Abstract
The expression of human leukocyte antigen (HLA) class I molecules on the cell surface is necessary for the presentation of peptide antigens to cytotoxic CD8+ T lymphocytes of the immune system. Down-regulation of HLA class I gene expression has been implicated in tumorigenesis, including squamous cell carcinoma of the head and neck (SCCHN). Loss of MHC class I antigens may be one mechanism by which tumor cells escape immune detection. We performed prospective immunostaining of 26 primary SCCHN tumors and samples of normal mucosa harvested several centimeters away from the primary tumor, using a large panel of antibodies directed against allele-specific as well as monomorphic determinants of HLA class I molecules. Loss of expression of HLA class I proteins in the tumor was found in 50% (13 of 26) of primary tumors and was highly correlated with HLA loss in the corresponding normal mucosa (P < 0.0001). Further analysis demonstrated that the loss of HLA class I expression in the tumor was significantly associated with regional lymph node metastases (nodal stage; P = 0.0388), and that the number of HLA class I alleles lost in the normal mucosa was associated with subsequent development of a new primary aerodigestive tract cancer (P = 0.042). A patient with two metachronous cancers available for analysis had no evidence of HLA loss in the first tumor, demonstrated allelic loss in the second cancer, and subsequently died of disease. These results suggest that the loss of expression of HLA class I alleles may have prognostic implications.
INTRODUCTION
The immune response against tumors is, in many cases, dependent on an efficient CTL response. CTLs recognize antigenic peptides bound in the cleft of MHC class I molecules and can be stimulated to kill cells expressing peptides for which they are specific. During the course of tumor establishment, cancers evolve strategies to evade the specific CTL response. These strategies have included the secretion of immunosuppressive cytokines, such as transforming growth factorβ (1) , and the induction of apoptosis via tumor cell expression of Fas ligand (2) .
It has been known for some time that certain tumors can down-regulate or lose the expression of MHC class I proteins (3) . HLAs2 are cell surface glycoproteins, encoded by a sequence of genes, which are inherited as codominant alleles. Decreased expression of HLA class I molecules by malignant cells may be an effective strategy for evading host immunosurveillance. Several potential mechanisms for MHC class I loss have been identified, including the loss of expression of the TAP proteins, which are important in the MHC class I antigen-processing pathway, resulting in the tumor cell losing expression of all MHC class I alleles (4) . Global loss of MHC class I alleles occurs infrequently in human tumors, which may be explained by expression of killer inhibitory receptors on NK cells (5) . Killer inhibitory receptor molecules recognize MHC class I, and, when they engage their ligand, NK cell-mediated lysis is inhibited (6 , 7) . Therefore, a cell that is totally devoid of MHC class I expression may be able to evade specific CTL recognition but would remain a good NK cell target. Some tumors seem to lose expression of single HLA-A or B alleles while retaining the expression of the other HLA alleles (3 , 8 , 9) . Such a partial HLA loss may be advantageous to the tumor because it potentially allows for the escape of tumor cells from CTLs directed against immunodominant epitopes, presented by the lost HLA allele, while maintaining resistance to NK cell lysis. Several studies have shown that MHC class I allelic loss may occur frequently in human malignancy, being reported in almost 90% of breast cancers (10) and 85% of prostate tumors (11) . Although data are limited, in some cases, the loss of HLA expression has been correlated with tumor aggressiveness and metastatic potential (12) .
SCCHN is the most common malignancy of the upper aerodigestive tract. Individuals who develop one head and neck tumor are at high risk for additional cancer formation at the rate of 3–6% per year (13) . Patients who develop a second aerodigestive tract cancer demonstrate only a 25% five-year survival (14) . Because the histologically normal mucosa of SCCHN patients is at risk for tumor development, genetic alterations in the aerodigestive tract mucosa are likely to represent early events in head and neck carcinogenesis. Therapies that target these alterations may effectively prevent tumor formation. Previous studies examining HLA expression in SCCHN specimens have chiefly used antibodies recognizing monomorphic determinants shared by HLA class I molecules (15, 16, 17) . In general, one-third to one-half of head and neck tumors demonstrate aberrant HLA class I expression (18, 19, 20) . The recent availability of allele-specific antibodies allows a more detailed analysis. Using a wide panel of allele-specific antibodies in conjunction with the HLA and Cancer Committee of the 12th International Histocompatibility Workshop and Conference, we investigated the association between HLA class I loss in the primary SCCHN tumor and derangements of HLA expression in the uninvolved normal mucosa from SCCHN patients with established clinical and pathological parameters.
MATERIALS AND METHODS
Patients and Tissue Samples.
All of the patients were treated at the University of Pittsburgh Medical Center. Tissue and blood were harvested under the auspices of the head and neck cancer tissue bank protocol, which was approved by the Institutional Review Board. All of the patients meeting the following criteria were selected for study: (a) a histologically confirmed diagnosis of SCCHN (oral cavity, oropharynx, hypopharynx, or larynx); (b) primary surgical resection with curative intent occurring during the period from August 1994 through November 1995; (c) absence of distant metastases; and (d) availability of PBMCs and tumor tissue snap-frozen at the time of resection (i.e., neither formalin-fixed nor paraffin-embedded). Thirty-one patients met entry criteria. Of the participating patients, complete HLA typing could not be completed on 4 patients, and there was no tumor detected in the tissue harvested from 1 patient, which made 26 patients available for complete analysis. One patient developed a second head and neck cancer during the course of the study, which was included for immunohistochemical staining, for a total of 27 tumors. Normal tissue was not available for two patients. Twenty-four patients had immunohistochemical analysis of HLA expression of both tumor and corresponding normal mucosa. Clinical follow-up was available for all of the patients until May 1999. Clinical follow-up was generally performed every 2–3 months for the first 2 years and every 6 months thereafter. Pertinent patient information was abstracted from the computerized head and neck tumor registry.
All of the patients were treated with curative intent. Each underwent complete surgical resection of the primary tumor with negative surgical margins and 23 (88.5%) of the 26 patients underwent dissection of the cervical lymph nodes with pathological staging of the regional lymphatics (N stage). Clinical staging was conducted in accordance with the American Joint Committee on Cancer tumor-lymph node-metastasis (TNM) classification at the time of surgery (21) . Of the 26 patients, 2 (8%) were initially treated with radiation therapy followed by surgery, 18 (69%) received postoperative external beam radiation, and 4 (15%) were treated with adjuvant chemotherapy. The vital status of each patient was classified as: alive, dead of disease, or dead of other causes. All of the patients who were alive at the end of the follow-up period had no evidence of disease. The clinical and pathological characteristics of the 26 patients are outlined in Table 1⇓ .
Clinicopathological characteristics of 26 head and neck cancer patients evaluated for HLA typing
HLA Typing.
PBMCs were isolated from each patient by centrifugation on a Ficoll Hypaque gradient. Serological typing for HLA-A, -B, and -C was performed on the PBMCs using the standard microlymphocytotoxicity assay with commercially available serum trays (One Lambda, Canoga Park, CA).
Immunohistochemical Analysis.
At the time of surgery, part of the tumor and a sample of normal mucosa harvested several centimeters away from the tumor were snap-frozen in liquid nitrogen and stored until use. Sections (7-μm) were cut from each tumor and normal mucosa sample, air-dried for 4–18 h, and fixed in acetone for 10 min. The slides were stored at 4°C until further use. The slides were stained for the expression of HLA class I alleles using the Vectastain ABC kit (Vector laboratories, Burlingame, CA). Briefly, the slides were first rinsed in PBS, preincubated for 30 min at room temperature with 1.5% normal horse serum, and then incubated with the test antibody for 1 h at room temperature. After washing, the slides were incubated with biotinylated goat antimouse IgG for 30 min at room temperature and were washed and further incubated with avidin-peroxidase (Vectastatin ABC reagent) for 1 h. The slides were washed and incubated with the enzyme substrate (DAB, Dako). The slides were washed and counterstained with Meyer’s H&E stain (Sigma).
The mAbs were initially tested for suitability for use in immunohistochemical analysis and used in the 12th International Histocompatibility Workshop (22) . These mAbs were pan-HLA class I-specific, locus-specific, or allele-specific, and many of the common HLA-A alleles could be identified with these antibodies. There were fewer HLA-B-specific mAbs, but anti-Bw4 or anti-Bw6-specific mAbs were used facilitating discrimination in cases in which the individual was Bw4/Bw6 heterozygous. A listing of the antibodies used is provided in Table 2⇓ . An additional mAb (16.1)—a mouse IgM mAb kindly provided by Dr. Marilyn Marrari (University of Pittsburgh, Pittsburgh, PA)—was used to detect the HLA-A1 and -A11 specificities. Negative control antibodies included isotype-matched murine mAbs, and, in addition, a mAb was included of a specificity that was known, from the HLA typing, not to be expressed by that particular patient.
mAbs used in immunohistochemical analysis
The slides were read in a blinded fashion by two individuals, (J. R. G., M. F. M.) including a pathologist, and were scored according to the criteria set up by the 12th International Histocompatibility Workshop (22) . The tissue (tumor or normal mucosa) was scored as negative for a particular allele only if staining of the surrounding stroma could also be visualized, which demonstrated that the lack of staining in the target tissue was specific.
Statistical Analysis.
Statistical analysis was performed on the entire patient population. Survival was measured in months from the date of surgery to the date of death or to the last follow-up. Disease-free survival was defined as the time from tumor resection until the first evidence of tumor recurrence or the development of a new primary tumor of the upper aerodigestive tract. Cause-specific survival was based on death attributable to the progression of the head and neck cancer as distinct from death attributable to other causes. All of the surgical resections were considered curative rather than palliative. Prognostic covariates included in the analysis were sex, age, tumor site, tumor grade, stage lymph node stage, presence of ECS, and the loss of HLA class I antigen expression in the tumor and normal mucosa. For statistical analysis, loss of class I antigen expression was classified as the loss of none, one, or two or more alleles. Allele loss in the tumor was cross-classified with allele loss in normal tissue, and each was cross-classified with the clinical and pathological variables: T-stage, N-stage, ECS, recurrence of original tumor, or growth of a new primary tumor. Each pair of cross-classified variables was tested for independence as well as for specific hypotheses appropriate for the levels of measurement, including tests of symmetry or trend (23) . All of the statistical tests of cross-classified data were derived from the exact permutation distribution of the test statistic using the software package Statxact (24) . Differences in progression-free survival and cause-specific overall survival attributable to allele loss were examined by the log rank statistic.
RESULTS
Loss of Expression of HLA-A or -B Alleles Is Common in SCCHN Tumors.
Full HLA typing was obtained on 26 of the 31 patients, and the HLA types are shown in Table 3⇓ . The frequency of specific HLA-A or -B alleles in this group of patients was similar to that in previously reported cohorts of normal Caucasian controls (25) . Immunohistochemical analysis of tumor tissue from SCCHN patients revealed no complete losses of MHC class I expression, because staining with pan-specific mAb or with mAb specific forβ 2-microglobulin was positive in all of the cases. Loss of expression of one or more alleles at the HLA-A or -B locus was observed in 50% (13 of 26) of the primary tumors tested (Table 3)⇓ . Staining was heterogeneous, and the tissue was considered to express the class I allele if >25% of the epithelial cells (normal or transformed) stained with the allele-specific antibody (Fig. 1)⇓ . The most common form of loss was that of a single HLA-A allele, which was observed in 54% (7 of 13) of the primary tumors that exhibited loss of HLA expression. Of the remaining tumors, two had lost expression of a single HLA-B allele, two appeared to exhibit a haplotype loss with lack of expression of one HLA-A and one HLA-B allele, and the remaining two tumors had lost expression of two HLA-A alleles in conjunction with a single HLA-B allele. Four (57%) of the seven HLA-A11-positive individuals had lost expression of A11 on their tumor, whereas only two (25%) of the eight A2-positive individuals had lost tumor expression of A2.
HLA class I allelic loss in a representative SCCHN patient. A, immunohistochemical analysis of HLA class I expression in the tumor using a mAb specific for monomorphic determinants (W6/32) demonstrating strong tumor staining. B, the same tumor stained with the Bw4-specific mAb, 116-5-2, demonstrating loss of HLA-B57 expression. No staining is seen in the same sampled tumor (C) or normal mucosa (D) stained with negative control antibody (IgG).
HLA typing and allelic loss in the primary tumor (T) and normal mucosa (N) for 26 patients
Loss of HLA Class I Expression in the Tumor Is Associated with HLA Loss in the Normal Mucosa.
Genetic alterations found in normal mucosa from SCCHN patients are generally considered early events in head and neck carcinogenesis. Histologically normal mucosa, harvested several centimeters away from the primary tumor, was obtained in most cases (24 of 26). Of the 13 primary tumors demonstrating HLA class I loss, normal mucosa was available in 11 cases, and 8 (73%) of 11 also showed loss of the same allele(s) in the normal tissue (Fig. 2)⇓ . Statistical analysis demonstrated that HLA class I loss in the tumor was highly correlated with HLA loss in the corresponding normal mucosa from the same patient (P < 0.0001; Table 4⇓ ).
Loss of HLA class I alleles in the SCCHN tumor and corresponding normal mucosa. A, immunohistochemical staining of a representative tumor and B, normal mucosa using mAb specific for monomorphic determinates (W6/32) showing strong staining. C, staining of the same patient’s tumor and (D) normal mucosa using mAb specific for A2A28, demonstrating loss of allelic expression.
Allele loss in the primary tumor and normal mucosa for 24 head and neck cancer patientsa
HLA Allelic Loss Is Tumor-Specific and Loss in the Normal Mucosa Correlates with Subsequent Development of a New Primary Cancer.
All of the patients underwent curative surgical resection as treatment for their tumors. A second primary tumor of the upper aerodigestive tract developed in 7 patients during the course of the study period, and 9 patients developed local or regional recurrence or distant metastases. Three patients had both a recurrence of their original tumor and the development of a new primary aerodigestive tract tumor. Two primary metachronous tumors from one patient were available for analysis. In this case, there was no loss of HLA class I allele expression in the first tumor. However, the second tumor, arising 15 months later, demonstrated loss of HLA-A3, and the patient succumbed to that cancer (Fig. 3)⇓ . Of the 26 patients entered in the study, 14 died—12 from their head and neck cancer and 2 from other causes (i.e., not as a result of carcinoma of the upper aerodigestive tract). The median follow-up for the 12 surviving patients was 41 months (range, 20–50 months).
HLA class I loss is tumor-specific. A, representative staining with the HLA-A3-specific mAb, 160-30 of the initial SCCHN tumor demonstrating expression of HLA-A3 in the first tumor. B, staining of the second primary tumor 15 months later in the same patient showing loss of expression of A3.
We investigated whether HLA class I allelic loss in the tumor (or normal mucosa) correlated with clinical and pathological parameters. HLA loss in the tumor or normal mucosa was not associated with tumor stage, ECS, recurrence of the primary tumor, disease-free survival, or overall survival. However, loss of HLA class I expression in the tumor was significantly associated with nodal stage (P = 0.0388). The degree of allelic loss in the normal mucosa was associated with the development of a new primary head and neck cancer (Table 5)⇓ . When allele loss was coded as 0, 1, or 2-or-more alleles lost, the exact χ2 test of independence was significant (P = 0.0075) as was the Mantel trend test (P = 0.0421), which suggested that the risk of developing a new primary tumor increases with the number of alleles lost in the “at risk” normal mucosa.
Allele loss in normal tissue and disease progression for 24 head and neck cancer patients
DISCUSSION
Loss of HLA antigen expression is considered to be one of the mechanisms whereby tumor cells escape immune surveillance. Many human tumors appear to lose expression of one or more HLA alleles. In most cases, the loss is restricted to a single HLA-A or HLA-B allele and rarely involves a total loss of HLA class I expression. We found that 50% of SCCHN tumors demonstrated HLA allelic loss that was highly correlated with loss of HLA expression in the normal mucosa harvested from the same patient. An earlier report examined MHC class I and β2 microglobulin loss in premalignant oral mucosa and found no decrease in HLA expression in tumors or dysplasias on the same specimen (26) . Our findings of allelic loss in both tumors and normal mucosa may be explained, in part, by use of a large panel of recently available allele-specific antibodies. None of the patients in our series had lost expression of all HLA alleles, which demonstrates the importance of assessing HLA expression with allele-specific antibodies. The observed HLA losses in this series would not have been detected using mAbs specific for monomorphic HLA determinants or forβ 2-microglobulin. A recent report confirmed that the total loss of HLA class I and/orβ 2-microglobulin is not detected in SCCHN tumors (27) . Similarly, loss ofβ 2-microglobulin was not detected in renal cell carcinomas nor in corresponding nontransformed renal tissue (28) . These data probably reflect the fact that tumor cells deficient in MHC class I expression become highly susceptible to NK cell lysis and are eliminated.
Previous studies have reported conflicting data regarding the association of HLA class I allelic loss and clinical outcome in cancer patients. In breast and bronchogenic tumors, the loss of HLA expression has been found to be associated with poor differentiation of the tumor and an increased frequency of aneuploidy (10 , 12 , 29) . These results suggest that the loss of HLA expression enables the tumor to evade the host immune response and to develop more aggressive characteristics. Previous studies of HLA loss in head and neck cancer have revealed a relatively high frequency of allelic loss that did not correlate with survival (16 , 18 , 20) . Others have reported an association between HLA class I expression and degree of differentiation, T stage, and/or degree of leukocyte infiltration into the head and neck tumor (17 , 30) . Cancer survival is multifactorial and unlikely to be determined by a single genetic alteration. Our results demonstrate that allelic HLA loss in the tumor or normal mucosa did not correlate with either overall or disease-free survival. The association between HLA loss in the primary head and neck tumor with the presence and extent of regional lymph node metastases (N stage) suggests that loss of HLA class I alleles may contribute to metastatic potential. Such an association has also been demonstrated in cervical cancer (8) . In studies in which both tumor tissue and metastatic lymph nodes were available for analysis, the degree of HLA loss in the draining lymph nodes was greater than that observed in the primary tumor (4 , 10 , 31) . The higher incidence of antigenic loss in the metastatic tumors suggests that recognition of HLA class I antigens by the host immune system could contribute to solid tumor metastasis.
Earlier reports suggest an association between certain HLA loci and the risk of developing squamous cell carcinoma (32) . In the present study, we have confirmed the moderately high frequency (50%) of HLA allelic loss in SCCHN. Such a loss may influence the induction or cytotoxicity of CTLs in their effector phase. The molecular basis for reduced HLA class I expression is incompletely understood. Prior studies have failed to detect mutations or gene rearrangements of MHC class I genes orβ 2-microglobulin in human tumors although a mutated β2-microglobulin gene has been reported in a colon cancer cell line (12 , 15 , 27 , 33) . By examining tumor tissue and corresponding early passage cell lines, Giacomini et al. (34) were able to determine that loss at the protein level was not accompanied by complete, selective loss of HLA-A or HLA-B allomorphs. Peptide-binding by MHC class I molecules has been shown to be dependent on the cytosolic interaction with TAP (35) . Although we did not examine expression of TAP proteins, data from earlier studies suggest that down-regulation of TAP-1/TAP-2 may contribute to deficient antigen processing in lung or head and neck cancers (27 , 36) . The high proportion of HLA-A11-positive individuals who have lost expression of this allele in their tumor suggests that this allele may be an important restriction element for cytotoxic T cells specific for squamous cell carcinoma antigens. Few specific antigens have been identified as targets in squamous cell carcinomas, although there have been several reports of CTLs specific for these tumors (37 , 38) . These CTLs, derived from individual patients, were found to be restricted by HLA-A2 or -B44, but the specific antigens recognized by these cells have not been specifically identified (39 , 40) . Our results suggest that HLA-A11 may bind a tumor-derived peptide that can drive an effective CTL response. A more comprehensive study of tumors is necessary to determine the relative role, if any, of HLA-A11 in the presentation of peptides derived from SCCHN cells.
Few studies have examined HLA class I expression in premalignant lesions or tissue at high risk for tumor formation. In both familial and sporadic colon carcinogenesis, HLA antigens are reduced in adenomas, as well as in histologically normal mucosa distant from the adenoma (41, 42, 43) . Decreased expression of MHC class I alleles have been shown in dysplastic epithelium adjacent to the oral cancer (26) . Histologically normal mucosa from patients with a primary SCCHN tumor at one aerodigestive tract site, is at high risk for subsequent malignant transformation. The finding of genetic alterations in this mucosa supports its malignant potential. We have previously reported up-regulation of transforming growth factor α and epidermal growth factor receptor in normal mucosa from SCCHN patients as well as in premalignant dysplastic epithelium (44, 45, 46) . Previous studies have demonstrated an interaction between MHC antigens and the EGFR system in squamous cell carcinoma cells in vitro (47) . The high degree of concordance between HLA loss in the tumor and in the corresponding normal mucosa from the same SCCHN patient suggests that HLA allelic loss may be an early event in squamous epithelial carcinogenesis. The potential biological significance of this finding is underscored by the strong correlation between loss in the normal mucosa and subsequent development of a new primary head and neck tumor. Introduction of MHC class I antigens into tumors has resulted in abrogation of tumor growth in preclinical models and is currently under investigation in head and neck cancer patients (48) . The loss of individual HLA class I alleles has additional therapeutic implications for immunomodulation strategies, such as the use of IFN-γ and peptide-pulsed dendritic cells in these patients.
Footnotes
-
The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
-
↵1 To whom reprint requests should be addressed, at Department of Otolaryngology, University of Pittsburgh School of Medicine, 200 Lothrop Street, Suite 500, Pittsburgh, PA 15213.
-
↵2 The abbreviations used are: HLA, human leukocyte antigen; NK, natural killer; TAP, transporter associated with antigen presentation; SCCHN, squamous cell carcinoma of the head and neck; PBMC, peripheral blood mononuclear cell; mAb, monoclonal antibody; ECS, extracapsular spread.
- Received October 21, 1999.
- Revision received March 23, 2000.
- Accepted March 23, 2000.
References
Volume 6, Issue 7
