“Hard” and “soft” lesions underlying the HLA class I alterations in cancer cells: Implications for immunotherapy


  • Federico Garrido,

    Corresponding author
    1. Departamento de Bioquímica, Biología Molecular 3 e Inmunología, Facultad de Medicina, Universidad de Granada, Granada, Spain
    2. Servicio de Análisis Clínicos, Hospital Universitario Virgen de las Nieves, Granada, Spain
    • Hospital Universitario Virgen de las Nieves, Servicio de Análisis Clínicos. Avenida de las Fuerzas Armadas 2, 18014 Granada, Spain
    Search for more papers by this author
    • Fax: +34-958-020069

  • Teresa Cabrera,

    1. Departamento de Bioquímica, Biología Molecular 3 e Inmunología, Facultad de Medicina, Universidad de Granada, Granada, Spain
    2. Servicio de Análisis Clínicos, Hospital Universitario Virgen de las Nieves, Granada, Spain
    Search for more papers by this author
  • Natalia Aptsiauri

    1. Servicio de Análisis Clínicos, Hospital Universitario Virgen de las Nieves, Granada, Spain
    Search for more papers by this author


The ability of cancer cells to escape from the natural or immunotherapy-induced antitumor immune response is often associated with alterations in the tumor cell surface expression of Major Histocompatibility Complex (MHC) Class I antigens. Considerable knowledge has been gained on the prevalence of various patterns of MHC Class I defects and the underlying molecular mechanisms in different types of cancer. In contrast, few data are available on the changes in MHC Class I expression happening during the course of cancer immunotherapy. We have recently proposed that the progression or regression of a tumor lesion in cancer patients undergoing immunotherapy could be predetermined by the molecular mechanism responsible for the MHC Class I alteration and not by the type of immunotherapy used, i.e., interleukin-2 (IL-2), Bacillus Calmette-Guèrin (BCG), interferon-alpha (IFN-α), peptides alone, dendritic cells loaded with peptides, protein-bound polysaccharide etc. If the molecular alteration responsible for the changes in MHC Class I expression is reversible by cytokines (“soft” lesion), the MHC Class I expression will be upregulated, the specific T cell–mediated response will increase and the lesion will regress. However, if the molecular defect is structural (“hard” lesion), the MHC Class I expression will remain low, the escape mechanism will prevail and the primary tumor or the metastatic lesion will progress. According to this idea, the nature of the preexisting MHC Class I lesion in the cancer cell has a crucial impact determining the final outcome of cancer immunotherapy. In this article, we discuss the importance of these two types of molecular mechanisms of MHC Class I–altered expression.

In spite of the great efforts in generating new and more advance cancer vaccines, the various immunization protocols used in clinical trials led to only limited clinical improvement with disease progression in the majority of the patients.1, 2 Circulating T cells capable of recognizing tumor antigens can be induced in cancer patients by peptide-based immunotherapy, but the efforts to achieve regression of established tumors or a clear positive clinical response in the treatment of metastatic disease has been below expectations.3, 4 In the meantime, different tumor escape mechanisms have been described and defined over the past years, including loss of tumor-specific antigens, lack of costimulatory signals and/or adhesion molecules, expression of immunosuppressive factors such as Fas ligand, vascular endothelial growth factor (VEGF), transforming growth factor-beta (TGF-β), or interleukin-10 (IL-10), induction of 2, 3 indoleamine dioxygenase and deficiencies in the dendritic cell function or in the signal transduction pathway of CD8+ cytotoxic T cells (CTL).5–9

One of the important tumor escape mechanisms represents loss or downregulation of Major Histocompatibility Complex (MHC) Class I antigens in tumor cells that have been frequently observed in a variety of mouse tumors and human malignancies derived from HLA (Human Leukocyte Antigen) Class I positive epithelia.10–12 These MHC Class I–deficient tumor variants lose the antigen-presenting molecule and are, therefore, resistant to T cell cytotoxicity. At the same time, cells with total loss of MHC Class I surface expression can potentially become susceptible to natural killer (NK) cell antitumor activity induced by the lack of inhibitory MHC antigens (“missing self”).13 Another immunoselection route is provided by the partial loss of HLA Class I antigens that allows tumor cells to escape both CTL and NK attack. For instance, it has been reported that colorectal cancer with intermediate HLA Class I expression are associated with poor prognosis as compared to HLA positive or totally negative tumors, suggesting that such tumors with intermediate HLA Class I expression may avoid both NK- and T cell–mediated immune surveillance.14 In addition, as we discussed above, some tumor lesions might escape independently of HLA Class I expression as it has been documented for renal cell carcinomas where tumors have higher expression than the corresponding autologous normal tissue.15

The discovery of MHC losses in human and mouse tumors goes back to the middle of the 70s using allo-antisera, tumor cell lines and tissues.12, 16 The detected abnormal anti-MHC serological reactivity was at the beginning controversial due to poor quality of the reagents and/or the purity of the mouse strains used.17, 18 The appearance of monoclonal antibodies directed against H-2 and HLA molecules confirmed the existence of such loses and the idea of tumor escape variants began to emerge.19 During the past years, the MHC Class I expression in cancer have been widely analyzed using a panel of specific anti-MHC antibodies on tumor biopsies or in tumor cell lines. Frozen tissue obtained from cancer patients in coordination with pathologists is commonly analyzed by immunohistology. Microdissection of tumor tissue is currently used to obtain DNA and RNA from particular stroma or tumor areas to define the molecular defects responsible for MHC Class I alterations.20

A more precise definition of the tumor phenotype and of the underlying mechanism of HLA Class I defects can be obtained by the combined use of immunohistochemistry with tissue microdissection, polymerase chain reaction (PCR), comparative genomic hybridization, fluorescence in situ hybridization and loss of heterozygosity (LOH) analysis with specific markers spanning the chromosomal region of interest.

Our laboratory for many years has been studying MHC Class I expression in cancer, developing methodologies and selecting reagents that are capable of defining altered expression of these molecules in solid tumors and cancer cell lines.21, 22 The molecular mechanisms that have been found to underlie these alterations in MHC Class I expression vary and can occur at any step required for MHC synthesis, assembly, transport or expression on cell surface. These defects can occur at the genetic, epigenetic, transcriptional and post-transcriptional levels and represent either regulatory abnormalities that can be recovered with cytokine treatment or structural defects that cannot be reversed. Thus, MHC alterations can be classified into two main groups: reversible defects (regulatory or “soft”) and irreversible defects (structural or “hard”) (Table 1). Recently, our group obtained new information about the HLA changes occurring in tumor tissues from patients undergoing immunotherapy.54, 55 This article will focus on the clinical impact that these 2 types of MHC Class I alterations might have in the final outcome of T cell–mediated cancer immunotherapy.

Table 1. Molecular defects underlying MHC class I–altered expression
inline image

Irreversible vs. Reversible MHC Class I Alterations in Tumors

Various defects in HLA Class I expression with different underlying molecular mechanisms, tumor-type distribution and frequency have been documented extensively in a large variety of tumors.56 In some cases, up to 90% of tumor cells have HLA Class I–altered phenotypes.57 These phenotypes can be classified as: total loss of HLA Class I expression (phenotype I), the loss of one HLA haplotype due to LOH in chromosome 6 (phenotype II), HLA-A, -B or -C locus product downregulation (phenotype III), loss of one HLA Class I heavy chain allospecificity (phenotype IV), the accumulation of several alteration leading to a tumor cell expressing one single HLA Class I allele (phenotype V) or, finally, a tumor cell unable to upregulate HLA Class I molecules due to a blockade in the IFN signaling pathway (phenotype VI). These altered HLA phenotypes have been analyzed in detail and classified previously.11, 56

Abnormalities in cell surface expression of HLA Class I antigens in tumor cell lines can be detected by flow cytometry using a panel of specific anti-HLA Class I antibodies.41 The use of a panel of anti-HLA monoclonal antibodies defining monomorphic, locus specific or allelic determinants is absolutely necessary as tumor cells can lose all or only some HLA alleles. The same altered HLA Class I phenotypes described in solid tumor tissues can be defined in tumor cell lines.42 Cells can be treated in vitro with various immunomodulators [i.e., interferons (IFNs)] or pharmacological agents in attempts to recover normal HLA expression. Tumor cells can upregulate or leave without changing the HLA Class I phenotype after such manipulations indicating that they have reversible or structural defects needed to be identified. From the functional point of view, the altered HLA tumor phenotypes can be divided in two groups: (i) those capable of recovering or upregulating HLA Class I antigens after cytokine treatment or (ii) those that cannot recover HLA I expression. The first one is harboring a reversible mechanism; whereas, the second one has a structural defect.

Irreversible defects

Of the 6 classical MHC Class I genes normally expressed in a somatic cell, three can be lost in tumors simultaneously. This mechanism is associated with LOH in chromosome 6 harboring HLA-ABC genes, the most frequent mechanisms underlying MHC Class I irreversible alterations.23 LOH of chromosome 6p21 is an important mechanism that generates HLA haplotype loss in various human tumors with an average incidence of 35%58, 59 (Table 1).

Another mechanism leading to irreversible MHC defects is LOH in one chromosome 15 that carries the β2-microglobulin (β2m) gene along with mutation in the other homologous gene. This genetic lesion is associated with the total loss of HLA Class I expression.30, 31 A mutation hotspot located in the CT repeat region of exon 1 of the β2m gene has been proposed56, 60 reflecting an increased genetic instability in this region in malignant cells. This irreversible mechanism was originally described in the lymphoma cell line DAUDI61 and later in melanomas,62 colorectal carcinomas63 and other types of tumors.56 Coexistence of distinct mutations in the two β2m genes is a rare event that has been reported in colorectal carcinomas64, 65 and in a sarcomatoid renal carcinoma.66

Recently, we have found that LOH in chromosome 15 can be frequently detected in tumors (40% of colon, melanomas and laryngeal carcinomas; 50% of bladder carcinomas)23 (Maleno and Garrido, unpublished data). This lesion in chromosome 15 may be unnoticed as the tumor cells might have “normal” HLA Class I pattern and could represent one of the early events in malignant cells leading to the generation of precommitted tumors to become MHC escape variants. LOH in chromosome 15 in tumors can be found more frequently than mutations in β2m gene.32

Another irreversible genetic damage is caused by a selective loss of one MHC Class I allele due to mutations, deletions and somatic recombination.11 There are not many single HLA Class I allelic mutations described so far.27–29 One can assume that this is not a frequent event happening during the natural history of tumor development. Similarly, mutations in the transporter associated with antigen processing (TAP) leading to HLA Class I loss is a rear event and have been described in lung cancer67 and in melanoma.68 Finally, resistance to IFN-γ–mediated upregulation of HLA Class I expression can also be a mechanism producing tumor escape variants. It is caused by defects in the Jak-STAT components of IFN-mediated signaling pathway.37, 38 These tumor cells express low levels of MHC Class I that cannot be upregulated. This defect is probably sufficient to escape T cell recognition and destruction.39 It is important to emphasize that an HLA haplotype loss, an HLA allelic loss or an LOH in one β2m gene will not be detected using the IFN screening methodology.

Structural defects can only be corrected by replacement of the defective HLA or β2m gene. In this context, we have recently succeeded in restoring HLA Class I expression to several cell lines with a double knock out of both copies of the β2m gene. We transferred a functional β2m gene into HLA Class I negative tumor cells using a recombinant adenovirus vector.69 The adenoviral-mediated recovery of HLA Class I expression on tumor cells in combination with vaccination or adoptive T cell therapy can provide an additional approach to improve the clinical efficacy of cancer immunotherapy.

Reversible defects

Downregulation of HLA Class I antigen expression is frequently caused by defects in the gene regulation of HLA Class I heavy chain genes, β2m gene and the components of the antigen-processing machinery (APM)44 in human and experimental tumors in mice.45, 46 Such abnormal MHC Class I phenotypes show low mRNA levels of specific genes (heavy chain, β2m and APM) that seem to be coordinately downregulated. These defects are reversible as they can be corrected in vitro by IFN-γ or other cytokines.41

HLA-A, HLA-B or HLA-C locus–specific downregulation is another frequent finding in tumor tissues and tumor cell lines of different origin.40, 41 In this context, we have recently reported the analysis of the HLA Class I expression in 92 human melanoma cell lines (ESTDAB cell line bank). HLA Class I locus-specific downregulation was observed in 35% of the cell lines, affecting mostly the locus B expression.41 In melanoma tumors, such defect can also be observed in similar frequencies.11, 40, 56 This low expression is upregulated after IFN-γ treatment.41

Epigenetic events associated with tumor development and with cancer progression have been found to underlie changes in HLA antigen and APM components. Unlike genetic alterations, epigenetic modifications can, in some cases, be reversed in vitro with pharmacologic agents that induce DNA hypomethylation or inhibit histone deacetylation. We previously reported an epigenetic downregulation of HLA Class I antigens in melanoma cell line MSR-3, which has hypermethylation of HLA Class I genes50 (Table 1).

MHC Class I expression changes after immunotherapy

Considerable knowledge has been gained on the prevalence of various patterns of HLA Class I defects and the underlying molecular mechanisms in particular types of cancer.5, 11, 56 These changes are happening during the natural history of tumor development.10 We have obtained evidences in experimental mouse tumor clones that T cell immunoselection is responsible for these MHC Class I phenotypic changes.70 MHC Class I negative tumor clone with a reversible defect was capable of reexpressing H-2 antigens after in vitro IFN-γ treatment. When this mouse fibrosarcoma tumor clone was injected in syngeneic immunocompetent mice, the metastases obtained were MHC negative. Interestingly, the metastases obtained in T cell immunodeficient mice were H-2 positive.45, 70

The level and intensity of T cell immunoselection in a cancer patient or in a developing tumor in mice is responsible for the destruction of tumor cells with different MHC Class I levels, that is, tumor cells with higher Class I expression will be easier to destroy than those with intermediate or low expression (see Fig. 1). The T cell response and the tumor growth will probably reach an equilibrium.71 This equilibrium is broken when a particular immunotherapeutic protocol is applied. Systemic or locally administered vaccines, cytokines, BCG or other immunomodulators, such as protein-bound polysaccharide, will most likely activate additional T cells in a clonal or polyclonal way. These activated T cells might reach the tumor and locally release new cytokines in the site of the lesion. These cytokines will now be able to upregulate MHC Class I only in tumor cells with reversible defects. This upregulation may increase the total number of Class I peptide complexes on tumor cells (including the specific tumor-rejection antigen peptides) and change the proportion and the variety of presented epitopes creating broadened antitumor CTL responses (epitope spreading). Consequently, the new T cell immunoselection will now promote the growth of tumor cells with preexisting irreversible MHC Class I lesions (see Fig. 1) as well as the proliferation of cells with new additional MHC defects as part of the oncoprogression.

Figure 1.

The primary tumor is composed of a heterogeneous population of cells: HLA positive (green +++), HLA intermediate (orange ±) and HLA negative (red −). In addition, tumor cells are either able or unable to upregulate the HLA expression due to the existence of reversible (orange) or irreversible structural (red) defects. The T cell–mediated natural immune response of the host destroys the HLA positive tumor cells. HLA upregulation can take place after cancer immunotherapy if the tumor cells harbor reversible (“soft”) defects. In this case, tumor cell with low HLA expression (orange) becomes highly HLA positive (green) and can be now destroyed. Finally, the relapsed tumor or resistant metastases will be composed of tumor cell variants immunoselected by T lymphocytes with negative or very low HLA expression with irreversible defects.

We have recently obtained clinical data that support this hypothesis. We found that several subcutaneous metastases in a melanoma patient responded in a different way to immunotherapy using autologous tumor vaccine (M-VAX) plus BCG.54 Three of the lesions progressed and the other three showed considerable regression. All lesions presented a loss of HLA Class I haplotype. However, the progressing metastases showed reduced HLA Class I expression, an additional LOH in chromosome 15 and HLA B locus dowregulation.54 In contrast, the regressors showed high levels of HLA Class I expression with an HLA haplotype loss similar to the progressors. Comparable results were obtained in another melanoma patient treated with IFN-α-2b and with an autologous tumor vaccine.55 Five lesions were obtained after INF-α-2b treatment and five metastases after autologous tumor vaccination plus BCG. Eight metastases were regressing after immunotherapy, whereas two were progressing. The regressing metastases showed high level of HLA Class I expression; whereas, the two progressing lesions had low HLA levels as measured by real time PCR and immunohistological techniques.55 These results indicate a strong association between HLA Class I expression and progression or regression of the metastatic lesions.32

Lack of response to immunotherapy and the development of progressing metastases in cancer patients seem to be associated with immune selection of HLA-deficient tumor cell variants with irreversible structural defects, including mutations in the β2m gene and LOH in chromosomes 6 and 15.30, 32, 72 In this context, we have previously reported that the poor clinical response to vaccination with HLA-A1 restricted MAGE-encoded peptides in two melanoma patients was correlated with the presence of β2m gene mutations and the loss of HLA Class I surface expression in the tumor tissues and tumor cell lines.30 Similarly, another melanoma patient who did not respond to immunotherapy with IFN-α had a total loss of HLA Class I expression due to a two successive mutational events affecting both copies of β2m gene along with LOH in chromosome 15.31

Our observations suggest that no matter what type of immunotherapy is administered, specific immunization or systemic immune stimulation, the outcome should be expected the same, namely, a modification of tumor microenvironment of each metastatic lesion leading to a release of immunostimulating cytokines.32 Depending on the preexisting HLA Class I expression level and the presence of reversible or irreversible alterations in the HLA Class I system, its expression will be either upregulated by the cytokines leading to regression of metastatic nodule or will remain unchanged due to resistance to cytokine treatment, which will stimulate progression of metastases with irreversible HLA Class I alterations (Fig. 2). Our prediction is that “soft” and “hard” HLA Class I tumor lesions will coexist during the natural history of tumor development. However, after immunotherapy tumor cells with “soft” lesion will be rejected and only those with “hard” lesions will prevail and will progress killing the host.

Figure 2.

Different types of immunotherapy can be administered to a patient or to a tumor-bearing host. The expected effect is to attract CTLs to the tumor microenvironment and to increase cytokine production. This newly induced “cytokine storm” will interact with tumor target increasing the MHC Class I expression only if the tumor cell have reversible defects. In that case, the lesion will regress. On the contrary, if tumor cell has irreversible defect, MHC Class I expression will remain low or absent and the lesion will progress. Correction of irreversible structural abnormalities can only be done with the transference of the functional MHC Class I heavy chain or β2m gene.

Recovery of normal HLA Class I expression in patients with local cancer relapses or metastatic progression with β2m mutations, LOH in chromosome 6 or other structural genetic HLA defects is essential. To achieve this goal, it will be necessary in the future to develop methods to restore normal MHC Class I expression in tumor cells with structural MHC Class I defects. This can only be achieved if a wild type MHC Class I heavy chain or β2m genes are delivered to the tumor lesion. In this context, we have developed a replication-deficient adenoviral vector carrying human β2m gene. In in vivo experiments using human tumor xenograft model, the intratumoral injection of β2m-carrying vector led to the restoration of normal HLA Class I expression.69 We propose that the adenoviral-mediated recovery or even increase of HLA Class I expression on tumor cells in combination with vaccination or adoptive T cell therapy can provide a successful approach to improve the clinical efficacy of cancer immunotherapy.

Concluding Remarks

Impaired expression of HLA Class I is frequent event in cancer.72 It abolishes tumor antigen presentation to T lymphocytes and limits the efficacy of cancer immunotherapy. Both reversible and irreversible structural defects of HLA Class I have been described in tumor cell lines, in animal models of cancer and in primary tumors and in metastatic lesions obtained from cancer patients. However, the clinical implication of this phenomenon is not well understood. Latest analysis of the response of cancer patients to immunotherapy revealed important observations demonstrating that two types of metastatic lesions develop in such cancer patients: regressing lesions with the capacity to upregulate MHC Class I expression in response to therapy and progressing lesions with altered MHC expression due to underlying irreversible alterations. We propose that the prevalence of “soft” or “hard” MHC lesions in the tumor cells is a key factor that will determine the future ability of T lymphocytes to destroy these malignant cells. On the basis of these observations, we favor the idea that HLA-deficient tumor escape variants with irreversible structural defects will grow independently of the type of immunotherapy used. It is known that during natural cancer evolution, tumor cells continuously acquire new genetic and phenotypic defects allowing them to escape antitumor immune responses. In addition, the metastatic potential of any primary tumor is altered by the genetic background on which it arises, by the tumor microenvironment and by treatment, including immunotherapy. Hence, we cannot exclude a possibility that the threshold of the immunotherapy-mediated activation might not be enough to upregulate HLA Class I expression even in case of regulatory alterations.

Our findings emphasize the importance of carefully defining the molecular mechanisms responsible for a particular altered HLA Class I phenotype to design specific ways to restore in situ normal tumor HLA Class I expression. Therefore, the optimization of the existing immunotherapy approaches will greatly benefit form the characterization of the HLA class I alterations in primary tumors and from the analysis of the correlation of impaired HLA expression with metastatic progression. We believe that therapies aimed at reexpression of HLA Class I antigens might improve the outcome of immunotherapy-based treatments. The ideal therapeutic protocol would include initiation of immunotherapy that would lead to rejection of tumor lesions with reversible alterations and, as a second step, gene therapy of progressing hard lesions with structural defects. To achieve this goal, the specific structural defects in HLA Class I system have to be identified. It must be acknowledged that personalized cancer management remains a distant goal. On the other hand, to date, considerable knowledge has been gained on the prevalence of various patterns of HLA Class I defects and the underlying molecular mechanisms in particular types of cancer.21 On the basis of this information, it is feasible to predict putative molecular mechanisms and to design a patient-tailored diagnostic strategy for HLA Class I alterations, at least in some cancer patients.