive immune cells and by cancer cells [28].

Our laboratory has been analyzing HLA expression in solid tumors and cells lines for many years, defining procedures and methods to detect these polymorphic molecules on the tumor cell surface. HLA analysis in tumor tissues requires a different approach as compared to the HLA assays used for organ transplantation. Normally, monomorphic, locus and allele-specific anti-HLA monoclonal antibodies (mAb) able to recognize epitopes preserved in frozen or paraffin-embedded tissues are used for tumor HLA-I analysis [29].

This focused research review will discuss the expression pattern of HLA-I antigens in primary or metastatic malignant lesions with a special emphasis on the relationship between tumor HLA-I expression, leukocyte infiltration and tumor tissue structure. We will also highlight the clinical impact of the molecular mechanism (“hard” or “soft” lesions) responsible for HLA-I alterations in tumor rejection or escape in different types of immunotherapy.

T-cell immune selection of MHC-I-negative tumor variants

We have obtained evidence in both mouse and human tumors that T-cells are responsible for selecting MHC-I-negative variants during tumor development. An H-2-class I-negative mouse fibrosarcoma clone derived from a heterogeneous H-2 primary tumor GR9, induced H-2-negative spontaneous metastasis in immunocompetent syngeneic mice. In contrast, the same clone produced stable H-2 class I-positive spontaneous metastases in immune-deficient nude/nude mice lacking T-cells [30]. The H-2 class I-deficient metastatic tumor cells were produced by a coordinated downregulation of antigen presentation machinery (APM) components that could be reversed after IFN treatment [31]. Similarly, a primary melanoma lesion with heterogeneous pattern of HLA-I expression after a period of 10 months of cancer progression generated a homogeneously HLA-I-negative metastasis with the same molecular alteration in the B2M gene as the primary tumor [32].

The high incidence of HLA-I-negative tumor variants in a large variety of human tumors highlights the role that T-cells are playing in destroying HLA-I-positive solid tumor masses as a result of successful immune surveillance [33, 34]. It can also explain the failure of the rejection of HLA-I-deficient metastatic nodes in response to different types of immunotherapy, including peptide vaccination, dendritic cells loaded with peptides [13, 35], or new approaches, such as antibodies against immune checkpoint inhibitors [14, 36]. We have observed clear examples of both rejection and escape in two melanoma patients with mixed responses to autologous vaccination. We discovered that HLA-I-positive metastatic lesions heavily infiltrated by CD4+ and CD8+ T lymphocytes (TILs) were rejected. In contrast, HLA-I-negative metastatic nodes with few TILs were not rejected and progressed [37, 38]. There are other reports describing a positive correlation between the number of TILs and tumor HLA-I expression in different types of cancer, including pancreatic and lung carcinoma [39, 40]. These findings strongly suggest that T-cell-mediated immune selection of MHC-I-negative tumors is a major mechanism for the generation of tumor escape variants.

MHC-I expression in primary tumors

HLA-I analysis of cryopreserved human primary tumor tissues revealed that the expression of these molecules can range from positive (more that 75% cells stained), heterogeneous (between 25 and 75%), and negative (<than 25% cells stained). This classification was approved and used at the 12th International Histocompatibility Workshop in Paris in 1996 [41]. These patterns can be observed in different tumors derived from HLA-I-positive epithelia (breast, colorectal, bladder, lung, melanoma, prostate (Fig. 1), head and neck, pancreas, thyroid, cervix, endometrium) and have been extensively described and reviewed by different groups [20–26]. These different types of HLA-I expression could represent different stages of T-cell immune selection taking place during the natural history of tumor development. At early stages, the tumor is predominantly HLA-I-positive and TILs begin to infiltrate and attack the nascent tumor mass. Later, the tumor becomes heterogeneous and is composed of both HLA-I-positive and HLA-I-negative cells (HLA-I heterogeneity). Finally, tumor escape variant becomes homogeneously HLA-I-negative with few TILs (Fig. 2a–c).

Fig. 1.

Different patterns of HLA-I expression in human prostate cancer. Three different HLA-I immunolabeling patterns in prostate tumor tissues: homogeneously positive HLA-I expression, heterogeneous HLA-I staining and homogeneously HLA-I-negative pattern with a clear histological separation of HLA-I-negative tumor from HLA-positive stroma. W6/32 mAb directed against HLA-ABC/B2M complex have been used to characterize tumor tissue HLA-I expression patterns

Fig. 2.

Tissue architecture in HLA-I-positive and HLA-I-negative breast (a), lung (b) and bladder (c) carcinomas in relation with the patterns of tumor infiltration with CD3+ and CD8+ lymphocytes. HLA-I-positive tumor tissue architecture is “permissive,” with tumor infiltration by different mononuclear cells, including T-cells (phase I). At later stages, primary tumors are heterogeneous, composed of HLA-I-positive and HLA-I-negative tumor cells, and TILs. Finally, tumor masses become uniformly HLA-I-negative. Leukocytes and other cells are now located outside the tumor in the stroma with a clear morphological separation of tumor cells from the stroma generating an encapsulated “non-permissive” structure (phase II). W6/32 mAb were used to analyze the expression of HLA-I-ABC/B2M complex

MHC-I expression in metastases

We have observed that some metastatic tissue samples and cell lines derived from them are uniformly HLA-I-negative when compared to autologous primary tumors. We have previously reported two such examples, one of which represents a heterogeneous primary melanoma with HLA-I-positive and HLA-negative areas. Molecular analysis of the DNA obtained from microdissected HLA-I-negative areas allowed us to identify loss of heterozygosity (LOH) in one chromosome 15 (harboring B2M) together with a point mutation in codon 67 in the second copy of B2M gene, producing a total loss of B2M and of HLA-I complex expression on the cell surface [32]. Analysis of an autologous metastasis detected 10 months later showed a “homogeneous” HLA-I-negative immunolabeling pattern and the same molecular lesion in the B2M gene as in the HLA-I-negative primary tumor, strongly supporting the idea that active T-cell immunoselection takes place during tumor/metastasis development (Fig. 3a). In another example, a primary colorectal tumor was heterogeneous for HLA-I immunolabeling, while autologous liver metastases were uniformly HLA-I-negative (Fig. 3b). These findings do not necessarily suggest that this is a common tendency and that all the metastatic nodes in a given cancer patient have the same HLA-I phenotype. There are reports describing that distant metastasis derived from different types of cancer can also have HLA-I-positive phenotypes. It all probably depends on various immunological and non-immunological factors, including the route of dissemination, which could be via blood circulation or the lymphatic system. We have previously reported that autologous melanoma metastases can have distinct HLA-I expression patterns [37, 42]. There are similar reports describing positive HLA-I expression in metastases originated from colorectal cancer [43] and in experimental tumors [44].

Fig. 3.

HLA-I immunostaining patterns in primary melanoma (a) and colorectal tumor (b) and in corresponding autologous metastases. a Primary melanoma lesion has heterogeneous HLA-I labeling pattern with positive and negative areas surrounded by stroma. Autologous melanoma metastasis is uniformly negative for HLA-I expression, while the stroma is HLA-I-positive. b Primary colorectal cancer is clearly heterogeneous for HLA-I expression with positive and negative areas, while autologous liver metastasis is homogeneously HLA-I-negative

Two patterns of MHC-I expression, leukocyte infiltration and tumor tissue architecture

We have recently obtained new evidence indicating that tumor tissue structure and the pattern of leukocyte infiltration are directly related to the expression of HLA-I molecules by tumor cells [45].

Tumor tissue structure in MHC-I-positive tumors (phase I)

When tumor cells express HLA-I molecules at early stages of tumor growth there is a “permissive” interaction with TILs, allowing recognition and destruction of the target tumor cells (Fig. 5a). Tumors are infiltrated by CD3+, CD4+, and CD8+ lymphocytes to various degrees depending on the HLA-I-positive/negative tumor cell ratio within the heterogeneous tumor mass. We present examples obtained in breast, lung and bladder carcinoma in which HLA class I expression is associated with intra-tumor T-cell infiltration (Fig. 2).

Fig. 5.

Three-dimensional schematic demonstration of the two phases of tumor development with the corresponding tissue organization patterns: a phase I or “permissive” and b phase II or “non-permissive/encapsulated”. a In phase I, tumor cells are HLA-I-positive (green) and are surrounded and killed by CD8+ T lymphocytes (yellow). HLA-I-negative or HLA-I-deficient tumor cells (red) escape the recognition and destruction by TILs. The tumor is now heterogeneous and is composed of HLA-I-positive and HLA-I-negative cells. Cells other than T-cell tumor-infiltrating cells, such as macrophages (blue), can also be found within the tumor tissue. This pattern of the tumor/stroma tissue architecture is “permissive,” since it permits leukocytes and other cells to get in close contact with tumor cells. b In phase II as a result of T-cell immune selection, the tumor becomes homogeneously negative for HLA class I expression (red cells). These HLA-I-negative tumor cells are encapsulated by the stroma rich in tumor-specific T lymphocytes (yellow) and other cells (blue), such as Tregs and MDSCs. A clear physical separation between these two structures (tumor cells and the stroma) creates a “non-permissive” tissue structure which prevents immune cell infiltration into the tumor mass. Thus, due to this immunosuppressive tumor microenvironment, HLA-I-negative cells are in an immune privileged tumor structure

Tumor tissue structure in MHC-I-negative tumors (phase II)

In contrast, when HLA-I-positive cells are destroyed and tumors are composed only of HLA-I-negative tumor cells, tumor architecture changes (Fig. 5b). Now tumor nodes are surrounded by the stroma with different types of T-cells/leukocytes/macrophages and probably other elements of the tumor microenvironment, including Tregs and MDSCs, creating a “non-permissive” tissue structure. In Fig. 4, we present an example of different types of HLA-I-negative tumors encapsulated by the stroma.

Fig. 4.

HLA-I-negative tumors encapsulated by surrounding fibrous stroma. a Breast carcinoma homogeneously negative for HLA-I with a clear separation from HLA-I-positive stroma surrounding tumor nests. CD3+ cells are outside the tumor tissue in the fibrous stromal capsule. b The “non-permissive” tumor tissue structure observed in HLA-I-negative tumors (lung, colorectal, laryngeal, bladder). HLA-I-negative tumor markedly contrasts with the HLA-I-positive stroma. Note that these tumor tissue examples represent primary lesions with a homogeneous negative HLA-I pattern. Tissue architecture with tumor nodes encapsulated by the stroma is clearly observed. W6/32 mAb were used to analyze the expression of HLA-I-ABC/B2M complex

We propose that in phase II the cellular immune system is trying to isolate tumor cells from the rest of the body by creating a structure with a tumor node surrounded by stromal capsule that resembles “granuloma”-type tissue organization seen in different pathological conditions [46, 47]. Additionally, we favor the idea that in phase II anti-tumor lymphocytes are “out of work.” They reside outside the tumor mass in the stroma, because cannot enter tumor parenchyma and cannot see HLA-I-negative tumor target cells, but are primed and ready to attack tumor cells if they recover HLA-I expression (Fig. 4a) [48].

A three-dimensional schematic representation of these two phases is presented in Fig. 5. In the permissive phase I (Fig. 5a), T-cells are entering tumor tissue attacking and destroying tumor cells, while HLA-I-negative tumor cells escape T-cell immune destruction. In this phase, the tumor is heterogeneous and consists of both HLA-I-positive and HLA-I-negative cells, depending on the stage of tumor destruction.

In phase II (Fig. 5b), tumor cells are homogeneously HLA-I-negative and encapsulated by the stroma. Anti-tumor T lymphocytes have finished their work destroying all HLA-I-positive cells, but the tumor is still growing because is composed of HLA-I-negative escape variants (Fig. 5b). We have seen this type of tumor tissue structure in different HLA-I-negative tumor tissues: breast, bladder, colorectal, lung, melanoma, etc. (Figure 4b). This phase II is associated with peculiar tumor tissue architecture observed in different growing tumors with a clear separation between the tumor tissue and the stroma. However, such tumor tissue organization, well defined by pathologists a long time ago, has not been previously associated with the absence of HLA-I molecules in tumors.

MHC-I tumor phenotypes

The complexity of the HLA system can also be appreciated during the analysis of altered tumor HLA-I phenotypes. Our experience obtained after many years of detailed analysis of HLA expression in human tumors indicates that, in most cases, a combination of two different hits affecting the HLA/B2M genes or the transcriptional regulation of any gene participating in tumor antigen presentation takes place. We classified different types of HLA-I alterations in tumors in order to facilitate the understanding of this phenomenon [21, 22].

Tumor phenotype I, HLA class I total loss

Out of the six HLA-I alleles presented in normal cells, tumor cells can lose all. We define it as phenotype I, or total loss of expression. It can be produced by: (a) a structural “hard” genetic alterations affecting both B2M genes (point mutation or deletions of large DNA segments affecting chromosome 15 [49, 50]) or (b) transcriptional downregulation affecting HLA-I, B2M or APM genes [50]. This regulatory alteration, “soft lesion,” can be reversed in vitro by different cytokines. The frequency of phenotype I varies between different tumors and can be observed from 15% in melanoma and colorectal cancers up to 40–50% in breast, bladder or prostate carcinomas [21, 22].

Tumor phenotype II, HLA haplotype loss

Tumor cells can lose an HLA haplotype, which includes three HLA-I alleles (A, B, and C). This is a frequent mechanism observed in primary tumors of different histological type [51]. This “hard” defect can only be corrected if the missing HLA-I alleles are re-introduced into the tumor cell. We have reported that chromosome 6 loss is the most frequent mechanism producing an HLA haplotype loss [52].

Tumor phenotype III, HLA class I locus A, B, C selective loss

Tumors can coordinately downregulate HLA-A, B or C locus-specific products. In this case, the expression of HLA-A, B or C genes is switched off on a transcriptional level and can be restored by IFN-γ or other TH1-type cytokines. It is not defined accurately how frequent these altered phenotypes are present in tumors of different histological type [21, 22].

Tumor phenotype IV, HLA class I allelic loss

Tumors can lose one single HLA-I allele; however, it is not a frequent alteration. Nevertheless, there are well-defined HLA-I allelic mutations reported in various tumor cell lines [53–55].

Compound phenotype V

It has been frequently observed that tumors during cancer progression accumulate different types of HLA-I alterations simultaneously. For instance, some tumor cells can express only an HLA-C allele on the cell surface as a result of a combination of an HLA haplotype loss produced by a LOH in chromosome 6 and HLA-A and B locus transcriptional downregulation [22].

Phenotype VI: unresponsiveness to IFN

Tumors can escape T-cell recognition by developing strategies affecting the capacity to upregulate HLA-I molecules. Among them, the mutations of genes involved in IFN signaling pathways have been described in different tumor cell lines [56]. The resistance to IFN can be functionally analyzed in tumor cell lines, but the frequency of this defect in solid tumor tissues is not known.

Frequency of HLA-I loss in human tumors

The analysis of HLA-I antigens in tumor tissues requires a complex approach because the HLA-I heavy chain is highly polymorphic and requires the analysis of the expression of six HLA-I alleles on the tumor cell surface which differ among cancer patients. Frozen tumor samples obtained from cancer patients are commonly analyzed by immunohistology. Microdissection of tumor tissue is currently used to obtain DNA and RNA from particular stromal or tumor areas to define the molecular defects responsible for HLA-I alterations [29, 50]. A more precise definition of the tumor phenotype and of the underlying mechanism of HLA-I defects can be obtained by the combined use of immunohistochemistry together with PCR, comparative genomic hybridization and microsatellite analysis with specific markers spanning the chromosomal region of interest (detection of LOH). Analysis of various tumor HLA-I phenotypes detected in human tumors derived from HLA-I-positive epithelia produced an overall frequency of HLA-I alterations in tumors of different histological type, ranging from 15 to 93% (Fig. 6).

Fig. 6.

Frequency of HLA-I loss in different types of cancer. Summarized results of the immunohistological analysis of tumors using different antibodies directed against monomorphic, locus and allele-specific HLA determinants. For details, see Refs. [21, 22, 29, 39, 65, 67]

Relevance of “hard” and “soft” molecular lesions for the recovery of HLA-I antigens and induction of tumor rejection

Assuming that most human solid tumors have one particular or a combination of different altered HLA-I phenotypes, which they use to escape specific anti-tumor T-cell responses, the recovery of these antigen-presenting molecules is crucial to restore the capacity of T-cell to recognize and reject tumor cells [57] and it depends on the molecular mechanism responsible for the HLA-I alteration. Treatment with TH1 cytokines can potentially recover alterations caused by deregulation of the transcription of different HLA-A, B, C, B2M or APM genes (“soft” lesion) [38, 50]. The recovery of HLA-I expression in tumor cells with “soft” lesions can be induced by different types of immunotherapies and accounts for the clinical responses observed in some regressing metastatic lesions. In contrast, if HLA-I loss is caused by structural damage of these genes, such as mutations or deletions (“hard” lesions), HLA-I recovery would be possible only by replacement of the altered gene [50, 58, 59].

Recovery of HLA-I in tumor cells with “soft” lesions: broadening of the specificity of anti-tumor T-cell responses

It is well established that in some HLA-I-deficient tumor cell lines, normal HLA-I expression can be recovered after IFN treatment. In fact, in vitro IFN screening is an efficient approach to distinguish “soft” regulatory HLA-I alterations from the structural ones. Figure 7a shows an example of two melanoma cells lines (ESTAB-025 and 042) with IFN-γ-induced HLA-I upregulation determined by flow cytometry. We believe that different cancer immunotherapies (IL-2, different IFNs, BCG or autologous vaccines) traditionally used to treat cancer, including melanoma or bladder cancer, lead to activation of the tumor microenvironment and upregulation of HLA-I expression, and promote tumor rejection [60].

Fig. 7.

Recovery of HLA-I expression in melanoma cells: by IFN in the case of “soft” lesions and by adenovirus-mediated B2M gene transfer in cells with “hard” HLA-I lesions. a “soft” lesions—a selective HLA-B locus downregulation in ESTDAB-25 cells and total HLA-ABC downregulation in ESTDAB-42 cells—are corrected by IFN-γ (as demonstrated on flow cytometry plots); b resistance to IFN-γ-mediated HLA-ABC upregulation in melanoma cell line ESTDAB-004 (“hard” lesion); c recovery of HLA-I expression in melanoma cells with total loss of HLA-I expression due to B2M mutations (“hard” lesion) after adenovirus-mediated transfer of a wild-type human B2M gene (as demonstrated by confocal microscopy). W6/32 mAB were used to examine cell surface HLA-ABC/B2M complex; L368 mAb were used to detect B2M protein

It has been reported that peptide vaccination in melanoma patients induces proliferation of high numbers of anti-tumor T-cells that are not specifically directed against the peptide used for vaccination. This phenomenon, observed by different authors, has been named “epitope spreading” [61]. It is not known which peptide/peptides define the specificity of newly induced anti-tumor T-cells. We have proposed that this commonly observed phenomenon could be due to the recovery of HLA-I expression in HLA-I-deficient tumor cells harboring “soft” lesions after the release of TH1 cytokines in the tumor microenvironment as a consequence of immunotherapy [50, 57]. This HLA-I recovery allows primed T-cells to see again the original tumor peptide hidden as a result of the tumor HLA-I loss. A similar phenomenon has been recently observed in cancer patients treated with anti-CTLA4 monoclonal antibodies [62].

Recovery of HLA-I in tumor cells with “hard” molecular lesions

HLA-I upregulation would not be possible if malignant cells contain irreversible structural alterations in HLA genes or demonstrate resistance to cytokine-mediated upregulation due to defects induced by mutations in signal transduction pathways. The in vitro IFN screening becomes useful for detection of such alterations. “Hard” lesions can appear as a result of genetic defects in the genes coding for B2M, HLA-ABC or molecules involved in IFN activation pathways. In Fig. 7b, we present one melanoma cell line (ESTDAB-004) with a resistance to IFN-γ due to alterations in IFN signaling pathways [56, 63]. When the mechanism responsible for HLA-I loss is due to structural genetic aberrations, cytokines are unable to upregulate HLA-I expression, rendering primed T-cells incapable of detecting tumor antigens. Hence, tumor cells with structural (“hard”) HLA aberrations may escape immune recognition even after immunomodulatory treatment and can represent a major obstacle to anticancer immunity [48, 57]. In this context, we have reported that the in vitro and in vivo transfer of wild-type human B2M or HLA-A2 genes in HLA-I-deficient tumor cells using a recombinant adenovirus is capable of recovering a functional HLA-I molecule on the tumor cell surface leading to recognition and elimination by cytotoxic T-cells [58, 59] (Fig. 7c).

Conclusions

It is well documented that human and experimental tumors can induce T-cell-mediated responses [64]. This natural immunosurveillance mediated by T-cells is directed against native tumor antigens represented by unique mutated peptides. At this stage, tumor cells are infiltrated by specific anti-tumor CD8+ T lymphocytes. We named this stage “permissive, phase I.” This period of tumor cell killing by T-cells can end in the total destruction of the tumor with no clinical evidence. Alternatively, MHC-I-negative tumor cells can escape the powerful T-cell immune selection at early stages of tumor development and proliferate. Homogeneously, MHC-I-negative tumors actively create an immunosuppressive microenvironment that prevents intra-tumor T-cell infiltration and produces an “encapsulated” tumor structure defined by us as “phase II”. The transition from phase I to phase II probably lasts for a short period of time and can be observed in tumors of different origin as demonstrated in this review. The molecular mechanism responsible for MHC-I loss or downregulation plays a crucial role in defining the capacity of tumor cells to upregulate MHC-I molecules after cytokine release. If the molecular mechanism is reversible or “soft,” different immunotherapies are able to upregulate MHC-I and induce T-cell-mediated tumor rejection [65]. In contrast, if the molecular mechanism is irreversible or “hard”, with structural gene damage affecting antigen presentation, tumor cells will likely demonstrate a resistance to any type of T-cell-mediated immunotherapy, resulting in cancer progression [50, 66, 67]. The tumor MHC-I dilemma determines tumor “rejection” versus tumor “escape” since MHC-I “expression” or “loss” is a central event during tumorigenesis (“dilemma” in Latin, comes from Greek δίλημμα or “double proposition,” is a situation in which a choice has to be made between two opposite alternatives). After looking back eighty years ago, it is very enlightening to see that the transplantation genetic systems in mice (the H-2 complex) were defined using tumors [1]. Now we see that the H-2/HLA genes and molecules play a pivotal role in cancer immunology. Are they finally paying back for being discovered? Who knows, but it is certain that MHC and cancer will be moving ahead together in the years to come.

Abbreviations

APM

Antigen presentation machinery

B2M

Beta-2-microglobulin

BCG

Bacille Calmette–Guerin

CTL

Cytotoxic T lymphocyte

CTLA4

Cytolytic T lymphocyte-associated antigen 4

DNA

Deoxyribonucleic acid

FACS

Fluorescence-activated cell sorter/sorting

FITC

Fluorescein isothiocyanate

HLA

Human leukocyte antigen

IFN

Interferon

IL

Interleukin

LOH

Loss of heterozygocity

mAb

Monoclonal antibodies

MDSC

Myeloid-derived suppressor cell

MFI

Mean fluorescence intensity

MHC

Major histocompatibility complex

PCR

Polymerase chain reaction

TH1

T helper 1

TIL

Tumor infiltrating lymphocyte

Compliance with ethical standards

Conflict of interest

Authors declare no conflict of interest.