Multi-national Gene Expression Profiling of Oral Squamous Cell Carcinomas. Biological Pathways Regardless of Differences Related to Life-style and Ethnicity

Oral squamous cell carcinoma (OSCC) is a major health problem in many developing countries, representing more than 25% of all new cancer cases in some countries such as India and Sri Lanka. Cigarette smoking, smokeless tobacco (ST), alcohol use and chewing of betel quid (BQ) are the main risk factors associated with OSCC development. Infection with highrisk Human papillomavirus (HPV) is an emerging risk factor, particularly for oropharyngeal cancers. Most of the OSCC cases are diagnosed at advanced stages, being one of the factors related to the high mortality rate of this cancer. A better understanding of the molecular biology of OSCC development might lead to improved methods related to detection, assessing prognosis and novel treatments of this malignancy. Over the recent years, microarray-based technologies have become commonly used techniques for analyzing gene expression and chromosomal alterations in human cancers and other disease conditions. These high-throughput technologies enable genome-wide analysis of changes in gene expression or chromosomal deletions/amplifications in the pathological samples to be studied. In the search for possible molecular biomarkers for OSCCs, a series of studies were carried out in the work described here, where cDNA microarrays and arraycomparative genomic hybridization (array-CGH) were applied to examine possible changes in gene expression and DNA copy number alterations in OSCC samples compared to their pairwised normal controls and between OSCC samples from different populations. In addition to examining gene expression in OSCC, the study included analysis of samples from a potentially malignant disorder known as oral submucous fibrosis (OSF), common in Asian populations, to ascertain whether genetic aberrations found in OSCC could be observed in early phases of carcinogenesis. Gene expression profiles and chromosomal alterations were studied in OSCC/OSF samples from Sri Lankan, Indian, Swedish and UK patients using cDNA microarrays and array-CGH. For gene expression profile (Paper I), 15 cases of OSCCs from Sri Lanka and their pairwised normal controls were examined. Following RNA extraction from all samples, cDNA was synthesized and labeled with Cy3 (tumor cDNA) and Cy5 (normal cDNA). Labeled tumor and normal cDNA were hybridized to 31k cDNA microarrays, slides were scanned and images were subjected to analysis with Genepix and J-Express computer software programmes. 262 genes (189 up-regulated and 73 down-regulated) were found to be differentially expressed between tumors and normal controls with 66 genes of known function and 66 novel genes. Among the group of genes of known function those found were CAV1, CAV2, COL4A1, MMP1, MMP3, PLAU, SPARC, TNC (all up-regulated) and AZGP1, KRT19 and S100A1 (all down-regulated). Microarray results for nine genes were verified with RTqPCR. Hierarchical clustering of the samples based on the differentially expressed genes did not show any clear relationship between sample clustering and the clinicopathological data, except for two samples (one verrucous carcinoma and one advanced tumor) that were clustered separately. In Paper II, gene expression profiles from 19 OSCCs from Sweden (n=8) and UK (n=11) were examined and compared between these two populations. RNA was extracted from all OSCCs and cDNA was synthesized and labeled with Cy3. For controls, human universal reference RNA was used for cDNA synthesis and labeling with Cy5. Labeled cDNAs were hybridized to 21k human oligonucleotide microarrays, slides were scanned and images were subjected to analysis with GenePix and J-Express. Here, 42 genes (including APOL3, NT5E, HMGA1, FASN and FOS) were found as being differentially expressed between the two populations compared to controls. Expression of three genes was validated with RT-qPCR. Upon hierarchical clustering, there was a tendency for the samples from the same population to group together. For chromosomal alterations (Paper III) 24 cases of OSCCs (12 from Sri Lanka and 12 from India) and 6 OSF (India) samples were studied. Following total DNA extraction from all samples, tumor and control DNA (Human Universal Reference DNA) were digested and labeled with Cy3 (tumors) and Cy5 (control) and further hybridized to arrays containing 4500 Bacterial artificial chromosomes (BAC) and P1 artificial chromosomes (PAC) clones. Array- CGH resulted in 349 candidate genes located in deleted (34 genes) or amplified (30 genes) different chromosomal regions in these samples common to both populations, in addition to 285 genes located in 66 chromosomal regions found as deleted or amplified in either Indian or Sri Lankan samples. Further, we selected one gene found to be deleted in the samples from both countries, namely S100A14, for further validation analysis using immunohistochemistry (IHC) and genetic variation study by RFLP (Restriction fragment length polymorphism). IHC showed decreased expression of S100A14 in OSCC archival samples compared to normal oral mucosa, and a relocalization from membrane to cytoplasmic expression. RFLP for one SNP (461A>G) demonstrated a significant difference in genotypes of OSCC and OSF. These results together with our previous findings on cases of OSCCs studied from Sudan and Norway, demonstrate that 72 genes were found to be common for all six populations studied. Among these were BAX, CCND1, COL4A1, DAPK1, FGF3, FGF4, JUNB, MMP1, MMP3, PLAU, SPARC, TNC, TGFB1, several S100 gene family members (including S100A14) and TP53. Of particular note, there were only small differences in gene aberrations in OSF compared to OSCCs. These results suggest that genetic alterations occur early during OSCC development, and that there are genes commonly involved in OSCC development regardless of the life-style and source of the material to be studied. We suggest S100A14 as a possible tumor biomarker for OSCCs together with COL4A1, MMP1, MMP3, PLAU, SPARC and TNC. Findings in the present work further suggest that there is a specific genotype of OSF which might be related to an increased risk of OSCC development