2353 The Levels of Expression of Galectin-1, Galectin-3, and the Thomsen–Friedenreich Antigen and Their Binding Sites Decrease as Clinical Aggressiveness Increases in Head and Neck Cancers Georges Choufani, M.D.1 Nathalie Nagy, M.D.2 Sven Saussez, M.D.1 Hadelin Marchant, M.D.3 Pierre Bisschop, M.D.3 Maria Burchert, Ph.D.4 André Danguy, Ph.D.5 Stéphane Louryan, M.D.6 Isabelle Salmon, M.D.2 Hans-Joachim Gabius, Ph.D.4 Robert Kiss, Ph.D.5 Sergio Hassid, M.D.1 1 Department of Otolaryngology and Head and Neck Surgery, Cliniques Universitaires de Bruxelles, Hôpital Erasme; Brussels, Belgium. 2 Department of Pathology, Cliniques Universitaires de Bruxelles, Hôpital Erasme, Brussels, Belgium. 3 Department of Otolaryngology and Head and Neck Surgery, Hôpital Brugmann, Brussels, Belgium. 4 Institute of Physiological Chemistry, Faculty of Veterinary Medicine, Ludwig-Maximilians University; Munich, Germany. 5 Laboratory of Histopathology, Université Libre de Bruxelles, Brussels, Belgium. 6 Laboratory of Anatomy and Embryology, Faculty of Medicine, Université Libre de Bruxelles, Brussels, Belgium. BACKGROUND. The aim of this study was to investigate whether an increase in malignancy level is accompanied by significant modifications of the expression of galectin-1, galectin-3, and Thomsen–Friedenreich antigen (T antigen) as well as the expression of binding sites for these three markers in head and neck squamous cell carcinomas (HNSCCs). METHODS. Immunohistochemical and glycohistochemical staining reactions were carried out with antibodies, labeled lectins, and a custom-made neoglycoprotein on the basis of histologic slides from a retrospective series of 40 normal and 75 HNSCC formalin fixed, paraffin embedded tissues, and were quantitatively described with the aid of computer-assisted microscopy. RESULTS. Whatever the histologic type, the epithelial tissues in HNSCC exhibited very significantly (P , 0.01 to P , 0.0001) lower amounts of galectin-1, galectin-3, and T antigen and their respective binding sites than their corresponding normal counterparts. The tumors of the larynx differed very significantly (P , 0.0001 to P , 0.000001) from all the other tumor types. A loss of differentiation in the HNSCCs is accompanied first by the loss of expression of galectin-3 and galectin-3-reactive sites and then by that of the T antigen and its binding site(s). The opposite feature was observed when the parameters associated with the TNM classification were taken into account. The negative lymph node HNSCCs could be distinguished (P 5 0.02) from the positive lymph node HNSCCs on the basis of a loss of galectin-3 expression. The modifications occurring in the extent of expression of galectin-1 and galectin-1-reactive sites were relatively marginal in comparison with those observed for galectin-3-dependent and T- antigen-dependent staining. CONCLUSIONS. The decrease in the extent of expression of galectin-3 and galectin3-reactive sites, T antigen and T antigen-binding sites, and, to a lesser extent, galectin-1 and galectin-1-reactive sites correlates significantly with an increasing level of clinically detectable HNSCC aggressiveness. Cancer 1999;86:2353– 63. © 1999 American Cancer Society. Supported by grants from the Dr. M. Scheel-Stiftung für Krebsforschung (Germany) and the Fonds de la Recherche Scientifique Médicale (FRSM, Belgium). KEYWORDS: galectins, lectins, head and neck cancers, prognosis, quantitative measurements, computer-assisted microscopy. The authors are especially indebted to Dr. J.L. Wang for his kindness in providing the expression vector for murine galectin-3. C R.K. is a Senior Research Associate with the Fonds National de la Recherche Scientifique (FNRS, Belgium). Address for reprints: Robert Kiss, Ph.D., Laboratory of Histopathology, Faculty of Medicine, Free University of Brussels, 808 Route de Lennik, 1070 Brussels, Belgium. Received June 24, 1999; accepted July 19, 1999. © 1999 American Cancer Society ancers of the upper aerodigestive tract are heterogeneous in their neoplastic processes, each of which requires its own unique set of epidemiologic, anatomic, pathologic, and therapeutic considerations, as stated by Schantz et al.1 The predominant lesions in the case of head and neck cancers are squamous cell carcinomas, which constitute a rather heterogeneous class and can be divided into welldifferentiated, moderately differentiated, and poorly differentiated lesions (see Materials and Methods). In accurately determining prognoses in head and neck squamous cell carcinomas (HNSCCs), the monitoring of defined biochemical characteristics can lead to new insights. Prognoses usually are esti- 2354 CANCER December 1, 1999 / Volume 86 / Number 11 mated by means of staging, but the role and actual relevance of staging in the case of HNSCCs is undergoing continuous analysis.1 The common staging reported by the American Joint Committee on Cancer Staging (AJCC)2 is based on the tumor-lymph nodemetastases (TNM) classification, with T representing the extent of the primary disease, N representing the extent of the regional lymph node metastases, and M representing the measure of distant metastases. Similar to other tumor entities, the quest to define markers with a demonstrable impact on clinical courses is a potentially promising and therapeutically relevant research topic. Several well-defined classes of biomolecules have been monitored due to the assumed importance of cell-cell and cell-extracellular matrix adhesion processes, cytoskeletal rearrangements for tumor cell invasiveness, and nuclear proteins for proliferative activity. For example, modifications in the expression levels of keratins,3–5 cadherins,6,7 cortactin,8 syndecan,9 nuclear matrix proteins,10 and the organization of the basement membrane11 have been reported as being correlated to various extents with prognoses in patients with HNSCC. In view of the potential of oligosaccharides to serve as information-bearing molecules, attention should be paid not only to protein-protein interactions. Obviously, the recognitive interplay between these coding units and receptors, such as endogenous lectins, likewise can regulate intercell interactions and various cell activities.12 The monitoring of one class of epitopes commonly has been performed by exogenous lectins. In this line of research, the detection of a diagnostic and/or prognostic value for HNSCCs includes the peanut agglutinin (PNA), which is purified from Arachis hypogaea and binds to the glycoproteins that expose the galactose-b 1,3-N-acetylgalactosamine sequence referred to as the Thomsen–Friedenreich antigen (T antigen).13–15 Having explicitly implied the functional significance of such epitopes, it is a logical step to complement the application of an agglutinin with the assessment of binding sites for the localized sugar structure in the tissue. Chemical synthesis of the ligand part and its immobilization on a suitable carrier gains access to the required tool, i.e., a neoglycoconjugate.16,17 Such a probe already has been proven useful for localizing endogenous binding sites for the T antigen.18 Thus, the concomitant analysis with an agglutinin and an appropriate neoglycoprotein is expected to advance our knowledge of distinct systems of protein-carbohydrate interactions. Due to the remarkable progress in animal lectinology, it is becoming possible to determine the expression of distinct endogenous lectins in tissues. If the same concept is applied, then the concomitant monitoring of lectin expression by an antibody and of lectin-reactive sites by the labeled tissue lectin constitutes a promising method for inferring physiologic activities. Due to their involvement in various activities, including splicing, apoptosis, and cell-matrix interactions, members of the galectins group may influence aspects of tumor cell behavior.20 This reasoning has prompted the monitoring of galectins in malignant human tumor specimens, such as sections from breast, thyroid, and brain carcinoma, and also the initial demonstration of the presence of galectin in HNSCC cell lines and 35 primary HNSCCs.21–24 However, no study so far has applied biochemically purified galectins simultaneously to determine the extent of lectin-reactive sites and galectin specific antibodies. The objective of the current work was to answer the pertinent question of whether increasing clinical aggressiveness in HNSCCs, as evidenced by TNM staging, is correlated with modifications in the expression of galectin-1, galectin-3, and galectin-reactive sites and of T antigen/PNA-binding sites. The series of specimens comprised 75 HNSCCs. Immunohistochemical and glycohistochemical staining reactions were quantified by means of computer-assisted microscopy. MATERIALS AND METHODS Histopathologic and Clinical Data The current series included 40 normal cases and 75 cancer cases. The 40 normal cases included 28 specimens from the oral cavity, 7 from the larynx, and 5 from the hypopharynx. All of these cases were fresh material obtained either from autopsy or surgery performed for nontumor pathologies. There were no differences in immunohistochemistry patterns between samples obtained directly from surgery and those obtained from cadavers (data not shown). The 28 oral cavity cases included 6 samples from the tongue, 8 from the tonsils, 7 from between the tongue and the tonsils, and 7 from the buccal mucosa. The patients were age and gender matched with respect to the cancer sample series. The 75 HNSCC tumor specimens included 41 from the oral cavity, 23 from the larynx, and 11 from the hypopharynx. The 41 oral cavity tumor specimens included 10 from the tongue, 10 from the tonsils, 9 from between the tongue and the tonsils, and 12 from the buccal mucosa. All of the patients (62 men and 13 women, ages 43– 83 years) had undergone partial or radical surgery. The diagnoses were established on the basis of the histologic criteria described by Hyams et al.25 All of the 75 HNSCC tumor specimens used for analysis were from primary tumors and not from metastases or recurrences. The 75 specimens from patients with HNSCC were categorized as well differentiated (n 5 35 speci- Galectin Expression in Head and Neck Cancers/Choufani et al. mens), moderately differentiated (n 5 30 specimens), and poorly differentiated (n 5 10 specimens) on the basis of the definitions given below. The well-differentiated lesions were characterized by the presence of intercell bridges and numerous keratinization foci, whereas the poorly differentiated lesions were characterized by the rare presence (or a notable absence) of keratinization foci and the absence of intercell bridges. Individual cell keratinization and the presence of intercell bridges are consistent with the moderately differentiated phenotype. Clinical staging was carried out by using the TNM classification system.2 There were 5 patients T1 tumors, 23 patients with T2 tumors, 19 patients with T3 tumors, and 28 patients with T4 tumors. Of these 75 HNSCC patients, 52 were lymph node negative, and 23 were lymph node positive. All 75 patients were M0. Of the 75 HNSCC patients under study, 31 had undergone surgical treatment alone, and 38 had undergone surgical treatment followed by radiotherapy. We checked to determine whether these two therapeutic regimens led to significantly distinct prognoses. According to what was observed during the relatively short period of clinical follow-up, no such difference existed (data not shown). We collected complete clinical follow-up data for 34 of the 75 HNSCC patients. Of these 34 patients, 14 died ,12 months after their initial diagnosis (the D , 12M group), whereas the other 20 patients were still alive at the time of the analysis and had survival periods exceeding 36 months from the initial diagnosis of their tumors (the SURV . 36M group). The remaining 41 patients were not included in the survival status-related analysis, because their clinical follow-up periods ranged only between 1 month and 21 months. Immunohistochemistry and Ligandohistochemistry The expression of galectin-1 and galectin-3 in the 40 normal cases and the 75 HNSCC patients under study was demonstrated by means of specific polyclonal antigalectin-1 and antigalectin-3 antibodies, respectively, and the expression of galectin-1-reactive and galectin-3-reactive sites was assessed by biotinylated galectin-1 and galectin-3 using routine procedures.26 The expression of the T antigen was detected by means of the application of biotinylated PNA, whereas that of the T antigen-binding sites was revealed by means of a biotinylated carrier-immobilized T antigen (i.e., by the use of a synthetic T antigen derived from a chemical preparation as a p-aminophenyl derivative18) as a ligand part of a biotinylated neoglycoprotein. Methodologically, tissue specimens were fixed in 4% formaldehyde and embedded in paraffin, and 5-mm-thick sections were subjected to processing with the various histochemical probes and kit reagents un- 2355 der study. The markers were prepared and biotinylated under activity-preserving conditions, as detailed elsewhere.19,26,27 The galectins were purified, as described previously in detail.28,29 Incubation with labeled probes was carried out at room temperature for 60 minutes at a concentration of 10 mg/mL. The extent of specifically bound markers (the biotinylated probes or antibodies) was demonstrated by avidin-biotinperoxidase complex (ABC) kit reagents (Vector Laboratories, Burlingame, CA) with diaminobenzidineH2O2 as a chromogenic substrate, as detailed elsewhere.17,19,30 –32 The control reactions included competitive inhibitions to ascertain sugar specificity, and the omission of the incubation step with a labeled marker served to exclude any staining by the binding of kit reagents, such as the mannose-rich glycoproteins horseradish peroxidase and avidin. Counterstaining with Toluidine blue concluded the processing. Computer-Assisted Microscopy The variables from the quantitative histochemical stainings were computed by means of a SAMBA 2005 computer-assisted microscope system (Alcatel-TITN, Grenoble, France) with a 320 magnification lens (aperture, 0.50). The way in which this system was used to quantify the histochemical staining is detailed elsewhere.30 –32 The computer-assisted microscope and related quantitative analyses were standardized as detailed elsewhere.33 Briefly, a negative histologic control slide (on which the primary antibody of the biotinylated glycoprobe was omitted) was analyzed for each probe assayed for each of the 40 normal samples and the 75 HNSCC samples under study. The software included in the computer-assisted microscope automatically subtracted the labeling index (LI) and mean optical density (MOD) values (see below) of the negative control sample from each corresponding positive sample. Some CCD camera-based systems, like that used in this study, exhibit inherent shading, i.e., equal levels of incident illumination that do not produce equal signal values from all of the image areas to be analyzed. Shading, therefore, is one factor that limits the performance of a computer-assisted microscope. Specific software in our system was set up to evaluate the level of magnitude of the shading phenomenon. This software was used to measure optical densities on a uniform sample (i.e., a test slide with different neutral density Kodak Wratten filters; Eastman-Kodak, Rochester, NY) in five different parts of the field of vision. The test results showed that the computerassisted microscope was not affected significantly by shading features (data not shown). Glare phenomenon also can affect computer-assisted microscoperelated measurements, because it concerns all types of 2356 CANCER December 1, 1999 / Volume 86 / Number 11 TABLE 1 Description of the 18 Quantitative Histochemical Variables Under Study Quantitative parametersa Tissue target (lectin or lectin-binding sites) Probe used Designation Abbreviation Galectin-1 Antigalectin-1 antibody Galectin-1-binding sites Biotinylated galectin-1 Galectin-3 Antigalectin-3 antibody Galectin-3-binding sites Biotinylated galectin-3 T antigen Biotinylated PNA T-antigen binding sites (endogenous PNA-like lectin) Biotinylated T antigen-bearing neoglycoprotein Labeling index Mean optical density Concentrational heterogeneity Labeling index Mean optical density Concentrational heterogeneity Labeling index Mean optical density Concentrational heterogeneity Labeling index Mean optical density Concentrational heterogeneity Labeling index Mean optical density Concentrational heterogeneity Labeling index Mean optical density Concentrational heterogeneity aGal1-LI aGal1-MOD aGal1-CH Gal1-LI Gal1-MOD Gal1-CH aGal3-LI aGal3-MOD aGal3-CH Gal3-LI Gal3-MOD Gal3-CH PNA-LI PNA-MOD PNA-CH T-LI T-MOD T-CH PNA: peanut agglutinin. a The labeling index (LI) relates to the percentage of tissue area specifically stained by a glycohistochemical probe. The mean optical density (MOD) relates to the staining intensity. The concentrational heterogeneity (CH) represents the extent of variability calculated for the MOD. phenomena due to extra light gathered in the image. The quantity of glare is system dependent and is a function of the relative size of the object in the illuminated area. Glare phenomenon is greater for highly absorbent particles and may introduce errors into some quantitative measurements. In the system used in this study for the shading evaluation, as explained below, specific software also was set up to evaluate the level of magnitude of the glare phenomenon. This phenomenon did not modify our results significantly (data not shown). In cases in which an object to be analyzed fills the whole field (like in the current study), the glare phenomenon remains of very low magnitude. Fifteen fields of between 60,000 mm2 and 120,000 mm2 epithelial cells were scanned per histologic slide. The LI refers to the percentage of epithelial tissue area specifically stained by a glycohistochemical probe, the MOD denotes staining intensity, and the concentrational heterogeneity (CH) feature expresses the concentration spread of individual epithelial areas and is evaluated by the coefficient of variation (CV 5 standard deviation value/mean value) of the MOD values. Thus, 18 quantitative histochemical variables were available for each of the 40 normal samples and each of the 75 HNSCC samples under study. These are listed and explained in Table 1. Data Analysis Discriminant analysis, a multivariate statistical procedure that is described elsewhere,34 was used to evaluate the contribution of the measured quantities (i.e., LI, MOD, and CH) for each of the six probes to the characterization of the various histopathologic groups considered in the series of 75 HNSCC samples. In the current study, the stepwise discriminant analysis program Statistica (Statsoft, Tulsa, OK) was used. RESULTS Expression of Galectin-1, Galectin-3, and T Antigen and their Respective Binding Sites in Normal Samples versus HNSCC Samples The percentages (determined by the LI variable) of epithelial cells that exhibited significant amounts of acceptor sites for the biotinylated T antigen were very significantly lower (P , 0.01 to P , 0.0001) in the HNSCC samples compared with the normal samples regardless of the histologic site considered (Fig. 1A). In contrast, the amounts (determined by the MOD variable) of T antigen-binding sites expressed by the glycohistochemically T antigen positive epithelial cells were relatively similar (P , 0.05 to P . 0.05) in the normal samples and the HNSCC samples (Fig. 1B). The application of biotinylated PNA showed that the percentages of cells exhibiting significant amounts of T antigen also were significantly lower in the Galectin Expression in Head and Neck Cancers/Choufani et al. 2357 Mean (bars) 6 standard error of the mean values of the labeling indices (A,C) and the mean optical densities (B,D) determined by using computerassisted microscopy for biotinylated T antigen (revealing glycohistochemically the presence of T-antigen binding sites) and biotinylated peanut agglutinin (PNA) (revealing glycohistochemically the presence of T antigen) in normal cells and in epithelial cells from patients with head and neck squamous cell carcinoma. Six histologic sites were analyzed: the buccal mucosa (B-M group), the tongue (TONG group), the area between the tongue and the tonsils (T-T group), the tonsils (TONS group), the larynx (LAR group), and the hypopharynx (HYPH group). Each histologic site included a minimum of 5 cases and a maximum of 12 cases, either normal (open bars) or malignant (solid bars), as detailed in Materials and Methods. The number of asterisks indicates the level of statistical significance: 1 asterisk, P , 0.05; 2 asterisks, P , 0.01; three asterisks, P , 0.001; four asterisks, P , 0.0001. a.u.: arbitrary unit. FIGURE 1. HNSCC samples compared with the normal samples (Fig. 1C), but the differences were less pronounced than what was observed for the biotinylated T antigenrelated glycohistochemical staining (Fig. 1A). In the application of biotinylated T antigen (Fig. 1B), the amounts of T antigen expressed by the glycohistochemically PNA positive epithelial cells were similar (P . 0.05) in the normal samples and the HNSCC samples (Fig. 1D). The percentages of epithelial cells that were positive with respect to the galectin-1 binding sites (revealed by biotinylated galectin-1; Fig. 2A) and galectin-1 (revealed by antigalectin-1 antibody; Fig. 2C) were significantly lower in the HNSCC samples compared with the normal samples. The amounts of galectin-1 binding sites in the biotinylated galectin-1 positive epithelial cells were similar (P . 0.05) in the normal samples and the HNSCC samples except for the buccal mucosa, in which the amounts of galectin-1 binding sites were significantly lower (P , 0.01) in the epithelial cells from the HNSCC samples compared with the normal samples (Fig. 2B). The same feature was observed with respect to the amount of galectin-1 (Fig. 2D). In addition, the epithelial HNSCC cells from the hypopharynx also exhibited significantly lower (P , 0.05) amounts of galectin-1 compared with normal epithelial cells from the hypopharynx (Fig. 2D). The percentages of epithelial cells that were positive with respect to the galectin-3 binding sites (revealed by biotinylated galectin-3; Fig. 3A) and galectin-3 (revealed by antigalectin-3 antibody; Fig. 3C) were significantly lower in the HNSCC samples compared with the normal samples except for the larynx, in which the percentages of immunohistochemically positive galectin-3 epithelial cells were similar (P . 0.05) in the normal samples and the HNSCC samples (Fig. 3C). The amounts of galectin-3 binding sites (Fig. 3B) and galectin-3 (Fig. 3D) were similar (P . 0.05) in both HNSCC and normal epithelial cells. Expression of Galectin-1, Galectin-3, and T Antigen and their Respective Binding Sites in Relation to HNSCC Histologic Types Figure 4A–D illustrates the development of 4 of the 18 quantitative histochemical variables that were available for each of the 75 HNSCC samples under study. It is very difficult to formulate precise conclusions from such huge amounts of data. Therefore, a discriminant analysis was performed to select the more informative variables. Together with the P value of statistical significance, Table 2 reports the 3 most discriminatory quantitative histochemical variables (ranked in decreasing order of discriminatory power) selected by discriminant analysis for 2358 CANCER December 1, 1999 / Volume 86 / Number 11 Mean (bars) 6 standard error of the mean values of the labeling indices (A,C) and the mean optical densities (B,D) determined by using computerassisted microscopy for biotinylated galectin-1 (revealing glycohistochemically the presence of galectin-1 binding-sites) and the antigalectin-1 antibody (revealing immunohistochemically the presence of galectin-1) in normal cells and in epithelial cells from patients with head and neck squamous cell carcinoma. For abbreviations and conventions, see the legend to Figure 1. FIGURE 2. Mean (bars) 6 standard error of the mean values of the labeling indices (A,C) and the mean optical densities (B,D) determined by using computerassisted microscopy for biotinylated galectin-3 (revealing glycohistochemically the presence of galectin-3 binding sites) and for the antigalectin-3 antibody (revealing immunohistochemically the presence of galectin-3) in normal cells and in epithelial cells from patients with head and neck squamous cell carcinoma. For abbreviations and conventions, see the legend to Figure 1. FIGURE 3. each of the paired comparisons. Each histologic group harbored specific (P 5 0.023 to P , 0.000001) glycohistochemical features except for the tumors from the tongue-tonsil group when those tumors were compared with those from either the tongue group or the tonsil group (Table 2). Expression of Glycohistochemical Markers in Relation to the Embryologic Origins of HNSCC Figure 4 shows that the origins of the tumors in this series varied widely. Tumors from different sites have different clinical courses that may be due to inherent characteristics of the origin of the tissue. Some of the tumors under study had an ectodermal origin, whereas others had an endodermal origin. Thus, the data were analyzed in terms of the embryologic origins of the HNSCC tumors. The endodermal group includes tumors from the larynx, the base of the tonsils, the junction between the tongue and the tonsils, the pharynx, and the hypopharynx. The ectodermal group includes tumors from the mouth and those corresponding to the anterior two-thirds of the tongue. Despite the mixed embryologic origins of the tonsil tumors, these tumors are considered to be Galectin Expression in Head and Neck Cancers/Choufani et al. 2359 FIGURE 4. Illustration of the development of the mean (open squares) and the related standard error (black rectangles) and deviation (bars) values in patients with head and neck carcinoma of the buccal mucosa (BUC-MUC group; n 5 12), the tongue (n 5 10), the tonsils (n 5 10), the transitional area between the tongue and the tonsils (TONG-TONS group, n 5 9), the larynx (n 5 23), and the hypopharynx (HYPOPHAR group, n 5 11). (A) The labeling index quantified with respect to the T antigen. (B) The mean optical density for staining by the antigalectin-1 antibody. (C,D) The labeling index for staining by biotinylated galectin-3 and the antigalectin-3 antibody, respectively. endodermal because of their epithelial origins (squamous cell carcinoma). Figure 5A, which should be compared with Figure 4A, shows that the HNSCC samples of ectodermal origin (n 5 12 samples) exhibited a significantly higher (P 5 0.008) mean LI value in relation to the T antigenbearing biotinylated glycoprotein compared with the HNSCC samples of endodermal origin (n 5 63 samples). In contrast, Figure 5B, which should be compared with Figure 4D, shows that the reverse feature was observed with respect to the anti-galectin-3 antibody (P 5 0.001). Expression of Galectin-1, Galectin-3, and T Antigen and their Respective Binding Sites in Relation to HNSCC Differentiation Levels Table 3 shows that the well-differentiated tumors (independent of histologic type; data not shown) present significantly different glycohistochemical characteristics compared with the moderately differentiated (P 5 0.009) or the poorly differentiated (P 5 0.0004) tumors. The moderately differentiated tumors differed weakly, although significantly (P 5 0.02), from the poorly differentiated tumors (Table 3). Expression of Galectin-1, Galectin-3, and T Antigen and their Respective Binding Sites in Relation to the Clinical Stage of HNSCC Table 4 shows that tumors belonging to each of the four clinical T stages exhibited significantly different (P 5 0.02 to P 5 0.002) glycohistochemical characteristics except for stage T3, which did not differ significantly (P . 0.05) from stage T4. Lymph node negative (N2) HNSCC tumors were weakly, although significantly, different (P 5 0.02) from lymph node positive (N1) tumors. The three most discriminatory variables in decreasing order of power were antigalectin-3 MOD, galactin-3 LI, and PNA MOD. All 75 HNSCC samples under study included an M0 classification in terms of TNM staging. 2360 CANCER December 1, 1999 / Volume 86 / Number 11 TABLE 2 Results Obtained by Means of Discriminant Analysis with the Emphasis on Identifying the Most Informative Quantitative Variables Describing the Expression of Galectin-1, Galectin-3, T Antigen, and their Respective Binding Sites in Relation to Head and Neck Squamous Cell Carcinoma Histologic Typea Buc-Muc Tongue PNA-LI (1) aGal1-LI (2) T-LI (2) P 5 0.006 Tong-Tons PNA-MOD (2) Gal3-MOD (1) aGal3-MOD (2) P 5 0.035 Tonsil PNA-CH (2) PNA-LI (1) aGal1-MOD (2) P 5 0.005 Larynx aGal3-LI (2) T-LI (1) aGal3-MOD (2) P , 0.000001 Hypophar aGal3-LI (2) T-LI (1) aGal1-LI (2) P 5 0.0006 Tongue Tong-Tons Tonsil Larynx — — — — — — — — — — — — — — — — — ns — — — — — — — — — — — — — — PNA-CH (2) T-MOD (2) T-CH (2) P 5 0.023 — ns — — — — — — — — — — T-LI (1) aGal1-LI (2) aGal3-MOD (2) P , 0.000001 aGal3-LI (2) T-LI (1) aGal1-LI (2) P , 0.000001 aGal3-LI (2) T-LI (1) Gal3-CH (1) P , 0.00001 — — — — aGal3-LI (2) aGal1-LI (2) T-LI (1) P 5 0.00003 aGal3-LI (2) aGal1-LI (2) Gal1-MOD (1) P 5 0.002 aGal3-LI (2) Gal1-LI (2) Gal1-MOD (1) P 5 0.03 T-LI (2) T-CH (2) PNA-CH (1) P 5 0.004 a The head and neck squamous cell carcinoma histologic types include 12 tumors of the buccal mucosa (the Buc-Muc group), 10 of the tongue (the Tongue group), 10 of the tonsils (the Tonsil group), 9 between the tongue and the tonsils (the Tong-Tons group), 23 of the larynx (the Larynx group), and 11 of the hypopharynx (the Hypophar group). The quantitative variables are described in Table 1. The three most discriminatory quantitative histochemical variables in decreasing order of discriminatory power are listed for each paired comparison along with the P level of statistical significance (ns, not significant; i.e., P . 0.05). The (2) and (1) signs to the right of each quantitative parameter indicate whether the mean parameter value in a given column group was lower (2) or higher (1) than the mean value of this parameter in the corresponding line group. For example, the (1) value to the right of the PNA-LI parameter under the Buc-Muc group column and in the Tongue group line indicates that the mean PNA-LI value was higher in the Buc-Muc than in the Tongue group. Expression of Galectin-1, Galectin-3, and T Antigen and their Respective Binding Sites in Relation to Patient Survival The 14 patients with HNSCC in the D , 12M group displayed significantly different (P 5 0.009) characteristics compared with the 20 patients in the SURV . 36M group. The three most discriminatory variables in decreasing order were PNA LI, galectin-3 MOD, and T LI. DISCUSSION The current study characterizes the patterns of expression of 6 glycohistochemical markers in 40 normal samples and in 75 patients with HNSCC. These 6 FIGURE 5. The development of the mean (open squares) and the related standard error (black rectangles) and deviation (bars) values in patients with head and neck carcinoma according to their embryologic origins, i.e., endodermal (ENDO) (n 5 62) versus ectodermal (ECTO) (n 5 13). For conventions, see Figure 4. markers include the binding sites for the T antigen, as revealed by lectinhistochemistry with biotinylated PNA; the endogenous binding sites for this disaccharide, as revealed by glycohistochemistry with a biotinylated T antigen-carrying neoglyoprotein; the binding sites for galectin-1, as revealed by lectinhistochemistry with biotinylated galectin-1; the binding sites for galectin-3, as revealed by lectinhistochemistry with biotinylated galectin-3; and galectin-1 and galectin-3, as detected by immunohistochemistry with specific polyclonal antigalectin-1 and antigalectin-3 antibodies, respectively. When monitoring tissue lectin expression, it is essential to determine the presence of lectin and lectin-reactive sites in serial sections. The percentage of tissue area specifically stained by the labeled probe, the histochemical concentration of the probe, and the heterogeneity-homogeneity distribution level of the probe’s histochemical staining were Galectin Expression in Head and Neck Cancers/Choufani et al. TABLE 3 Results Obtained by Means of Discriminant Analysis with the Emphasis on Identifying the Most Informative Quantitative Variables Describing the Expression of Galectin-1, Galectin-3, T Antigen, and their Respective Binding Sites in Relation to Head and Neck Squamous Cell Carcinoma Differentiation Levela Well Mod Mod Gal3-MOD (1) aGal3-MOD (1) Gal3-LI (1) P 5 0.009 Poor PNA-LI (1) T-LI (2) Gal1-CH (1) P 5 0.0004 — — — — PNA-LI (1) Gal3-MOD (1) Gal1-CH (1) P 5 0.02 The data presentation is identical to that in Table 2. The variables are listed in Table 1. The Well (n 5 35), Mod (n 5 30), and Poor (n 5 10) groups cover well, moderately, and poorly differentiated head and neck squamous cell carcinoma, respectively. a TABLE 4 Results Obtained by Means of Discriminant Analysis with the Emphasis on Identifying the Most Informative Quantitative Variables Describing the Expression of Galectin-1, Galectin-3, T Antigen, and their Respective Binding Sites in Relation to the T Variable of the TNM Classification Systema T1 T2 T-MOD (2) PNA-LI (2) Gal3-MOD (1) P 5 0.03 T3 aGal3-MOD (1) Gal3-MOD (1) P 5 0.002 T4 aGal3-MOD (1) Gal3-MOD (1) P 5 0.002 a T2 T3 — — — — — — — — T-MOD (2) aGal1-MOD (2) aGal3-CH (1) P 5 0.04 — — — — T-MOD (2) T-CH (2) PNA-LI (1) P 5 0.009 — ns — — For conventions and abbreviations, see Table 3. quantified by using computer-assisted microscopy, as detailed elsewhere for other marker types.30 –32 Thus, 18 quantitative histochemical variables described each of the 40 normal samples and the 75 HNSCC tumors under study. Data treatment was carried out by means of discriminant analysis for the 75 HNSCC tumors. A comparison of the normal samples and the HNSCC samples revealed that the percentage of cells exhibiting 1 of the 6 biological markers under study was systematically lower in HNSCC samples than in 2361 normal samples with respect to a given histologic tissue type. One exception appears in the larynx, in which the percentage of cells exhibiting galectin-3 was similar in HNSCC and normal samples. In contrast, the amounts of each marker exhibited by cells that remained glycohistochemically positive or immunohistochemically positive with respect to a given maker were relatively similar in normal and HNSCC samples. The data obtained for the expression of the T antigen/T antigen-binding site are in agreement with reports previously published in the literature by other authors. Gussack14 reported that the T antigen belongs to the family of oncofetal antigens in HNSCC. A particularity of this class of tumors is that the expression of the oncofetal antigen decreases as HNSCC aggressiveness increases at both the biologic level and the clinical level.13–15 A similar correlation was observed in the current study. Having introduced the corresponding neoglycoprotein, we extended these observations to the T antigen-binding site level. With respect to the galectins, we confirm the reported presence of these endogenous lectins in head and neck tumors.22 From this basis, we have been able to extend the field of the knowledge of galectin expression and, especially, of the galectin ligand sites. Our data indicate that there are quantitative parameters to describe modifications in galectin-3 and galectin-3reactive sites that mainly distinguish well-differentiated HNSCC from moderately differentiated HNSCC, whereas there are mainly quantitative variables to describe modifications in the T-PNA expression pattern that distinguish well-differentiated from poorly differentiated HNSCC (see Table 3). In fact, the mean value (6 standard error of the mean) for each of these quantitative variables decreased significantly (P , 0.05 to P , 0.01) from well-differentiated HNSCC through moderately to poorly differentiated HNSCC (data not shown). This suggests that a concomitant decrease in galectin-3 and galectin-3-binding site expression is an event that occurs early in the loss of HNSCC differentiation, whereas a concomitant decrease in PNA-T expression occurs later in the process of transformation of well-differentiated HNSCC into moderately to poorly differentiated HNSCC. With respect to the clinical T stage, an inverse feature was observed compared with what appeared to occur with respect to the levels of differentiation. Indeed, the T1 HNSCC tumors differed mainly from the T2 tumors on the basis of a decrease in T antigen and T antigen-reactive site expression (see Table 4), whereas the T1 HNSCC tumors differed mainly from the T3 and T4 tumors on the basis of a decrease in galectin-3 and galectin-3-reactive site expression (see Table 4). The grading and staging of the tumors from the different sites are not presented in Materials and 2362 CANCER December 1, 1999 / Volume 86 / Number 11 Methods, because the grades and stages covered all of the sites. For example, not all 5 T1 tumors were from a specific site: One tumor was from the oral cavity, 1 tumor was from the tonsils, 1 tumor was from the pharynx, and 2 tumors were from the larynx. In addition, the 10 poorly differentiated tumors also were from different sites. Thus, the data interpretation presented here from the statistical analyses was not skewed by any bias. The progressive loss of differentiation and the progression from T1 clinical stage to T4 clinical stage apparently was not reflected in the same way at the level of the modification of the glycohistochemical characteristics monitored. Table 4 also shows that modifications to the levels of expression of galectin-1 and galectin-1-binding sites correlate only marginally with the progression from T1 through T2, T3, and T4 stages. Galectin-1 and galectin-3 along with their ligands evidently behave disparately. In this connection, we observed that all 5 T1 HNSCC tumors under study were either well differentiated or moderately differentiated. In contrast, 5 of 23 T2 tumors under study were poorly differentiated, whereas 14 of 28 T4 HNSCC tumors were well differentiated (data not shown). This feature means that the differentiation level and clinical T stage clearly reflected the distinct biologic status in HNSCC. The clinical progression of patients with HNSCC from N2 to N1 was paralleled biologically by a decrease in the amount of galectin-3 expression (as evidenced by the antigalectin-3 MOD variable, which quantifies the MOD of the antigalectin-3 antibodyrelated immunohistochemical staining), a decrease in the percentage of the cells exhibiting galectin-3-reactive sites (as evidenced by the galectin-3 LI variable), and a decrease in the amount of T antigen expression (as evidenced by the PNA MOD variable; data not shown). Once again, the expression of galectin-1 and galectin-1-binding sites did not appear to play a major role in the distinction that could be made between N2 and N1 HNSCC (data not shown). It should be emphasized that fewer than half of the cases that we studied had any long term clinical follow-up that enabled prognostic conclusions to be made. These cases covered all of the anatomic sites and tumor grades and stages. We are now enlarging our series significantly, i.e., up to 300 cases, so that independent prognostic analyses can be performed for each anatomic site. With respect to the 34 patients for whom we were able to collect sufficient clinical data regarding survival status (those who died in ,12 months [the D , 12M group] and those who survived for .36 months [the SURV . 36M group] after the initial diagnosis of HNSCC), it appeared that the HNSCC tumors from the D , 12M group displayed weak but nevertheless significant differences in terms of glycohistochemical characteristics. The weak level of statistical significance is due to the rather small number of patients included in the analysis (14 patients vs. 20 patients), a factor that precludes making any definitive conclusions. However, the current results raise initial evidence for the suggestion of an independent prognostic value for the quantitative determination of lectin-related characteristics compared with conventional clinical prognosticators like the TNM variables. There were 10 T3 or T4 tumors for the 14 D , 12M patients, as expected; however, there were 4 T2 tumors and 7 N1 tumors compared with 7 N2 tumors. This means that these T2 and/or N2 tumors were not accounted for by the prognostic value of the TNM classification system, which, in the current study, underestimated the unfavorable prognosis of the patients analyzed. In the same way, there were 5 T3 tumors, 2 T4 tumors, and 2 N1 tumors in the SURV . 36M group. Thus, the TNM classification system overestimated the potentially unfavorable prognosis of the patients analyzed. Having indicated by this study the usefulness of monitoring endogenous lectin expression, other lectin-related characteristics should be taken into consideration to contribute independent prognostic values to the TNM classification system. One example may be sarcolectin, which is a sialic acid-binding and lymphokine-binding protein35 that acts as a human interferon a/b antagonist and a growth regulator,36 for which we recently demonstrated a prognostic value in oral cavity and lung tumors.32,37 In conclusion, the characterization of the expression of galectin-3 and the T antigen along with their respective binding sites and, to a lesser extent, of galectin-1 and galectin-1-binding sites appears to contribute additional prognostic value to conventional clinical staging of patients with HNSCC. Nevertheless, the fact remains that the current study, which included 75 patients with HNSCC, is preliminary. Indeed, there are 6 different anatomic sites (which would be complemented by further sites), there are 3 different tumor grades, and there are potentially 8 different tumor stages, representing a potential tumor diversity of at least 144 tumor groups. We are now collecting HNSCC samples from various hospitals to set up a series of cases that include several hundred specimens so that significant numbers of cases can be reviewed from each tumor group. REFERENCES 1. Schantz SP, Harrison LB, Forastiere AA. Tumors of the nasal cavity and paranasal sinuses, nasopharynx, oral cavity, and oropharynx. In: DeVita VT Jr., Hellman S, Rosenberg SA, editors. Cancer: principles and practice of oncology. vol 1, 5th ed. Philadelphia: JB Lippincott Co., 1997:741– 847. Galectin Expression in Head and Neck Cancers/Choufani et al. 2. 3. 4. 5. 6. 7. 8. 9. 10. 11. 12. 13. 14. 15. 16. 17. 18. 19. 20. 21. American Joint Committee on Cancer. Manual for staging of cancer. 3rd ed. Philadelphia: JB Lippincott Co., 1988. Vaidya MM, Borges AM, Pradhan SA, Rajpal RM, Bhisey AN. Altered keratin expression in buccal mucosal squamous cell carcinoma. J Oral Pathol Med 1989;18:282– 6. Klijanienko J, El-Naggar A, De Braud F, Micheau C, Janot F, Luboinski B, et al. Keratins 6, 13 and 19. Differential expression in squamous cell carcinoma of the head and neck. Anal Quant Cytol Histol 1993;15:335– 40. Balm AJ, Hageman PC, Van Doornewaard MH, Groeneveld EM, Ivanyi D. Cytokeratin 18 expression in squamous cell carcinoma of the head and neck. Eur Arch Otorhinolaryngol 1996;253:227–33. Mattijssen V, Peters HM, Schalkwijk L, Manni JJ, van’t HohGrrotenboer B, de Mulder PH, et al. E-cadherin expression in head and neck squamous cell carcinoma is associated with clinical outcome. Int J Cancer 1993;55:580 –5. Schipper JH, Unger A, Jahnke K. E-cadherin as a functional marker of the differentiation and invasiveness of squamous cell carcinoma of the head and neck. Clin Otolaryngol 1994; 19:381– 4. Patel AM, Incognito LS, Schechter GL, Wasilenko WJ, Somers KD. Amplification and expression of EMS-1 (cortactin) in head and neck squamous cell carcinoma cell lines. Oncogene 1996;12:31–5. Inki P, Joensuu H, Grenman R, Klemi P, Jalkanen M. Association between syndecan-1 expression and clinical outcome in squamous cell carcinoma of the head and neck. Br J Cancer 1994;70:319 –23. McGaffery JD, Gapany M, Faust RA, Davis AT, Adams GL, Ahmed K. Nuclear matrix proteins as malignant markers in squamous cell carcinoma of the head and neck. Arch Otolaryngol Head Neck Surg 1997;123:283– 8. Sakr WA, Zarbo RJ, Jacobs JR, Crissman JD. Distribution of basement membrane in squamous cell carcinoma of the head and neck. Hum Pathol 1987;18:1043–50. Gabius HJ. Concepts of tumor lectinology. Cancer Invest 1997;15:454 – 64. Wolf GT, Carey TE, Schmaltz SP, McClatchey KD, Poore J, Glaser L, et al. Altered antigen expression predicts outcome in squamous cell carcinoma of the head and neck. J Natl Cancer Inst 1990;82:1566 –72. Gussack GS. Oncofetal antigen expression in head and neck carcinoma. Laryngoscope 1992;102:168 –76. Bryne M, Gravdahl C, Koppang HS, Kjaerheim A, Dabelsteen E. Is the carbohydrate sialosyl-Tn a marker for altered, nonmalignant activity in squamous epithelium in the head and neck region? J Pathol 1995;175:237– 42. Gabius HJ. Tumor lectinology: at the intersection of carbohydrate chemistry, biochemistry, cell biology and oncology. Angew Chem Int Edit 1988;27:1267–76. Danguy A, Kayser K, Bovin NV, Gabius HJ. The relevance of neoglycoconjugates for histology and pathology. Trends Glycosci Glycotechnol 1997;7:261–75. Gabius HJ, Schröter C, Gabius S, Brinck U, Tietze LF. Binding of T-antigen-bearing neoglycoprotein and peanut agglutinin to cultured tumor cells and breast carcinomas. J Histochem Cytochem 1990;38:1625–31. Gabius HJ, Gabius S, editors. Lectins and glycobiology. Heidelberg: Springer Verlag, 1993. Gabius HJ. Animal lectins. Eur J Biochem 1997;243:543–76. Gabius HJ, Brehler R, Schauer A, Cramer F. Localization of 22. 23. 24. 25. 26. 27. 28. 29. 30. 31. 32. 33. 34. 35. 36. 37. 2363 endogenous lectins in normal human breast, benign breast lesions and mammary carcinomas. Virch Arch B Cell Pathol Mol Pathol 1986;52:107–15. Gillenwater A, Xu XC, El-Naggar AK, Clayman GL, Lotan R. Expression of galectins in head and neck squamous cell carcinoma. Head Neck 1996;18:422–32. Xu XC, El-Naggar AK, Lotan R. Differential expression of galectin-1 and galectin-3 in thyroid tumors. Potential diagnostic implications. Am J Pathol 1995;147:815–22. Bresalier RS, Yan PS, Byrd JC, Lotan R, Raz A. Expression of the endogenous galactose-binding protein galectin-3 correlates with the malignant potential of tumors in the central nervous system. Cancer 1997;80:776 – 87. Hyams VJ, Batsakis JG, Michaels L, editors. Tumors of the upper respiratory tract and ear. Atlas of tumor pathology. Washington, DC: Armed Forces Institute of Pathology, 1988. Gabius HJ, Bardosi A. Neoglycoproteins as tools in glycohistochemistry. Progr Histochem Cytochem 1991;22:1– 66. Gabius HJ, Gabius S, editors. Glycosciences: status and perspectives. London: Chapman & Hall, 1997. Gabius HJ. Influence of type of linkage and spacer on the interaction of b-galactoside-binding proteins with immobilized affinity ligands. Anal Biochem 1990;189:91– 4. Siebert HC, Adar R, Arango R, Burchert M, Kaltner H, Kayser G, et al. Involvement of laser photo CIDNP (chemicallyinduced dynamic nuclear polarization)-reactive amino-acid side chains in ligand binding by galactoside-specific lectins in solution. Eur J Biochem 1997;249:27–38. Remmelink M, Salmon I, Goldschmidt D, Decaestecker C, Nemec E, Berthe JV, et al. Quantitative measurements of desmin and vimentin immunostains and cell density in leiomyomas and leiomyosarcomas. Anal Cell Pathol 1996; 12:25– 44. Goldschmidt D, Decaestecker C, Berthe JV, Gordower L, Remmelink M, Danguy A, et al. The contribution of computer-assisted methods for histopathological classification of adipose tumors. Lab Invest 1996;75:295–306. Saussez S, Marchant H, Nagy N, Decaestecker C, Hassid S, Jortay A, et al. Quantitative glycohistochemistry defines new prognostic markers in cancers of the oral cavity. Cancer 1998;82:252– 60. Camby I, Nagy N, Lopes MB, Schäfer BW, Maurage CA, Ruchoux MM, et al. Supratentorial pilocytic astrocytomas, astrocytomas, anaplastic astrocytomas and glioblastomas are characterized by a differential expression of S100 proteins. Brain Pathol 1999;9:825– 43. Johnson RA, Wicheren DW. Applied multivariate statistical analysis. 3rd ed. New York: Prentice-Hall International, 1992. Zeng FY, Gerke V, Gabius HJ. Characterization of the macrophage migration inhibitory factor-binding site of sarcolectin and its relationship to human serum albumin. Biochem Biophys Res Commun 1994;200:89 –94. Chany-Fournier F, Jiang PH, Chany C. Sarcolectin and interferon in the regulation of cell growth. J Cell Physiol 1990; 145:173– 80. Kayser K, Bovin NV, Korchagina EY, Zeilinger C, Zeng FY, Gabius HJ. Correlation of expression of binding sites for synthetic blood group A-, B- and H-trisaccharides and for sarcolectin with survival of patients with bronchial carcinoma. Eur J Cancer 1994;30A:653–7.