Defects in the Human Leukocyte Antigen Class I Antigen Processing Machinery in Head and Neck Squamous Cell Carcinoma: Association with Clinical Outcome

Effective CTL responses require adequate MHC class I antigen surface expression. This mechanism accounts for the negative impact of MHC class I antigen expression abnormalities, which are frequently found in many different types of human tumors, on tumor immunosurveillance, and in the course of the disease as well as on the outcome of T cell–based immunotherapies in malignancies (1–3). Abnormalities in the MHC class I phenotype of tumors are caused by distinct molecular defects, which interfere with several known steps within the antigen processing pathway. These include structural alterations, methylation, or dysregulation of genes coding for the MHC class I heavy chain and/or β₂-microglobulin as well as of those coding for antigen processing machinery (APM) components, such as IFN-γ inducible immunoproteasome subunits LMP2 and LMP7, the peptide transporter TAP, and tapasin (3–5).

Head and neck cancer of the oral cavity, pharynx, and larynx represents only about 6% of newly diagnosed cancers in the United States (6), but is more common in other parts of the world. Despite advances in therapeutic procedures, the benefit of surgery, radiotherapy, and chemotherapy for patients with head and neck squamous cell carcinoma (HNSCC) has not improved in the last three decades (7). The high rate of recurrence, grave functional implications of radical surgery resulting in morbidity and a poor 5-year-survival rate have emphasized the need for alternative therapies for HNSCC patients. T cell–based immunotherapy represents one such alternative, which aims at the induction and/or activation of human leukocyte antigen (HLA) class I antigen–restricted, tumor-associated antigen (TAA)–specific CTL (8, 9). Recent evidence suggests that HNSCC may be a suitable target for T cell–based immunotherapy (10).

One major prerequisite for the successful implementation of such an immunotherapy is the appropriate surface expression of MHC class I molecules on tumor cells. Thus, MHC class I antigen expression was studied in surgically removed HNSCC lesions using, in most cases, monoclonal antibodies (mAb), which recognize monomorphic determinants shared by various HLA class I molecules or in a limited number of samples of mAb which recognize allele-specific determinants (11–13). Aberrant HLA class I surface expression accompanied by a significant down-regulation or loss of certain HLA class I alleles (14) as well as by a loss of heterozygosity at chromosome 6p21 (15, 16) has been frequently found in head and neck cancer lesions. The HLA class I allelic loss in HNSCC lesions was significantly correlated with the presence of lymph node metastases, HLA class I defects in the adjacent normal mucosa and subsequent development of new primary cancers (14, 17, 18). In contrast, β₂-microglobulin mutations have not been detected in HNSCC with defective MHC class I antigen surface expression (19).

Altered HLA class I antigen surface expression is likely to play a major role in the escape of HNSCC cells from immune recognition. Furthermore, this mechanism may account for the difficulties in generating HLA class I antigen–restricted, TAA-specific CTL from peripheral blood mononuclear cells of patients with HNSCC (20–25). Therefore, characterization of the molecular defects underlying HLA class I antigen expression abnormalities in HNSCC cells is not only important from the biological viewpoint, but it may also have substantial clinical relevance. Nevertheless, only scanty information is available in the literature in this regard (26). In particular, the role of APM component abnormalities in defective HLA class I antigen surface expression rates by HNSCC cells has been investigated only to a limited extent (27). The lack of information about the mechanisms underlying HLA class I antigen abnormalities has a negative impact on our ability to design strategies to correct these defects in HNSCC lesions. Therefore, in the present study, we have examined a series of surgically removed HNSCC lesions and HNSCC cell lines for both the constitutive and IFN-γ-inducible expression levels of various APM components, such as proteasome subunits, TAP subunits, calnexin, tapasin, calnexin, HLA class I heavy chain, and β₂-microglobulin. In addition, we have correlated the APM component expression rates in HNSCC lesions with their histopathologic characteristics and with patients' survival. Thus, this study provides insights into molecular mechanisms underlying APM deficiencies in HNSCC lesions and, for the first time, assesses their association with patients' survival.

Cell lines, cell culture, and cytokine treatment. With the exception of the HNSCC cell line SCC-4, which was purchased from American Type Culture Collection (CRL-1624), all other tumor cell lines studied were established from patients with HNSCC at the University of Pittsburgh and have been described elsewhere (24, 27). The characteristics of the HNSCC cell lines in terms of tumor localization, stage, and grade as well as HLA phenotype determined by DNA typing are summarized in Table 1. The human keratinocyte cell line HaCaT (28) and the renal cell carcinoma cell line MZ1257RC (29) served as controls. All cell lines were maintained in DMEM (Biochrom KG, Berlin, Germany) supplemented with 10% FCS (Greiner, Munich, Germany), 2 mmol/L l-glutamine, 100 μg/mL streptomycin, and 100 units/mL penicillin (Life Technologies, Karlsruhe, Germany) at 37°C in an atmosphere of 5% CO₂ in air.

Patients and tissue samples. The study group consisted of a sample from all patients between the ages of 37 and 79 who had a histologically confirmed diagnosis of HNSCC at the Department of Oral and Maxillofacial Surgery at the Johannes Gutenberg University, Mainz, between 1999 and 2001. From a total of 75 patients meeting these criteria, a sample of 25 was randomly selected. Twenty of the 25 patients were treated with surgical resection for primary diseases; the remaining 5 underwent surgery for recurrent disease. Each tumor was histopathologically classified according to the WHO classification system. The TNM status, tumor stage, grade, localization, and lymph node involvement of all the studied patients are listed in Table 2.

Surgically removed biopsies were immediately snap-frozen in liquid nitrogen and stored at −70°C for immunohistochemical staining. In addition, paraffin-embedded HNSCC tissues were available from which autologous tumor cell lines had been established. Tonsils were used as control tissues. The study was approved by the Institutional Review Committee at the University of Mainz. Informed consent was obtained from all patients enrolled in this study.

Monoclonal and polyclonal antibodies. The mouse mAb W6/32, which recognizes a framework determinant expressed on β₂-microglobulin associated HLA-A, -B, and -C heavy chains, the mAb HC10 recognizing HLA class I heavy chains, the mouse anti-TAP1 mAb 148.3, the mouse anti-TAP2 mAb 429.4, and the mouse anti-tapasin mAb have been described elsewhere (30–33). The mouse anti-PA28, mouse anti-calnexin, and mouse anti-calreticulin as well as the rabbit anti-LMP2 and anti-LMP7 antibodies were purchased from Affinity (Mamhead, United Kingdom). FITC-conjugated anti-HLA-A, -B, -C mAb, an IgG_2a, and an irrelevant FITC-conjugated mAb IgG_2a were purchased from Beckman/Coulter, Krefeld, Germany. Alkaline phosphatase-conjugated rabbit anti-mouse immunoglobulin antibodies and peroxidase-conjugated goat anti-rabbit immunoglobulin antibodies were purchased from Dako (Hamburg, Germany). All antibodies were stored at 4°C or at −20°C and were diluted with sterile PBS (pH 7.2) before use.

Human recombinant IFN-γ. IFN-γ (specific activity of 3 × 10⁶ units/mg protein) was purchased from Boehringer Ingelheim, Ingelheim, Germany.

RT-PCR analyses. Total cellular RNA was extracted using the Qiagen RNeasy kit (Qiagen, Hilden, Germany) followed by DNase digestion according to the manufacturer's instructions. The different primer sets for APM components used for PCR and the PCR conditions employed were recently described elsewhere (30, 34).

Western blot analyses. Protein was extracted from 1 to 5 × 10⁶ HNSCC cells which were either left untreated or treated with 20 μg/mL IFN-γ for 24 hours at 37°C. Protein (40 μg per lane) was analyzed by Western blots analyses as recently described (30). Filters were incubated overnight with the anti-human (mAb) specific for TAP1 (mAb 148.3; ref. 31), TAP2 (mAb 429.4; ref. 33), PA28α, calnexin, calreticulin, or the polyclonal rabbit anti-human LMP2 antibody (Affinity). After extensive washing with TBS/0.1% Tween 20, membranes were incubated for 30 to 45 minutes either with a horseradish peroxidase–conjugated rabbit anti-mouse immunoglobulin antibody (Dako) for TAP1, TAP2, PA28α, calnexin, calreticulin, or with a peroxidase-conjugated goat anti-rabbit immunoglobulin antibody as secondary antibody (Dako) for LMP2. After rinsing twice, filters were developed using nitroblue tetrazolium/X-phosphate (Roche Diagnostics, Mannheim, Germany) or the enhanced chemiluminescence system (ECL, Amersham, Braunschweig, Germany) as substrate.

Flow cytometry. Indirect immunofluorescence staining of cells was done as previously described (5). Briefly, 5 × 10⁵ cells were incubated with the appropriate concentration of FITC-conjugated anti-HLA-A, -B, -C mAb for 30 minutes at 4°C. An irrelevant FITC-conjugated mAb IgG_2a served as a control. Stained cells were analyzed on a flow cytometer (Epics XL MCL, Beckman/Coulter). The results were expressed as mean specific fluorescence intensity.

Immunohistochemistry. Snap-frozen HNSCC biopsies embedded in the OCT compound (Tissue-Tek, Sakura, Tokyo, Japan) as well as paraffin-embedded HNSCC tissues from which tumor cell lines were available were serially cut into 5 μm sections on a Leica CM1900 cryostat and mounted on electrostatically precharged slides (Superfrost Plus; Menzel-Gläser, Frankfurt, Germany). Consecutive sections from paraffin-embedded tissues were deparaffinized and rehydrated. After endogenous peroxidase quenching (3% H₂O₂ in PBS for 30 minutes), antigens were retrieved by microwave treatment.

Immunohistochemical staining was done at room temperature using the avidin-biotin technique with mAb and rabbit antibodies. Optimal working dilutions of the primary antibodies were determined in titration experiments using human tonsils. Staining was developed with the standard streptavidin-biotin-immunoperoxidase technique (LSAB 2 system-HRP, Dako) and visualized with 3,3′-diaminobenzidine tetrahydrochloride (Dako). The slides were counterstained with hematoxylin (Merck, Darmstadt, Germany) and mounted after dehydration in Eukitt (Kindler, Freiburg, Germany). All immunostained sections were coded and separately examined in a light microscope by two investigators (T.E. Reichert and M. Kunkel). Natural stromal cells and tumor-infiltrating lymphocytes were used as positive controls. Results were scored as positive, heterogeneous, or negative, when the percentage of stained tumor cells in each section was >75%, between 25% and 75%, and <25%, respectively. The staining intensity of nonmalignant oral epithelium adjacent to the carcinoma was used as a reference to define strong and weak staining intensity of HNSCC cells. Negative controls were done by omitting primary antibodies.

Statistical analysis. Overall survival and presence of disease recurrence were analyzed by a series of nonparametric tests for each of the five markers analyzed (LMP2, LMP7, TAP1, TAP2, and HLA class I antigens). The analysis was limited to 19 patients treated for primary disease having adequate follow-up data. Five patients treated with recurrent disease and one patient with unknown survival data were excluded.

(a) The expression level scores were dichotomized by combining scores of 0 and 1 or 1 and 2, depending on the score frequencies. A log rank test was applied to the dichotomized data. (b) The expression level scores were used as is and tested with an exact log rank test. Monte-Carlo sampling was used to simulate the permutation distribution. (c) Patients were classified as either disease-free or with recurrent disease 2 years after surgery. The Cochran-Armitage trend test was then used to identify a trend in the proportion of patients with recurrent disease by expression level score. All tests resulting in P values were adjusted by the step-down Bonferroni procedure.

Peptide transport assay. The peptide translocation experiments were done as recently described using the [¹²⁵I]-labeled model peptides, RYWANATRSI and TNKTRIDGQY (35) which were kindly provided by F. Momburg (Deutsches Krebsforschungszentrum, Heidelberg, Germany).

Antigen processing machinery component and human leukocyte antigen class I antigen down-regulation in head and neck squamous cell carcinoma lesions. The histopathologic characteristics of 25 surgically removed tumor lesions obtained from patients with either primary, recurrent, or metastatic HNSCC as well as the patients' survival time are summarized in Table 2. Immunohistochemical staining of snap-frozen tumor lesions and autologous normal epithelium of HNSCC patients with APM component–specific antibodies showed a highly variable expression profile for LMP2, LMP7, TAP1, TAP2, and HLA class I antigens (Figs. 1 and 2) ranging from a total loss to heterogeneous but decreased expression levels. Specifically, LMP2 was not detectable in 16% and down-regulated in 72% of the lesions. LMP7 was not detectable in 16% and down-regulated in 44% of the lesions. TAP1 was not detectable in 28% and down-regulated in 24% of the lesions. TAP2 was not detectable in 32% and down-regulated in 56% of the lesions. HLA class I antigens were not detectable in 12% and down-regulated in 36% of the lesions (Table 2). The three tumors with a total HLA class I antigen loss also lacked TAP protein expression, suggesting that in these lesions, deficient peptide translocation may be the molecular mechanism underlying this lack of HLA class I surface expression. A combined deficiency of various APM components was found in 80% of the tumor specimens analyzed. A coordinated down-regulation of the various proteasome subunits, TAP, and HLA class I antigens was the most frequent phenotype of the tumors studied (Table 2).

Additional analyses assessed the clinical significance of APM component down-regulation patterns in HNSCC lesions. For this study, only 19 lesions were utilized, since 5 tumor lesions were from surgically resected recurrent disease, and no information was available about the survival of one HNSCC patient. Clinical, pathologic, and survival data obtained from the 19 HNSCC patients were examined for association with the expression of LMP2, LMP7, TAP1, TAP2 and HLA class I antigens using three nonparametric tests for each of these five markers. The expression of all five markers was associated with patients' overall survival (Fig. 3); the association was stronger for LMP2, LMP7, and TAP2 expression with adjusted P values according to the two-group log rank test of 0.005, 0.045, and 0.0044, respectively than for TAP1 and HLA class I antigen expression with adjusted P values of 0.0775 and 0.0580, respectively (Table 3). In summary, the high expression levels of the five markers analyzed were consistent with improved patients' survival (Fig. 3). The association of APM component expression pattern with disease recurrence was less clear except for a statistically significant association of LMP7 down-regulation (P value, 0.0395) with disease recurrence at 2 years (Table 3).

Impaired expression and function of antigen processing machinery components in cell lines derived from head and neck squamous cell carcinoma lesions. In order to assess functional consequences of impaired APM gene expression rates observed in tumor biopsies, we analyzed 10 HNSCC cell lines for APM component mRNA and protein expression levels. The keratinocyte cell line HaCaT and the renal cell carcinoma cell line MZ1257RC, known to have a normal HLA class I phenotype, served as controls. When compared with HaCaT cells, all HNSCC cell lines displayed reduced levels of total HLA class I surface antigens, which was accompanied by heterogeneous mRNA and protein expression levels of the various APM components analyzed (Fig. 4). In the case of PC-15A/B cells, which exhibit a low HLA class I surface expression, it could not be determined whether the patient is HLA homozygous or whether there exists a HLA haplotype loss in the tumor cell, since normal cells from this patient were not available for such an analysis. Furthermore, highly variable TAP1 protein levels were often correlated with impaired TAP function as measured by the peptide translocation assay (Fig. 5). PCI-13 and PCI-15A cells displayed peptide transport rates ranging from 2.1% to 5.8%. The peptide transport rate in both HNSCC cell lines was dependent on the model peptides employed, since peptide #63 was more efficiently transported into the endoplasmic reticulum than peptide #600.

Antigen processing machinery component and human leukocyte antigen class I antigen down-regulation in both head and neck squamous cell carcinoma lesions and autologous cell lines. To determine whether the levels of the APM components and of the HLA class I antigen surface expression in HNSCC cell lines correlated with those in the lesions from which the cell lines had been originated, we compared LMP2, LMP7, TAP1, TAP2, and HLA class I heavy chain expression levels in three formalin-fixed, paraffin-embedded HNSCC lesions to that of the autologous HNSCC cell lines PCI-13, PCI-30, and PCI-52 (Table 4). With the exception of the HLA class I heavy chain expression in PCI-13 cells, the expression levels of APM components and HLA class I heavy chains in the three HNSCC lesions were similar to those found in the corresponding cell lines as determined by RT-PCR and Western blot analyses (Fig. 4).

IFN-γ-mediated restoration of antigen processing machinery component expression in head and neck squamous cell carcinoma cell lines. It is generally accepted that IFN-γ transcriptionally modulates the expression of most APM components, including the proteasome subunits LMP2, LMP7, and LMP10/MECL-1 as well as their activators PA28α and β, next to TAP, tapasin, and HLA class I antigens (3, 5, 36). To provide insights into the underlying mechanism(s) of APM component abnormalities in HNSCC cells, such as structural alterations or deregulation, APM component expression levels were analyzed by flow cytometry, RT-PCR, and/or Western blot in the 10 HNSCC cell lines incubated in the absence or presence of IFN-γ (200 units/mL for 24 hours at 37°C). As shown in Fig. 4, PA28α, LMP2, TAP, and tapasin expression was induced by IFN-γ, whereas calnexin and calreticulin transcription and/or translation were not up-regulated by this treatment. The IFN-γ-mediated increased expression of PA28α, LMP, TAP, and tapasin was directly associated with an enhanced HLA class I antigen expression (Table 1).

Abnormalities in MHC class I antigen surface expression and down-regulation of different APM components have been identified in a substantial number of human tumors of distinct histology (3) and may be related to the metastatic potential of some tumor types (37, 38). In HNSCC patients, MHC class I alterations were found in up to 50% of the tumor lesions (11, 14, 15, 18). Although the underlying molecular mechanisms of deficient HLA class I antigen expression rates in HNSCC cells have not yet been determined in detail, microsatellite instability analyses showed loss of heterozygosity of chromosome 6p21 in HNSCC lesions (16, 39). Here, we extended these findings to other HLA class I molecules and APM components showing a high frequency of down-regulation or even loss of LMP2, LMP7, TAP1, and HLA class I antigens in both tumor biopsies as well as cell lines of HNSCC patients when compared with normal control cells (Fig. 2; Tables 1 and 2). A similar coordinated down-regulation of APM components in tumor cell lines has been described by Johnson et al. (40). In addition, down-regulation of one or more APM components is often associated with deficient HLA class I antigen surface expression (41).

It is noteworthy that APM abnormalities in HNSCC seem to be mainly attributable to deregulation rather than to structural gene abnormalities, since normalization of APM component expression could be induced by IFN-γ treatment (Table 1). This is further underlined by the fact that structural alterations of APM components, such as TAP, seem to be a rather rare event (41–43). However, detailed sequence analyses of TAP as well as other APM components are needed to better understand the molecular basis of escape mechanisms utilized by human malignancies, including HNSCC.

Impaired APM component expression might influence both the induction and the effector phase of TAA-specific immune responses. Indeed, abnormalities of MHC class I surface expression attributable to APM deficiencies enable tumors to evade immunosurveillance and thereby developing a more aggressive phenotype (3). Such an immune escape mechanism has also been suggested for HNSCC cells and might be reflected by a low T cell infiltration of tumors or by inefficient activation and generation of HNSCC-associated antigen–specific CTL as described by several investigators (20, 22, 25). Restoration of antigen presentation by cytokine treatment and/or gene transfer of deficient APM component consequently renders cells sensitive to CTL-mediated lysis. These strategies should be considered for the development of novel T cell–based immunotherapies for HNSCC patients. In this context, it is noteworthy that other mechanisms besides APM down-regulation may be involved in selective HLA losses. Indeed, Maleno et al. (44) described multiple mechanisms of HLA class I down-regulation pattern in laryngeal carcinomas. These include a high frequency of haplotype loss associated with loss of heterozygosity as well as total HLA class I loss, HLA-A and -B locus–specific down-regulation and HLA class I allelic loss. Such structural alterations cannot be restored by cytokine treatment.

It has been suggested that the survival of cancer patients might be influenced by the functional state of their immune system. Clearly, the ability of the host immune system to control tumor growth is likely to be impaired in the presence of HLA antigen expression abnormalities. To the best of our knowledge, only a few studies have examined the correlation between loss of APM component expression and clinical outcome (26, 37, 38, 43). In melanoma, breast, lung, and cervical carcinoma, the lack of HLA class I antigen surface expression and/or TAP expression was reported to be associated with poor tumor differentiation, increased frequency of aneuploidy as well as reduced survival rate (37, 38, 45, 46). Conflicting information is available about the association between HLA class I antigen down-regulation and patients' survival in HNSCC patients (25, 26, 47). Our results are in accordance with those reported by one of the authors (26) showing a significant correlation between the survival time and a loss or down-regulation of APM components in primary maxillary sinus squamous cell carcinoma lesions. Although preliminary, our results suggest that the APM component levels might serve as valuable prognostic markers in this disease (Fig. 3). However, our study has limitations including the small number of HNSCC patients analyzed thus far and the fact that preoperative radiation was used in several of our patients, which might affect the pathologic stages as well as clinical responses and patient survival. In addition, the node stage, which is considered to be the best predictor of survival or recurrence in HNSCC patients, could not be correlated with survival or recurrence in this cohort of patients (48). To further confirm the intriguing possibility that the APM components could serve as independent predictors of survival in HNSCC patients, it will be necessary to perform a larger retrospective or prospective study specifically designed to measure survival. Nevertheless, the observed correlation between deficiencies in APM components and survival represents a novel finding suggesting that the host's immune system/tumor cell interactions determine the disease outcome and that long-term prognosis depends on these interactions.

We thank Dr. Robert Tampé (University of Frankfurt, Frankfurt, Germany) for his generous gift of antibodies, M. Post (German Cancer Research Center, Heidelberg, Germany) for technical assistance, and I. Schmidt and Liis Pellenen for help in the preparation of this manuscript.