Differential expression profiling of head and neck squamous carcinoma: significance in their phenotypic and biological classification

ª Oncogene (2002) 21, 8206 – 8219 2002 Nature Publishing Group All rights reserved 0950 – 9232/02 $25.00 www.nature.com/onc Differential expression profiling of head and neck squamous carcinoma: significance in their phenotypic and biological classification Adel K El-Naggar*,1, Hyung W Kim2, Gary L Clayman3, Madelene M Coombes1, Bao Le1, Syeling Lai1, Feng Zhan2, Mario A Luna1, Waun Ki Hong4 and J Jack Lee2 ONCOGENOMICS 1 Department of Pathology, The University of Texas MD Anderson Cancer Center, Houston, Texas, TX 77030, USA; 2Department of Biostatistics, The University of Texas MD Anderson Cancer Center, Houston, Texas, TX 77030, USA; 3Department of Head and Neck Surgery, The University of Texas MD Anderson Cancer Center, Houston, Texas, TX 77030, USA; 4Department of Thoracic/Head and Neck Medical Oncology, The University of Texas MD Anderson Cancer Center, Houston, Texas, TX 77030, USA The genetic events associated with the development and progression of head and neck squamous carcinoma (HNSC) are largely unknown. We analysed 12 matched pairs of histologically normal squamous mucosa and tumor specimens from six conventional and six phenotypic variants HNSC to define the differentially expressed genes in these tumors. Parallel expression analysis of 8055 unique genes was performed, and the level of the hybridization signal for each gene was measured after normalization. Hierarchical cluster analysis of the expressed genes showed distinct interand intra-tumoral patterns in and between conventional squamous carcinoma and squamous carcinoma variants. We also identified 26 (0.32%) differentially expressed genes that were consistently different between matched pairs of normal and tumor specimens; a selected set of the overexpressed genes was validated using real-time quantitative RT – PCR. The majority of the genes were associated with differentiation and proliferation. Our study defines a set of genes that could form the basis for the construction of limited HNSC targeted expression array and indepth studies and further highlights gene profile differences that may be useful in pathobiologic classification of HNSC. Oncogene (2002) 21, 8206 – 8219. doi:10.1038/sj.onc. 1206021 Keywords: gene expression; microarray analysis; head and neck squamous carcinoma; biological classification; squamous carcinoma subtypes Introduction Head and neck squamous carcinoma (HNSC) is the fifth most common cancer in the United States with an estimated 40 000 annual new cases (Greenlee et al., *Correspondence: AK El-Naggar, Department of Pathology, Box 85, The University of Texas MD Anderson Cancer Center, 1515 Holcombe Blvd. Houston, Texas, TX 77030, USA; E-mail: Received 12 December 2001; revised 28 August 2002; accepted 3 September 2002 2000) and a mortality rate of approximately 50% (Parkin et al., 1999). These tumors encompass a spectrum of phenotypic entities and variable biological behaviors within the currently used TNM staging system (Helliwell, 2001). The majority of HNSCs, represents the conventional keratinizing variant; less common are the verrucous, papillary, basaloid and the sarcomatoid subtypes. While the poorly differentiated grade of the conventional, the basaloid and the sarcomatoid subtypes are considered highly malignant, the verrucous, papillary and the well-differentiated grade of the conventional types are generally less aggressive (Banks et al., 1992; Ellis et al., 1987; Medina et al., 1984; Suarez et al., 2000; Wain et al., 1986). Despite advances in the diagnosis, management, and molecular understanding of HNSC, minimal improvement in patient survival and quality of life has been achieved (Califano et al., 1996; El-Naggar et al., 1995, 1997; Lai et al., 1995; Vokes et al., 1993). Factors that impede progress in this field are late detection, phenotypic heterogeneity, and the lack of surrogate biomarkers for better stratification of individual patients. Large-scale gene-expression profiling of experimental and primary tumors have recently been tested for the identification and understanding of the biological function of genes involved in critical pathways associated with tumorigenesis (Diehn et al., 2000; Emmert-Buck et al., 2000; Pollack et al., 1999; Zhang et al., 1997). Studies of different human cancer subtypes using cDNA microarray analysis have led to novel classification and biological profiling of certain tumors (Anbazhagan et al., 1999; Chaib et al., 2001; Huang et al., 2000; Ono et al., 2000; Petersen et al., 2000; Sallinen et al., 2000; Sgroi et al., 1999; Shim et al., 1998; Wang et al., 2000). Similar studies of head and neck cancer have been limited in tumor type and number and used different array platforms (Alevizos et al., 2001; Leethanakul et al., 2000; Villaret et al., 2000). None of the studies addressed the global genomic expression profiles in the different phenotypic variants of squamous carcinoma. To identify the differentially expressed genes associated with HNSC and to investigate their role in the classification of HNSCs, we analysed total RNA Gene-expression analysis of head and neck squamous carcinoma AK El-Naggar et al 8207 extracted from 12 matched, macrodissected, histologically normal squamous mucosa and carcinoma specimens using a membrane-based gene-expression array platform. Results Clinicopathologic and patient characteristics are summarized in Table 1. The 12 tumors were composed of six conventional squamous cell carcinoma (SCC) and six SCC variants. The expression levels of specific genes were determined by densitometric scanning of hybridized signals. Comparative analysis of hybridized blots of normal and tumor tissues revealed several genes that were differentially expressed in several patients. Figure 1 presents scatter plots of logarithm-10, transformed gene-expression intensity for tumor versus normal tissue in the first experiment of tissue from patient 2. Figure 1 shows that the raw gene expression data had higher intensity in the normal versus tumor tissue before normalization (left panel); after normalization, the loess line was shifted to the 458 line as intended (right panel). Most of the housekeeping genes (in orange) were also distributed along the 458 line. Selected genes are marked in circles and printed with the Genebank accession number. Table 2 presents the ratio of highly expressed genes in each HNSC patient. Only consistently differentially expressed genes identified by at least three of the five methods defined in the Statistical Considerations section were selected. Using those criteria, we identified 26 distinct genes with significantly different expression between matched pairs of tumor and normal squamous mucosa tissue. For membrane GF200, seven genes were highly expressed in tumor tissue (in bold), and none were highly expressed in normal tissue from conventional SCC patients. On the other hand, eight genes were highly expressed in normal tissue (in italic), and none were highly expressed in (...truncated)