Oncogene (2005) 24, 4155–4161 & 2005 Nature Publishing Group All rights reserved 0950-9232/05 $30.00 www.nature.com/onc SHORT REPORT Gene profiling reveals specific oncogenic mechanisms and signaling pathways in oncocytic and papillary thyroid carcinoma Olivier Baris*,1, Delphine Mirebeau-Prunier1, Fre´ de´ rique Savagner1, Patrice Rodien1,2, Benoit Ballester3,Be´ atrice Loriod3, Samuel Granjeaud3, Serge Guyetant4, Brigitte Franc5, Re´ mi Houlgatte3, Pascal Reynier1 and Yves Malthiery1 1INSERM EMI-U 0018, Laboratoire de Biochimie et Biologie Mole´culaire, CHU, 4 rue Larrey, Angers F-49033, France; 2Service d’Endocrinologie, Nutrition et Me´decine Interne, CHU, Angers F-49033, France; 3Laboratoire TAGC/INSERM ERM 206, Marseille F-13009, France; 4Laboratoire d’Anatomie Pathologique, CHRU, Tours F-37044 Cedex 1, France; 5Laboratoire d’Anatomie Pathologique, Hoˆpital A Pare´, Boulogne F-92104, France The oncogenic pathways in mitochondrial-rich thyroid more frequently in the thyroid gland. Oncocytic thyroid carcinomas are not clearly understood. To investigate the carcinomas are defined as malignant epithelial tumors of possible implication ofmitochondrial abundance in the the thyroid, composed exclusively or predominantly genesis ofthyroid tumors, we have explored the gene (75% or more) of oncocytic cells presenting cytoplasmic expression profile ofsix oncocytic carcinomas and six accumulations of dilated mitochondria. Several studies mitochondrial-rich papillary carcinomas using cDNA- have confirmed the increased activity of the mitochon- microarray technology. A supervised approach allowed drial respiratory chain enzymes in these tumors (Ebner us to identify 83 genes differentially expressed in the two et al., 1991; Savagner et al., 2001a). These lesions are types ofcarcinoma. These genes were classified according mainly classified into oncocytic follicular carcinomas to their ontologic profiles. Three genes, NOS3, alpha- (also known as Hu¨ rthle cell carcinomas). The identifica- actinin-2 and alpha-catenin, suspected ofplaying a role in tion of a Hu¨ rthle cell papillary carcinoma subgroup tumor genesis, were explored by quantitative RT–PCR suggests a close relationship between ordinary PTCs and analysis and immunohistochemistry. Ofthe 59 genes oncocytic carcinomas (Cheung et al, 2000). Also, overexpressed in papillary carcinomas, 51% were involved ordinary PTCs present frequently an increase in in cell communication. Ofthe 24 genes overexpressed in mitochondrial content that may be involved in the oncocytic carcinomas, 84% were involved in mitochon- tumor genesis (Haugen et al., 2003). drial and cellular metabolism. Our results suggest that There is no evidence to suggest that the etiology of mitochondrial respiratory chain complexes III and IV oncocytic malignancies of thyroid is distinct from that play a significant role in the regulation ofreactive oxygen of nononcocytic lesions. The somatic mutations in species production by oncocytic tumors. oncocytic malignancies of thyroid follicular cells are Oncogene (2005) 24, 4155–4161. doi:10.1038/sj.onc.1208578 similar to those of the ordinary papillary and follicular Published online 4 April 2005 carcinomas. Rearrangements involving the proto-onco- gene RET as well as anomalies in the B-RAF, NTRK1, Keywords: carcinoma; thyroid; mitochondria RAS, TC-1 and APC genes have been associated with the development of PTCs (Camiel et al., 1968; Fusco et al., 1987; Chua et al., 2000; Kimura et al., 2003). Mutations in the RAS and PTEN genes, and a fusion between Pax8 and PPARg, have been implicated in the Introduction pathogenesis of follicular thyroid carcinomas (Kroll et al., 2000; Moretti et al., 2000). Nevertheless, the basic ONCOGENOMICS The WHO classification distinguishes between two types mechanisms and signaling pathways involved in the of differentiated thyroid cancer originating from folli- genesis of oncocytic thyroid carcinomas are still poorly cular cells. Papillary thyroid carcinoma (PTC) is known. diagnosed on the basis of architectural and nuclear The role of mitochondrial proliferation in cancer features, whereas follicular thyroid carcinoma is defined development has been evoked (Augenlicht et al., 1999). by capsular or vascular invasion in follicular epithelial A recent cDNA-microarray analysis of oncocytic cell neoplasms. Oncocytic tumors are epithelial tumors thyroid tumors has demonstrated the profound mod- that affect a wide variety of human tissues but occur ification of energy metabolism in oncocytoma (Baris et al., 2004). In this study, the upregulation of genes *Correspondence: O Baris; coding for enzymes playing a role in the glycolysis, E-mail: [email protected] Received 1 June 2004; revised 27 October 2004; accepted 31 January the tricarboxylic acid cycle and oxidative phosphory- 2005; published online 4 April 2005 lation suggested the existence of an aerobic glycolytic Gene profiling of mitochondrial-rich thyroid tumors O Baris et al 4156 mechanism in oncocytic adenomas and, unexpectedly, in oxidative phosphorylation, mitochondrial proliferation oncocytic carcinomas as well. In another study, the and cell proliferation in these tumors (Savagner et al., upregulation of the PGC-1-related coactivator (PRC) 2003). indicated a close relationship between the regulation of In the present study, we compare the expression profiles of six oncocytic follicular carcinomas and six ordinary mitochondrial-rich PTCs selected from a Figure 1 (a) Hierarchical average linkage clustering of the 83 most discriminating genes between follicular oncocytic and PTCs. Samples were obtained from 12patients presenting sporadic thyroid tumors, six oncocytic follicular carcinomas and six ordinary mitochondrial-rich papillary carcinomas, diagnosed according to the WHO classification. All the samples were rendered anonymous before the study. We confirmed the rich mitochondrial content of the tumor cells by using a monoclonal antibody against a complex IV mitochondrial subunit of the respiratory chain (clone 113-1, Biogenex laboratories Inc., San Ramon, CA, USA). According to immunostaining scores, taking into account the signal localization and intensity and the percentage of positive cells, oncocytic carcinomas scored at 3 (more than 75% of staining), papillary carcinomas scored at 2(between 25and 75% of staining) and 3. Patients included three men and nine women, with a mean age of 48 years (range 19–75 years). The average tumor size for oncocytic carcinomas was 29715 mm (mean7s.d.; range 15– 55 mm) and 23710 mm (mean7s.d.; range 10–40 mm) for PTCs. Gene expression was analysed with several other thyroid tumors by hybridization of nylon Hybond N þ membrane (Amersham Pharmacia Biotech, Little Chalfont, UK) cDNA arrays with radioactive probes, according to the protocols described elsewhere (Thieblemont et al., 2004). The arrays displayed PCR products spotted from 6720 selected IMAGE human cDNA and control clones. The cDNA clones were selected because of their proven or putative implication in carcinogenesis (list available at http:// tagc.univ-mrs.fr/pub/). Total RNA was isolated from frozen thyroid samples using a standard guanidium isothiocyanate protocol (Trizol Reagent, Life Technologies, Gaithersburg, MD, USA). RNA integrity was determined on an RNA 6000 nano labchip (Agilent technologies, Waldbronn, Germany). cDNA from each thyroid sample was obtained from 5 mg of total RNA by simultaneous reverse transcription and [a-33P]dCTP labeling as described elsewhere (http://tagc.univ-mrs.fr/pub/). Each cDNA was hybridized on an individual array prepared as described below. After washing, hybridization images were obtained by scanning with an imaging device (Fuji BAS 5000, Raytest, Paris, France). Signal intensities were quantified using ArrayGauge software (Fujifilm, Japan). Supervised analysis was used to highlight the most important genes discriminating oncocytic carcinomas from papillary carcinomas. The discriminating score (DS) was calculated as follows, DS ¼ (m1Àm2)/(s1 þ s2), where m1 and s1, respectively, represent the mean and the s.d. of the expression levels of a given gene in sample subgroup 1, and m2 and s2, respectively, represent the mean and the s.d. of the expression levels of the same gene in sample subgroup 2(Golub et al., 1999). Owing to the small number of samples in each group, we had to increase the error risk to 1% to select discriminant genes (Magrangeas et al., 2003). This higher risk gives 3.2 false-positive genes meaning that, among 25 genes of our list (83/3.2), one gene could have been randomly selected. Clustering analysis was performed using the Pearson correlation and the average linkage method (Eisen et al., 1998). (b) Gene function distribution (in %) according to the ontologic analysis. Black: papillary carcinomas; Gray: oncocytic carcinomas. Only genes with a known function were considered (N ¼ 77 genes). We searched for the biological processes in which the 77 genes with a known function were involved, according to the GO database. We used the statistical methods available on the FATIGO web server (http://fatigo.bioin- fo.cnio.es, Al-Shahrour et al., 2004) to find over- and under- represented GO terms in the two types of carcinomas. The statistical contrast was performed at the GO level 3, which constitutes the best compromise between information quality and the number of gene annoted Oncogene Gene profiling of mitochondrial-rich thyroid tumors O Baris et al 4157 previous cDNA-microarray analysis on thyroid tumors gene profiling and immunostaining studies,