Carcinoembryonic Antigen as a Target for Specific Antitumor Immunotherapy of Head and Neck Cancer

Cancer of the head and neck represents ∼4% of all cancers, with an estimated 42,800 new cases and 12,300 expected deaths in the United States each year. Approximately half of all patients have advanced stage disease at the time of diagnosis, with an expected 5-year relative survival rate that ranges between 10–40%. Worldwide HNCs³ represent ∼8% of all cancers, with the expected death of 271,000 individuals annually. Long-term survival rates for all HNC victims are among the lowest of major cancers worldwide, with 5-year survivals ranging between 5 and 60%, depending on the site of the primary lesion (1). Most patients with HNC are treated with surgery, radiation, or combination therapy. Whenever possible, larger primary lesions (i.e., >4 cm), with or without invasion of adjacent structures, are surgically excised (2). Despite improvements in diagnostics and locoregional control during the past 3 decades, overall long-term survival rates for HNC patients remain essentially unchanged. For this reason alternative treatment modalities are actively being sought to improve survival. One potential treatment modality for HNCs is vaccine therapy. One of the targets being evaluated for vaccine therapy of other carcinoma types (principally gastrointestinal carcinoma) has been CEA. Recent clinical trials have focused on the use of recombinant poxviral vectors and peptides to generate immune responses against carcinomas of the digestive tract. In these early trials, human CEA, a member of the CEA gene family of oncofetal glycoproteins, has been a principal target for antigen-specific T-cell responses. A variety of poxviral vectors containing the CEA gene alone, or in combination with costimulatory molecules (3, 4) and a CEA-modified peptide (5, 6) are being evaluated currently in clinical trials. It is important to note that results of Phase I studies have shown these vectors and peptide-pulsed DC vaccines to be safe (7, 8, 9), and that patients can indeed mount CEA-specific T-cell responses to CEA (10, 11, 12). These CEA-specific T cells have, in turn, been shown to be capable of lysing colorectal carcinoma cells expressing CEA. Clinical responses and increased survival have also been observed using these CEA-based vaccines (11, 12). To date, none of these clinical trials include patients with carcinomas of the head and neck.

Overexpression of CEA in HNC is not widely recognized and has received little attention. Previous studies have focused on the usefulness of CEA as either a diagnostic or prognostic marker for HNC (13, 14, 15, 16, 17, 18). Despite these efforts, CEA is not considered a useful marker for either a diagnosis or prognosis in HNC. In many of those studies patients were found to have elevated serum levels of CEA; however, the polyclonal antiserum that was used lacked the specificity to discriminate reliably between CEA and CEA-related antigens (i.e., nonspecific cross-reacting antigen, CGM-6, and biliary glycoprotein; Ref. 19). Moreover, previous studies (20) involving colorectal cancer have shown no correlation between the level of CEA protein in the sera and the expression level of CEA in the tumor.

The aim of this study was to evaluate CEA as a possible target antigen for specific immunotherapy against HNC. Cell lines and histological specimens from HNC patients were both qualitatively and quantitatively assessed for CEA expression by use of specific immunological and molecular biological assays. Additional supportive evidence for CEA as a target was demonstrated by the ability of an HLA-A2-restricted/CEA-peptide-specific CTL to specifically lyse CEA-overexpressing HNC-derived SCC cells.

SCC-4 (ATCC CRL-1624), SCC-9 (ATCC CRL-1629), SCC-15 (ATCC CRL-1623), SCC-25 (ATCC CRL-1628), A253 (ATCC HTB-41), CAL-27, TT (ATCC CRL-1803), FaDu (ATCC HTB-43), DET-562 (ATCC CCL-138), SW-579 (ATCC HTB-107), HGF-1 (ATCC CRL-204), Hs840.T (ATCC CRL-7573), SW1463 (ATCC CCL-234), and SK-MEL-24 (ATCC HTB-71) (designated SK-MEL) were purchased from American Type Culture Collection (Manassas, VA). Cell cultures were maintained in complete medium [DMEM (Life Technologies, Inc., Grand Island, NY) supplemented with 10% fetal bovine serum, 2 mm glutamine, 100 units/ml penicillin, and 100 μg/ml streptomycin (Life Technologies, Inc.)]. Adult HEKas were purchased from Cascade Biologics (Portland, OR) and maintained in basal medium for keratinocytes (medium 154; Cascade Biologics) with growth supplement (human keratinocyte goat serum; Cascade Biologics). The C1R cell line is a human plasma leukemia cell line that does not express endogenous HLA-A or B antigens (21) but has been transfected for expression of HLA-A2.1 (22). These cells were obtained from Dr. William E. Biddison (National Institute of Neurological Disorders and Stroke, NIH, Bethesda, MD), and were maintained in RPMI 1640 complete medium (Life Technologies, Inc.). The T-V8 cell line, a CTL line directed against the CAP-1 epitope of CEA, was established from a patient with metastatic colon carcinoma who was enrolled in a Phase I clinical trial using recombinant vaccinia-CEA (23). T-V8 cells were cultured in RPMI 1640 complete medium containing 10% human antibody serum and interleukin 2 (20 units/ml; Hoffmann-LaRoche, Inc., Nutley, NJ). T-V8 cells were grown with CAP-1 peptide (25 μg/ml) at an effector:antigen-presenting cell ratio of 1:3 using a restimulation cycle of 16 days. Irradiated (23,000 rads) autologous EBV-transformed B cells were used as antigen-presenting cells (5). All of the cell lines were tested and found to be free of Mycoplasma contamination by two independent assays using the MycoTect (Life Technologies, Inc., Gaithersburg, MD) and Mycoplasma detection kit (Boehringer-Mannheim, GmbH, Mannheim, Germany).

Total RNA was isolated from 1 × 10⁷ tumor cells with RNA STAT-60 (Tel-Test, Friendswood, TX) according to the manufacturer’s instructions (24). Total RNA was treated with RNase-free DNase I (Boehringer-Mannheim) followed by phenol-chloroform extraction and ethanol precipitation. The RNA concentration was determined from the absorbance values of each sample using a DU 640 UV spectrophotometer (Beckman Instruments, Fullerton, CA).

CEA-specific oligonucleotide-nested PCR primers designated A, B, and C, GAPDH-specific primers and the clatherin-specific oligonucleotide primers were synthesized based on the published sequence information, as described previously (25, 26). CEA-specific primer sequences containing a reduced number of nucleotides from those originally described (27), GAPDH-specific primers (IT BioChem, Salt Lake City, UT), and clatherin-specific oligonucleotide primers (OLI TO GO; Cruachem, Aston, PA) were synthesized using the following nucleotide sequences: A: 5′-TTCTCCTGGTCTCTCAGCT-3′; B: 5′-TGTTGCAAATGCTTTAAGGAAGAA-3′; C: 5′-GGGCCACTGTCGGCATCA-3′; GAPDH 1: 5′-CAGGAGCGAGATCCCT-3′; GAPDH 2: 5′-GGTGCTAAGCAGTTGGT-3′; and Clatherin-F: 5′-GACAGTGCCATCATGAATCC-3′; Clatherin-R: 5′-TTTGTGCTTCTGGAGGAAAGAA3′. The size of oligonucleotide primers was confirmed using polyacrylamide Tris-borate EDTA-Urea (15%) gel electrophoresis (XCell II Mini-Cell system; Novex, San Diego, CA).

Total tumor cell RNA (1 μg) was used to synthesize cDNA using oligodeoxythymidylic acid and Reverse Transcriptase Superscript II (Life Technologies, Inc.). The reliability of first strand synthesis was monitored by the presence of the 570-bp clatherin product after standard PCR analysis with the clatherin-specific primers.

The presence of CEA transcripts was determined by performing nested PCR on tumor cDNA, using the nested primers A, B, and C, as described previously (27). Briefly, 50 ng of each cDNA were diluted to a volume of 20 μl of PCR mix (Roche Diagnostics, Mannheim, Germany) containing a final concentration of 5 mm MgCl₂ and 0.5 μm of primers. For amplification of CEA, the DNA was denatured at 95°C for 10 min after a three-step PCR with 10 s at 95°C, 10 s at 55°C, and 7 s at 72°C for 20 cycles. The outer primers A and B were used. The PCR product was then purified (Microcon PCR; Amicon, Bedford, MA) and diluted to 30 μl. For reamplification, 3 μl of the purified PCR product from the first round of PCR was used in a 20-μl reaction volume as described above. The primers used for reamplification were B and C. The outer CEA primers amplify a PCR fragment of 146 bp, whereas the primers B and C amplify one fragment of 126 bp.

Fifty ng of each cDNA were also used to amplify GAPDH. The PCR mix containing 4 mm MgCl₂ and 0.5 μm of primers also included the GAPDH-specific primers. Denaturation at 95°C for 10 min was followed by a three-step PCR with 10 s at 95°C, 10 s at 55°C, and 10 s at 72°C for 40 cycles. The GAPDH primers amplify a PCR fragment of 245 bp.

Specificity of the PCR products was also confirmed by Bioanalyzer system (Agilent 2100 Bioanalyzer; Agilent Technologies, Waldronn, Germany). As per the manufacturer’s guidelines, 1 μl of each PCR product, diluted in 5 μl of gel-dye mix (DNA 1000 LabChip kit; Agilent Technologies) was analyzed. Results were given as size and concentration (ng/μl) of the PCR product.

The method for single-color flow cytometric analysis has been described (28). Briefly, adherent tumor cells were treated for 5–10 min with 0.25% trypsin with 1 mm EDTA (Life Technologies, Inc.), washed three times with cold Ca²⁺ and Mg²⁺-free DPBS, and viable cells counted using 0.04% filtered trypan blue to exclude nonviable cells. Cells were stained for 1 h with a MAb against either HLA-A2 (A2, 69, 131HA-1; One Lambda, Canoga Park, CA) or CEA (COL-1; Ref. 29) using 10 μl of the 1× working dilution per ∼10⁶ cells. Additional tumor cells were stained for 1 h with appropriate isotype control MAbs that included MOPC-104E (Cappel/Organon Teknika, West Chester, PA) and UPC-10. The cells were again washed three times with cold DPBS and then incubated for 1 h in the presence of 1:100 dilution (100 μl of PBS containing 1% BSA) of FITC-conjugated goat antimouse immunoglobulin (Kirkegaard and Perry Laboratories, Gaithersburg, MD). After incubation, the cells were washed three times, resuspended in cold DPBS at a concentration of 1 × 10⁶ cells/ml, and immediately analyzed using a Becton-Dickinson FACScan equipped with a blue laser with an excitation of 15 mW at 488 nm. Data were gathered from 10,000 live cells, stored, and used for analysis.

Cell extracts were prepared by lysis in nonionic detergent as described previously (30). A near-confluent monolayer of tumor cells was scraped and centrifuged for 5 min at 1800 × g at 4°C. The cell pellet was resuspended in 200 μl of lysis buffer [100 mm Tris-HCL (pH 8.0), 100 mm NaCl, 0.5% NP40, and 0.2 mm phenylmethane sulfonyl fluoride], set on ice for 10 min, and nuclei were separated by centrifugation for 10 min at 4°C. The supernatant was collected for protein analysis. Tumor cell lysates were quantitatively analyzed for total protein by the Bradford Coomassie brilliant blue dye method (31) using Pierce protein assay reagent (Pierce Chemical, Rockford, IL) according to the instructions. Absorbance was measured at 595 nm using a Beckman DU 640 UV spectrophotometer, and the protein concentrations were calculated from a standard protein reference curve.

Tumor cell lysates were quantitatively analyzed for CEA using a two-site fluoroimmunometric assay according to the manufacturer’s instructions (DELFIA CEA assay; Wallac Oy, Turku, Finland). This europium-based direct sandwich technique has been described previously (32). Fluorescence was measured with a time-resolved 1234 DELFIA Fluorometer (Wallac Oy) using MultiCalc software.

Tumor cell lysates were boiled for 10 min in SDS sample buffer as described previously (33), loaded onto a 4–12% gradient SDS gel along with SeeBlue prestained protein standard (Novex, Encinitas, CA), and separated electrophoretically (125V constant for 60–90 min) in Tris-Glycine using the XCell II Mini-Cell system. Proteins were transferred to a nitrocellulose membrane using a semi-dry XCell II blotting system (25V constant for 60–90 min). After transfer, membranes were blocked in 4% Blotto (Instant Nonfat Dried Milk; Carnation, Los Angeles, CA) in PBS for 24 h at 4°C. Blocked membranes were washed in PBS and incubated for 1 h with biotinylated COL-1 (1 μg/ml), then washed three times with PBS and incubated for 1 h with a 1:500 dilution streptavidin-HRP. CEA was detected using single emulsion film (BioMax MS; Kodak, Rochester, NY) and the LumiGLO Chemiluminescence System (Kirkegaard and Perry Laboratories). The radiogram was acquired as a digital image using an HP ScanJet (Hewlett-Packard Corp., Palo Alto, CA), and bands were semiquantitatively analyzed using Kodak1D Image Analysis Software (Eastman Kodak Company, Scientific Imaging Systems, Rochester, NY).

ICC evaluation of CEA expression in human SCC tumor cells was performed as described previously with modification (34). Specifically, 1–2 × 10⁴ cells were plated on eight-well chamber slides (Lab-Tek, Naperville, IL) and incubated for 2–3 days. The CEA-expressing cell line SW-1463 was used as a positive control. Cells were fixed and permeabilized in freshly made cold methanol:acetone at a 1:1 concentration for 5 min at 4°C. The remainder of the procedure was carried out at room temperature. Ten-percent horse serum (Vector Lab, Burlingame, CA) was used to block the nonspecific binding sites for 20 min and removed without washing. The samples were incubated with the following primary antibodies at room temperature for 2 h: 10 μg/ml mouse anti-CEA MAb (Col-1), 10 μg/ml isotype-matched mouse MAb with irrelevant antigen specificity (UPC-10), 0.2 μg/ml mouse antihuman pancytokeratins (Novocastra Lab, Newcastle upon Tyne, United Kingdom) or 0.2 μg/ml of isotype control mouse IgG (Dako Corp, Carpinteria, CA). The samples were then washed and blocked again with 5% horse serum for 20 min and incubated with the secondary biotinylated antibody for 30 min (1:200 in PBS containing 5% horse serum), followed by 30 min of incubation with biotin/avidin HRP conjugates (Vectastain Elite ABC kit; Vector Lab). CEA expression was detected after 5–8 min with a chromogen (diaminobenzidine tetrahydro-chloride; Vector Lab) according to the manufacturer’s instructions. The slides were mounted with glycerol and examined at ×100 and ×400 magnifications.

After obtaining written informed consent in accordance with institutional review board guidelines at the National Cancer Institute and Regina Elena Cancer Institute, tissues were surgically removed from HNC patients or normal volunteers, fixed in 10% buffered formalin, and embedded in low-melting point paraffin. IHC staining of CEA was performed on 4 μm tissue sections and air-dried on poly-l-lysine-coated slides using a modification of the avidin-biotin complex method (35). Briefly, tissue sections were deparaffinized in xylene, rehydrated in graded ethanols, and treated for 20 min at room temperature with methanol containing 0.3% H₂O₂ to inhibit endogenous peroxide activity. After rinsing in PBS (pH 7.4), the sections were incubated in 10% normal horse serum for 15 min, and a biotinylated anti-CEA MAb (COL-1) was added at a concentration of 20 μg/ml (200 μg/slide) and incubated overnight at 4°C. An isotype-matched MAb with irrelevant antigen specificity was used as a negative control. After a PBS rinse, slides were incubated with avidin DH-biotinylated HRP H complex for 30 min at room temperature, washed in PBS, and the peroxidase reaction was initiated using 0.06% diaminobenzidine (Sigma Chemical, St. Louis, MO) and 0.01% H₂O₂ for 2 min. After the final PBS rinse, the sections were counterstained with hematoxylin, dehydrated in ethanol, cleared with xylene, and mounted under a coverslip using Permount. A routine histological examination was also conducted for each specimen on sections stained with H&E at ×100 and ×400 magnifications.

Two pathologists independently evaluated both tumor and control slide samples in a blinded, random manner as described previously (35). For each IHC slide, three to five different fields were evaluated for the presence of tumor cells, as well as the cell-associated diaminobenzidine tetrahydro-chloride precipitate staining pattern within those fields. The relative staining intensity was arbitrarily scored as absent (−), mildly (+), moderately (++), or strongly (+++) reactive.

For ICC slides, the pathologists scored each field independently for the percentage of MAb⁺-staining cells (i.e., number of cells with either moderate or strong reactivity per total number of tumor cells × 100). The percentage of MAb⁺-staining cells represents the analysis of the three to five different fields for each slide. The relative staining intensity represented the MAb⁺-staining intensity within the whole tissue sample. Interobserver agreement was determined by dividing the number of cases in which there was agreement by the total number of cases and multiplying by 100. More than 95% agreement was obtained by the two pathologists. Each pathologist reread all of the slides after 3–5 weeks, and the percentages were averaged for each case. For those slides in which the pathologists did not agree, the slides were unblinded and reread by both pathologists until a consensus was reached. In most cases, the discrepancy was because of the relative staining intensity not the percentage of tumor cells positive for MAb staining. The determination of the antigen index (product of the percentage of positive staining cells and intensity of staining) is as described previously (35, 36).

A 16-h ¹¹¹Indium-release assay was performed as described previously (23). Target cells were labeled with 50 μCi of ¹¹¹Indium-labeled oxyquinoline (Medi-Physics Inc., Arlington, IL) for 15 min at room temperature. Target cells (0.3 × 10⁴) in 100 μl of RPMI 1640 complete medium were added to each of 96 wells in flat-bottomed assay plates (Corning Costar, Inc., Corning, NY). The labeled target cells were incubated with peptides (25 μg/ml) for 60 min at 37°C in 5% CO₂ before adding effector cells. No peptide was used when carcinoma cell lines were used as targets. Effector cells were suspended in 100 μl of RPMI 1640 complete medium supplemented with 10% pooled human antibody serum and added to the target cells. The plates were then incubated at 37°C in 5% CO₂ for 16 h. Supernatant was harvested for gamma counting with the use of harvester frames (Skatron, Inc., Sterling, VA). Determinations were carried out in triplicate, and SDs were calculated. Specific lysis was calculated with the use of the following formula (all of the values in cpm): % Lysis [(Observed release − Spontaneous release)/(Total release − Spontaneous release)] × 100. Spontaneous release was determined from wells to which 100 μl of RPMI 1640 complete medium was added. Total releasable radioactivity was obtained after treatment of targets with 2.5% Triton X-100.

Statistical analysis of differences between means for the CTL assay was done using a two-tailed t test.

Initial experiments investigated whether or not CEA expression could be detected in tumor tissue from 69 cases of SCC of the head and neck using the monoclonal anti-CEA antibody COL-1. This antibody has been shown previously to react only with CEA and not with CEA-related proteins (37). Forty-four of 50 laryngeal tumors (88%) were reactive, and staining intensity ranged from mild to strong (Table 1). An average of 60% of tumor cells from positive tumors of the larynx were reactive, and most of these stained strongly (Table 2). Tumors of the paranasal sinus, oral cavity, and oropharynx were all reactive with COL-1 (Table 1). One of 2 hypopharyngeal tumors was reactive, whereas 9 of 11 esophageal cancers showed positive staining. Overall, a high percentage (87%) of tumors were found to be reactive (Table 1), and expression was primarily cytoplasmic. Six laryngeal and 2 esophageal tumors were determined to be nonkeratinizing SCCs, and none of these were found to be COL-1-reactive. Most importantly, keratinizing squamous epithelium, from normal donor tissue and nonmalignant tissue adjacent to tumor, was found to be nonreactive with COL-1.

The expression pattern of CEA antigen in HNSCCs was examined by immunohistochemistry using COL-1. Fig. 1 shows the pattern of CEA expression in a biopsy of a moderately differentiated SCC of the larynx and lack of CEA expression in surrounding normal mucosa. Fig. 2 shows representative morphology, and the immunostaining pattern of CEA and pancytokeratin in primary tumors removed from 3 other HNSCC patients. The H&E staining (Fig. 2, first row) showed morphology from more differentiated (patient I) to less differentiated (patient III) SCC tumors. As predicted, pancytokeratin staining was positive for all three of the tumors, consistent with epithelial cell origin, although some loss of expression was observed for the least-differentiated SCC as shown by the pancytokeratin staining presented in Fig. 2 (patient III, second row). We observed nonuniform CEA expression patterns at lower magnification (×100) with stronger expression seen in areas with higher degrees of differentiation (Fig. 2, third row). At higher magnification (×400), we observed primarily cytoplasmic staining patterns of CEA expression, which is consistent with the membrane and cytoplasmic expression patterns identified in cultured HNSCC cell lines.

CEA RNA expression was detected using a CEA-specific nested RT-PCR assay in 7 of 10 HNC cell lines comprising head and neck tumors of various cell types (Table 3). The presence of CEA mRNA was detected. The colon carcinoma cell line SW-1463, known to express high levels of CEA, served as a positive control. One cell line derived from a squamous papilloma (Hs840.T), and two lines derived from normal donor tissue (HGF-1 and HEKa) and a melanoma line (SK-MEL) were used as negative controls. Results of nested primer RT-PCR analysis of cell lines is shown in Fig. 3.

CEA-specific solid-phase fluoroimmunometry was used to detect CEA protein in the cytoplasm of each of the 7 HNC cell lines that tested positive for CEA transcript. Protein expression levels for CEA ranged from 93.0 ng/mg of protein in TT cells to 0.4 ng/mg in FaDu cells. The CEA protein expression level in the head and neck tumor cell lines SCC-4 and TT was equivalent to the CEA levels in the SW-1463 colon adenocarcinoma cell line known for its high level of CEA expression. CAL-27, FaDu, and SSC-15 with the lowest level of cytoplasmic CEA expression at 2, 0.4, and 1.8 ng/mg of protein, respectively, had undetectable levels of CEA shed into the tissue culture supernatant (Table 4). SDS-PAGE/Western blot analysis of cellular lysates of all of the head and neck tumor lines demonstrated a COL-1 immunoreactive product with a molecular weight of 180 kilodaltons. This value is equal to the M_r of the CEA glycoprotein molecule (Fig. 4,B). Also shown are the Western blot analyses of different concentrations of purified CEA, and positive control colon cancer cell line SW-1463 and negative control cell lines HEKa and SK-MEL (Fig. 4 A).

The percentage of CEA-positive cells when assayed by immunocytochemistry was higher for all 5 of the SCC lines than when assayed by FACS analysis (Table 5). Inspection at high magnification showed primarily intracellular staining of the cells. Of note, SCC-4 had approximately half the CEA expression of SW-1463 by ICC analysis. All of the cells that scored positively were either moderately or strongly reactive. All of the lines were negative (i.e., <1% reactive) for isotype control antibody reactivity. Squamous keratinocytes derived from normal donor tissue were negative for COL-1 reactivity. FACS analysis of the cell surface expression of CEA was low in SCC-4, CAL-27, A253, FaDu, and SCC-15 (Fig. 5). A comparatively high level of CEA expression was detected on the surface of three adenocarcinoma lines: DET-562, SW-1463, and TT (Table 5; Fig. 5). Interestingly, the SCC line SCC-4 displayed high levels of CEA in the cytoplasm much like the adenocarcinoma cell lines SW-1463 and TT; unlike TT and SW-1463, however, SCC-4 displayed low surface expression of CEA. This observation was not explainable on the basis of trypsin sensitivity of SCC-4 because FACS analysis after overnight incubation of trypsinized cells maintained in suspension yielded no change in surface expression of CEA (data not shown).

The HLA-A2-restricted, CEA epitope-specific human CTL designated T-V8 was tested for cytotoxic activity against HLA-A2⁺ and CEA⁺ HNC cell lines (i.e., SCC-4). As shown in Table 6, T-V8 killed 23.7% of SCC-4 cells at an E:T of 50 and 52.6% of SW-1463 (positive control) cells; no killing was observed against TT cells. The difference in CTL killing percentage was consistent with the difference in cytosolic CEA expression between the two lines. Specifically, the percentage of positive cytosolic CEA expression and CTL killing of SCC-4 was approximately half the percentage found with SW-1463. As expected, T-V8 failed to kill the SK-MEL and C1R-A2 lines (HLA-A2⁺ and CEA⁻); however, CTL activity was observed against C1R-A2 cells loaded with CAP1-6D (57.6% lysis), the CEA-specific CTL-enhancer agonist peptide. As shown in Table 7, the percentage of lysis of the SCC-4 head and neck cell line by the CEA-specific T-cell line was also shown to increase after the addition of IFN-γ to the tumor cells at all three of the E:T ratios tested.

The presence of tumor-associated antigens in the cytosol of head and neck carcinomas may make it possible for T cells to recognize and lyse such tumors (38). This pattern of expression is similar to that seen in other carcinomas such as colorectal carcinoma. The staining intensity seen in many of the head and neck tumors was also comparable with that seen in colorectal carcinomas. It should be pointed out that T cells do not recognize surface proteins but processed peptides in combination with surface MHC molecules. Thus, this cytosolic pattern of expression may well be suitable for T-cell recognition. Typically, HNC occurs in patients with a history of heavy carcinogen exposure from the use of tobacco and alcohol. Evidence suggests that prolonged exposure to such mutagens is associated with the expression of neoantigens (39). Indeed, a number of overexpressed antigens have been identified in SCC (40, 41, 42). These include two members of the immunoglobulin supergene family, namely, CEA and human E48 SCC antigen (43, 44). It was shown that CEA could function as a target for immunotherapy in a murine adenocarcinoma cell line transduced with a retroviral construct containing cDNA encoding the human CEA gene (29). The murine homologue of E48, known as Ly-6A/E, was found to be overexpressed after malignant transformation of the murine SCC line Pam 212. Subsequent loss of expression of Ly-6A/E occurred with metastatic tumor progression of Pam in vivo (45). It is possible that, when overexpressed, these immunoglobulin homologues may function as immune-recognition determinants, through efficient processing of MHC class I and II peptides.

Whereas evidence exists that highly malignant tumor cells may retain antigens recognized by T-helper cells and lose antigens recognized by cytotoxic T-cell-mediated CTLs (46), some of these antigens have been shown to elicit CTL responses. For instance, CTLs specific for wild-type peptide epitopes in p53 have been shown to kill human head and neck SCC lines overexpressing mutant p53 (47). In the present study, specific in vitro assays were used to detect and measure the expression of CEA in HNCs. Indeed, the majority of the HNC tumors that were evaluated expressed CEA. More importantly, this study also demonstrated, for the first time, the presence of CTL-mediated killing of human HNC cells overexpressing CEA.

The relative tissue specificity of CEA makes it an attractive target for active specific immunotherapy (29, 48). In preclinical studies, effective antitumor responses were demonstrated in CEA transgenic mice after immunization with recombinant poxvirus-expressing CEA (49). The antitumor response was enhanced by the presence of granulocyte macrophage colony-stimulating factor, and there was no associated autoimmune toxicity. Several recent clinical trials in patients with advanced carcinoma have shown that vaccine consisting of recombinant poxvirus-expressing CEA or CEA peptide-pulsed DCs can generate CEA-specific T-cell responses in patients with advanced gastrointestinal tumors (10, 50, 51, 52, 53, 54, 55). Survival in one of these studies was correlated with increases in CEA-specific T-cell responses (11). In another study, a CEA-modified peptide, designated CAP1-6D (5, 6), was used to pulse DCs as a vaccine. After vaccination, 2 of 12 patients experienced dramatic tumor regression, 1 patient had a mixed response, and 2 had stable disease. Clinical response correlated with the expansion of CEA-specific T cells (12). Therefore, under defined conditions, CEA is capable of generating an immunological response in humans without evoking autoimmune toxicity despite being a self-antigen.

A large body of scientific evidence now implicates the CD4 and CD8 cells in playing a key role in effecting tumor regression. It is now widely held that the objective of antitumor vaccines is to generate a robust cell-mediated or T_H1-type immune response to antigen. There is also evidence that inadequate expression of the costimulatory molecule B7.1 by head and neck and other cancer cell types may lead to tumor escape from immune defense (56, 57, 58, 59, 60). Moreover, recent vaccine constructs have been developed that contain the CEA gene and three different T-cell costimulatory molecules (B7.1, ICAM-1, and LFA-3; designated TRICOM; Refs. 3, 4). The avipox-CEA-TRICOM and or vaccinia-CEA-TRICOM vectors (3, 4) would be excellent candidates for vaccine trials of patients with CEA-positive head and neck carcinomas. In the study reported here, the high percentage of HNCs that expressed CEA provide supportive evidence to consider using CEA-based immunogens in clinical trials of HNC patients.

We thank Drs. Tomoko Makishima, Douglas Grosenbach, and Philip Arlen for critical reviews of the manuscript. We also thank Dr. Stephen Hewitt for assistance with scoring IHC slides. Garland Davis and Judith DiPietro are acknowledged for their outstanding technical assistance, and Debra Weingarten for help in manuscript preparation.