Gene Expression in Thyroid Autonomous Adenomas Provides Insight Into Their Physiopathology

Home , ALPI, C7 protein, Caveolin 1, DLG4, ELK4, FBLN2

Gene expression in thyroid autonomous adenomas provides insight into their physiopathology

Sandrine Wattel1, Hortensia Mircescu1, David Venet1, Agnes Burniat1, Brigitte Franc2, Sandra Frank1, Guy Andry3, Jacqueline Van Sande1, Pierre Rocmans1, Jacques E Dumont1, Vincent Detours1 and Carine Maenhaut*,1

1Institute of Interdisciplinary Research, School of Medicine, Free University of Brussels, Brussels, Belgium; 2Service d’Anatomie et de Cytologie Pathologiques, Hoˆpital A Pare´ (AP-HP), Universite´ de Versailles, St-Quentin-en-Yvelines, France; 3Department of Surgery, Institut Bordet, 1000 Brussels, Belgium

The purpose of this study was to use the microarray Oncogene (2005) 24, 6902–6916. doi:10.1038/sj.onc.1208849; technology to define expression profiles characteristic of published online 18 July 2005 thyroid autonomous adenomas and relate these findings to physiological mechanisms. Experiments were performed Keywords: thyroid; autonomous adenoma; gene expres- on a series of separated adenomas and their normal sion; microarray counterparts on Micromax cDNA microarrays covering 2400 genes (analysis I), and on a pool of adenomatous tissues and their corresponding normal counterparts using microarrays of 18 000 spots (analysis II). Results for Introduction genes present on the two arrays corroborated and several gene regulations previously determined by Northern Autonomous thyroid adenomas are monoclonal encap- blotting or microarrays in similar lesions were confirmed. sulated benign tumors that grow, metabolize iodide, and Five overexpressed and 24underexpressed genes were also secrete thyroid hormones independently of the normal confirmed by real-time RT–PCR in some of the samples thyrotropin (TSH) control, the main proliferative and used for microarray analysis, and in additional tumor functional stimulus of the thyroid gland (Corvilain specimens. Our results show: (1) a change in the cell et al., 2001). They can be either solitary or part of a populations of the tumor, with a marked decrease in multinodular goiter, and grow rather slowly. Their lymphocytes and blood cells and an increase in endothelial frequency is an inverse function of the iodine dietary cells. The latter increase would correspond to the supply, and they are therefore uncommon in the USA or establishment of a close relation between thyrocytes and in Japan, but very common in Europe (Knudsen et al., endothelial cells and is related to increased N-cadherin 2000; Tonacchera et al., 2000; Dremier et al., 2002). expression. It explains the increased blood flow in the Their autonomous secretion of thyroid hormones tumor; (2) a homogeneity of tumor samples correlating decreases TSH secretion by the classical negative feed- with their common physiopathological mechanism: the back exerted on the pituitary thyrotrophs and on TRH constitutive activation of the thyrotropin (TSH)/cAMP secretion. This leads to functional quiescence of the cascade; (3) a low proportion of regulated genes consistent nonaffected tissue. If untreated, in the presence of with the concept of a minimal deviation tumor; (4) a sufficient iodine supply to allow the synthesis and higher expression of genes coding for specific functional secretion of excess thyroid hormones and once they proteins, consistent with the functional hyperactivity of reach a certain size, autonomous adenomas cause the tumors; (5) an increase of phosphodiesterase gene hyperthyroidism (Ermans and Camus, 1972; Corvilain expression which explains the relatively low cyclic AMP et al., 2001). They therefore represent a major medical levels measured in these tumors; (6) an overexpression of problem and account for about half of the cases of antiapoptotic genes and underexpression of proapoptotic hyperthyroidism in Europe (Russo et al., 1995; Bauch, genes compatible with their low apoptosis rate; (7) an 1998; Corvilain et al., 2001). Clinically, they are overexpression of N-cadherin and downregulation of diagnosed by the presence of hot nodules that take up 99m caveolins, which casts doubt about the use of these high amounts of radioiodide or Tc pertechnetate, and expressions as markers for malignancy. are surrounded by poorly radioactive quiescent tissue, and by low TSH serum levels, with normal or elevated thyroid hormone levels. Their histology represents a mixture in various proportions of micro- and macro- *Correspondence: C Maenhaut, IRIBHM, University of Brussels, follicles surrounded by a well-defined capsula. The Campus Erasme, Building C, 808 route de Lennik, B-1070 Brussels, Belgium; E-mail: [email protected] majority have a monoclonal origin (Namba et al., Received 1 March 2005; revised 3 May 2005; accepted 11 May 2005; 1990; Krohn et al., 1998). A very minor proportion may published online 18 July 2005 degenerate into malignant follicular carcinomas, Gene expression in thyroid autonomous adenomas S Wattel et al 6903 whereas a much higher proportion of cold (i.e. non- Adenomas were identified by scintigraphy and in iodide trapping) follicular adenomas do. some cases by a higher in vitro iodide uptake than the TSH exerts most of its effects via a seven-transmem- control quiescent tissue. The analysed tissue RNAs were brane domain receptor positively coupled to adenylate validated for untreated patients by a NIS RNA cyclase, leading to a rise in intracellular cAMP levels expression higher than in control tissues, as evaluated (Dremier et al., 2002). In Europe and Japan, one of the by RT–PCR (see below). For control purposes, one main mechanisms responsible for the hyperfunction and multinodular goiter with no hyperfunctioning adenoma growth of thyroid autonomous adenomas is the (v6) was compared to normal adjacent tissue. For serial constitutive activation of the cAMP-dependent mito- real-time RT–PCR analysis, 10 samples from 10 new genic cascade, through mutations conferring constitu- patients, all of them untreated, were analysed. tive activity of the TSH receptor (50–80%) or an activating mutation of Gsa (8%) (Russo et al., 1995; Analysis of separated samples (analysis I) Van Sande et al., 1995; Fuhrer et al., 1997; Tonacchera et al., 1999; Vanvooren et al., 2002). Similar mutations Reproducibility of the TSA method in gene expression account for hyperfunctioning nodules in multinodular analysis Tyramide signal amplification in microarray goiters (Tonacchera et al., 1999, 2000; Krohn et al., gene expression is a very useful method when limiting 2000). Their phenotype is reproduced in transgenic mice amounts of RNA are available. To assess its reprodu- expressing the constitutive adenosine A2 receptor cibility, we labeled reverse-transcribed total RNA from (Ledent et al., 1992), constitutive Gsa (Michiels et al., the adenomatous tissue with biotin and total RNA from 1994) or the activator of Gsa choleratoxin (Zeiger et al., the control tissue with fluorescein. We performed 1997). Autonomous adenomas are thus a well-defined hybridization with the reverse combination of dyes as example of the results of long-term stimulation by the well (dye swap), and calculated the averages of the physiological TSH receptor/cAMP-dependent cascade intensity ratios obtained with the two opposite dye of human thyroid tissue in vivo (Dumont et al., 1989). As combinations (tumor/control) (Taniguchi et al., 2001; such, they illustrate the concept of minimal deviation Karsten et al., 2002). A high correlation (96%) between tumors: they are constituted of hyperfunctioning thyroid the ratios of the intensity values was obtained for cells whose operation is uncoupled from their normal duplicate spots on the same array. When comparing the physiological control (Van Sande et al., 1988; Weber, same spot on two different slides, correlation ranged 2002). However, data are scarce about specific gene between 70 and 80%. With microarray data generated expression modifications in these tumors and especially by two independent RNA preparations from the same about their physiopathological consequences. tumor sample (a11/v3), a correlation of 73% was Numerous studies in different tumor types have obtained (data not shown). We thus concluded that shown that gene expression profiling of clinical samples tyramide signal amplification is a consistent and by microarray offers unprecedented opportunities to reproducible method, adequate for our analyses. obtain molecular signatures of the state of the tumors. In this study, we have analysed the pattern of gene Gene expression analysis shows high homogeneity between expression in hyperfunctioning autonomous thyroid different autonomous adenomas For each adenoma/ adenomas and validated the relative expression of control paired sample, experiments were duplicated with selected genes by real-time RT–PCR. The data show dye switching, and the averages of the intensity ratios that several physiological and morphological character- (tumor/control) resulting from each dye combination istics of these adenomas can be explained by their were calculated. A total of 14 tumor samples were transcriptional program. analysed (Table 1). We performed unsupervised hierarchical clustering of the samples to analyse the data. As depicted in Figure 1, a set of up- or downregulated genes Results could be identified, with similar expression in the majority of the tumors, suggesting that thyroid adeno- Two types of analyses were performed: an analysis of mas are relatively homogeneous tissues, based on gene gene expression in separated, individual samples of expression analysis. However, tumor classification adenomatous and adjacent tissues (analysis I), and a allowed us to distinguish a subset of samples, underlined general survey of gene expression on a large number of in blue (v1–a21), clustered more closely. The samples genes, for which a pool of equal amounts of RNA from that were on the most distant dendrogram branches adenoma or normal adjacent tissues from five patients were samples v4, a15 and v6. Sample v6, the multi- were compared (analysis II). Among these patients, two nodular goiter included as a control, clusters apart from had been treated with antithyroid drugs (v4, a15). The the adenoma samples. two pooled RNA preparations were also used as pooled We further examined the clinical data from these samples for real-time RT–PCR analysis. For analysis I, patients, listed in Table 1, to explore the possibility of a we used the Micromax cDNA system (PerkinElmer) and relationship between the dendrogram branching pattern analysed 13 patients, two of them being treated (v4, and one or several pathological features. We could find a15). Among the 13 adenomas, eight were solitary, while a relationship between the dendrogram pattern and the five were from a multinodular goiter. Three patients capacity of the tumor to transport iodide. Iodide were common to analysis I and analysis II. transport has been quantitated previously in the

Oncogene Gene expression in thyroid autonomous adenomas S Wattel et al 6904 Table 1 Clinical, biochemical and histological data of the samples used for microarray analysis Tumor sample Sex Birth year Age at surgery Diagnosis TSH levels T3/T4 levels T/M Ad/N

v4 M 1962 37 MTN High Normal 1.5 v1 M 1949 51 Adenoma Low Normal 27.4 v5 F 1968 29 Adenoma Low Normal ND a1 M 1970 24 Adenoma Normal Normal 3.8 a2 F 1978 17 Adenoma Low High 32.2 a8 F 1964 30 Adenoma Low Normal 4.8 a11/v3 F 1939 53 MTN Low Normal ND a4 M 1928 66 Adenoma Low High 3.6 a5 M 1944 50 MTN Low Normal 5.3 a12 F 1962 32 Adenoma Low Normal 4.8 a18 F 1956 38 MTN Low Normal 8.5 a21 F 1958 37 MTN Low Normal ND a15 F 1948 45 Adenoma Normal Normal 1.3 v6 F 1929 68 MTN, no adenoma Normal Normal ND

MTN: multinodular goiter; TSH and T3/T4 levels: levels at the time of surgery; T/M: ratio of radioiodide content between tissue (T) and medium (M), measured in slices in vitro; T/M Ad/N: ratio of T/M between adenoma (Ad) and corresponding control tissue (N); ND: not determined

Mitochondrial genes cluster: - mitoch. phosphate carrier protein - mitoch. ATP synthase sub. 9 - mitoch. 3-ketoacyl-CoA thiolase beta-sub. - nuclear aconitase mRNA, encoding mitoch. protein

Transcription factors cluster: -jun-B - tristetraproline (TTP) - CL 100 m RNA for tyrosine ppase -desmin - proto-oncogene Bcd orf1 and orf2 -ATF3 - vimentin N-terminal fragment - bone morphogenetic protein 2A -AF-9 - insulin rec. substrate-1 - filamin (ABP-280) -jun-D - transcription factor ETR-101

> 2X > 2X repressed induced Figure 1 Hierarchical clustering of the microarray data from 14 thyroid tumor samples and identiﬁcation of a subset of tumors (underlined in blue). The results represent the averages of the intensity ratios (tumor/control) resulting from each dye combination, and the genes considered are those with a median intensity >1 (B1200 genes). Each row represents a cDNA clone and each column a tumor RNA sample. Red indicates upregulation, green downregulation and black no change, according to the color scale shown. Two clusters, one indicating an upregulation of mitochondrial genes and the other a downregulation of transcription factors, are indicated

Oncogene Gene expression in thyroid autonomous adenomas S Wattel et al 6905 laboratory and was significantly higher in the adeno- downregulated. For expression ratios >3 in the 75% matous tissues than in the quiescent tissues in the group most expressed genes, 74 were upregulated and 100 of samples which clustered more tightly, but was downregulated (Table 2a and b), where the genes are comparable between normal and tumor tissue slices classified according to Gene Ontology (http://fatigo. for samples v4 and a15 of patients under methimazole bioinfo.cnio.es). treatment. Thus, the subset of tumors which cluster The following genes were found to be significantly together are also those which are characterized by a high overexpressed in both the separated pairs of samples rate of iodide transport, that is, the highest hyperactivity (analysis I) and in the pooled samples (analysis II): with regard to iodide uptake. In accordance with this MT1, FcGBP and CKB. hyperactivity, NIS mRNA expression was shown to be The following genes were found to be significantly increased in all the untreated autonomous adenomas, as downregulated in both analysis I and II: clusterin, bcd revealed by real-time RT–PCR (Figure 2) and pre- (COPEB), ATF3, cav1, DUSP1, junB, DNAJ, TR3 viously by Northern blotting (Deleu et al., 2000). As (NR4A1 or NGFIB), fibulin1 and C1s. expected for patients v4 and a15, under methimazole Quite a few upregulations confirm previously ob- treatment, adenomatous and normal tissues express NIS tained data: thyroperoxidase (Deleu et al., 2000), mRNA equally. deiodinase (dio1), sialylyltransferase, protein kinase Cb, Hierarchical clustering of the genes groups genes collagen IV and IX (Eszlinger et al., 2004). FcGBP has encoding mitochondrial proteins and immediate early been shown to be upregulated in follicular adenomas genes in the upper and lower parts of the panel, (O’Donovan et al., 2002). Alcohol DH (Eszlinger et al., respectively (Figure 1). A clustering of such genes has 2004) and solute carrier expressions are also increased in been previously demonstrated in oncocytomas (Baris the Tg-A2aR mice thyroids, an animal model of et al., 2004) and in papillary thyroid carcinomas autonomous adenomas (Goffard et al., 2004). Phospho- (Haugen et al., 2003). diesterase expression is classically increased in response We then focused on gene expression differences to cAMP and PKA activation (Cho-Chung, 2004). between normal and tumor tissues. Genes with greater Some other gene expressions increased by a factor than twofold and statistically significant differential between 2 and 3, such as wnt, PGF, fibromodullin and expression (Po0.05, among all the investigated tumors) ZNF transcription factors, are also increased in Tg- were identified: 11 genes were overexpressed and 22 A2aR mice thyroids (Goffard et al., 2004) or in other genes were underexpressed in most of the adenomas. cAMP-stimulated tissues (phosphodiesterases) (Lania These genes were included in the results of the general et al., 1998). Overexpression of GAPDH has been shown survey described below (Table 2). to be induced by TSH in thyroid cell cultures (Savonet et al., 1997). Decreased expression of several genes confirms General survey of gene expression (analysis II) previous data obtained in our group and by others: In a second approach, tumor RNA extracts from five ApoD, myosin (Eszlinger et al., 2004), c-fos, fosB, junB, adenoma samples were pooled and hybridized with their c-myc and TR3 (NGFIB) (Deleu et al., 2000). Several respective control pools on microarrays containing genes are downregulated in both analysis I and II, but 18 442 ESTs representing 13 000 human genes spotted only by a factor of less than 3 in analysis II, namely in duplicate. A dye-swap hybridization was performed, clusterin, bcd, ATF3, cav1, DUSP1, junB and TR3. and the few genes for which the color flip gave different Clusterin is also downregulated by TSH in human results were not considered. Considering up- or down- thyrocytes in culture (Yamazaki et al., 2003). MT1L and regulation expression ratios >2 and levels of expression MT1H, a metallothionein called MTIl in analysis I and above the first 25%, 348 genes were upregulated and 364 MT2H in analysis II, but which because of high

7 6 5 4 3 2 1

fold modulation 0 -1 B3 J3 V1 V2 V5 16 S2 S6 S7 S8 B7 B8 A15 V4 pool -2 treated tissue identification Figure 2 Measurement by RT–PCR of NIS mRNA expression in different adenomatous tissues compared to their normal adjacent tissues. The results are given in log 2 (treated: patients under methimazole treatment)

Oncogene Gene expression in thyroid autonomous adenomas S Wattel et al 6906 Table 2 Upregulated (a) and downregulated (b) genes in pooled and separated samples according to the microarray data GO at level 3 andbiological Genes (pool: ratio>3 andint>25% or confirmed;separatedsamples: ratio>2 andin most UniGene Ratio process samples; Colorflip verified)

(a) Extracellular signals Pool: CCL18: Chemokine (C–C motif) ligand 18 (pulmonary and activation-regulated) Hs.16530 6.0 IGFBP1: Insulin-like growth factor binding protein 1 Hs.102122 3.8 Receptors Pool: NPR3: Natriuretic peptide receptor C/guanylate cyclase C (atrionatriuretic peptide Hs.123655 3.3 receptor C) *FGFR3: Fibroblast growth factor receptor 3 (achondroplasia, thanatophoric Hs.1420 3.9 dwarfism) Cell–cell communication Pool: *CDH2: Cadherin 2, type 1, N-cadherin (neuronal) Hs.161 13.0 GJA4: Gap junction protein, alpha 4, 37 kDa (connexin 37) Hs.296310 3.3 Intracellular signal proteins Pool: PDE7B: Phosphodiesterase 7B Hs.283016 3.3 DUSP9: Dual-specificity phosphatase 9 Hs.144879 3.7 PRKCB1: Protein kinase C, beta 1 Hs.77202 3.5 GNAO1: Guanine nucleotide-binding protein (G protein), alpha activity Hs.296184 4.6 polypeptide O PRKWNK2: Protein kinase, lysine-deficient 2 (SDCCAG43: serologically defined colon Hs.232116 3.5 cancer antigen 43) Cell communication Pool: VWF: Von Willebrand factor Hs.110802 3.0 FLRT1: Fibronectin leucine-rich transmembrane protein 1 Hs.12523 3.2 TIAM2: T-cell lymphoma invasion and metastasis 2 Hs.529559 7.4 DLG4: Discs, large homolog 4 (Drosophila) Hs.23731 3.3 Locomotory behavior Pool: ANKH: Ankylosis, progressive homolog (mouse) Hs.168640 3.4 Metabolism Pool: MT1H: Metallothioneine 1 H Hs.2667 3.6 QPRT: Quinolinate phosphoribosyltransferase (nicotinate-nucleotide pyrophosphorylase Hs.8935 4.2 (carboxylating)) SLC27A4: Solute carrier family 27 (fatty acid transporter), member 4 Hs.248953 4.7 NMNAT2: Nicotinamide nucleotide adenylyltransferase 2 ¼ C1orf15 Hs.158244 4.0 CYP27B1: Cytochrome P450, family 27, subfamily B, polypeptide 1 Hs.199270 3.5 CILP: Cartilage intermediate layer protein, nucleotide pyrophosphohydrolase Hs.151407 4.5 SIAT1: Sialyltransferase 1 (beta-galactoside alpha-2,6-sialyltransferase) Hs.2554 8.6 CKMT2: Creatine kinase, mitochondrial 2 (sarcomeric) Hs.80691 3.4 DIO1: Deiodinase, iodothyronine, type I Hs.251415 4.5 PCSK2: Proprotein convertase subtilisin/kexin type 2 Hs.93164 3.5 TPO: Thyroid peroxidase Hs.2041 3.3 SLC25A15: Solute carrier family 25 (mitochondrial carrier; ornithine transporter) Hs.78457 4.1 member 15 Separatedsamples: GAPDH: Glyceraldehyde-3-phosphate dehydrogenase 1.95 *CKB: Creatine kinase B Hs.173724 2.9 MTÀ1l ¼ MT1X Hs.458273 4.0 Morphogenesis Pool: ESM1: Endothelial cell-specific molecule 1 Hs.41716 11.3 PARD6A: Par-6 partitioning defective 6 homolog alpha (Caenorhabditis elegans) Hs.112933 3.2 Death Pool: BIRC5: Baculoviral IAP repeat-containing 5 (survivin) Hs.1578 3.6 Growth Pool: MKI67: Antigen identified by monoclonal antibody Ki-67 Hs.80976 3.0 RBBP9: Retinoblastoma-binding protein 9 Hs.69330 3.1 RPL26: Ribosomal protein L26 Hs.406682 6.3 Extracellular structure Pool: organization and biogenesis COL4A2: Collagen, type IV, alpha 2 Hs.407912 3.3 COL9A3: Collagen, type IX, alpha 3 Hs.126248 19.5 COL17A1: Collagen, type XVII, alpha 1 Hs.117938 6.1 Transcription factors Pool: ZNF180: Zinc-finger protein 180 (HHZ168) Hs.130683 3.4 KIAA0478: KIAA0478 gene product Hs.528723 3.4 FLJ14009: Transducin-like enhancer protein 6 Hs.128417 5.9 HR: Hairless homolog (mouse) Hs.272367 3.9 TCEA2: Transcription elongation factor A (SII), 2 Hs.80598 4.0 LZK1: C3HC4-type zinc-finger protein (ZNF403) Hs.149055 3.0 Cellular physiological Pool: processes ENTPD1: Ectonucleoside triphosphate diphospho-hydrolase 1 Hs.205353 3.4 AQP5: Aquaporin 5 Hs.298023 4.7

Oncogene Gene expression in thyroid autonomous adenomas S Wattel et al 6907 Table 2 (continued ) GO at level 3 andbiological Genes (pool: ratio>3 andint>25% or confirmed;separatedsamples: ratio>2 andin most UniGene Ratio process samples; Colorflip verified)

CACNA1D: Calcium channel, voltage-dependent, L type, alpha 1D subunit Hs.23838 4.4 RAB3A: RAB3A, member RAS oncogene family Hs.27744 3.5 FLJ13265: Hypothetical protein FLJ13265 Hs.212640 4.5 KCND2: Potassium voltage-gated channel, Shal-related subfamily, member 2 Hs.202687 3.1 STX7: Syntaxin7 Hs.8906 2.0 ALPI: Alkaline phosphatase, intestinal Hs.37009 3.1 SNAP25: Synaptosomal-associated protein, 25 kDa Hs.221974 3.0 Separatedsamples: SERPING1: Serine (or cysteine) proteinase inhibitor, clade G (C1 inhibitor), member 1, Hs.384598 2.0 (angioedema, hereditary) Varia Pool: CENTB5: Centaurin, beta 5 Hs.21446 4.9 MGEA6: Meningioma expressed antigen 6 (coiled-coil proline-rich) Hs.117242 3.9 FCGBP: Fc fragment of IgG-binding protein Hs.111732 2.0 C19orf4: Chromosome 19 open reading frame 4 Hs.5012 7.4 FLJ20542: Hypothetical protein FLJ20542 Hs.6449 6.4 DXYS155E: DNA segment on chromosome X and Y (unique) 155 expressed sequence Hs.21595 6.1 LAK-4P: Expressed in activated T/LAK lymphocytes (EVER1) Hs.16165 5.0 KIAA1126: KIAA1126 protein Hs.44087 4.7 FLJ14351: Hypothetical protein FLJ14351 Hs.116104 4.6 MGC40157: Hypothetical protein MGC40157 Hs.270232 4.2 KIAA1394: KIAA1394 protein Hs.32156 3.9 MGC10772: Hypothetical protein MGC10772 (PRR7) Hs.130316 3.9 FLJ13236: Hypothetical protein FLJ13236 Hs.170298 3.9 DKFZP564C152: DKFZP564C152 protein (mucin20) Hs.69321 3.6 FLJ23186: Hypothetical protein FLJ23186 Hs.434247 3.5 FLJ21588: ASC-1 complex subunit P100 (ASCC2: activating signal cointegrator 1 complex Hs.334686 3.4 subunit 2) KIAA1157: KIAA1157 protein, PPM1H protein phosphatase 1H (PP2C domain Hs.21894 3.4 containing) KIAA1423: KIAA1423 protein Hs.274396 3.4 KIAA0690: KIAA0690 protein Hs.434251 3.3 SLAC2-B: SLAC2-B Hs.138380 3.2 LOC64150: Uterine-derived 14 kDa protein Hs.173780 3.1 MGC11316: Hypothetical protein MGC11316 (rab11-FIP4) Hs.7985 3.1 Separatedsamples: *S100 alpha protein a1 Hs.433503 3.1 IgG Fc-binding protein (FCGBP) Hs.111732 2.3 trp-1: Alternatively spliced trp-1 protein and unspliced trp-1 protein 2.1

(b) Signals Pool: FZD4: Frizzled homolog 4 (Drosophila) Hs.19545 3.3 CYR61: Cysteine-rich, angiogenic inducer, 61 Hs.8867 3.6 CXCL9: Chemokine (C–X–C motif) ligand 9 Hs.77367 3.5 LTBP2: Latent transforming growth factor beta-binding protein 2 Hs.83337 5.0 SFRP4: Secreted frizzled-related protein 4 Hs.105700 3.1 *CTGF: Connective tissue growth factor (IGFBP8) Hs.75511 2.9 Receptors Pool: GPR83: G protein-coupled receptor 83 Hs.272385 4.3 CD2: CD2 antigen (p50), sheep red blood cell receptor Hs.89476 3.4 IL1RL1: Interleukin 1 receptor-like 1 Hs.66 16.0 IL20RA: Interleukin 20 receptor, alpha Hs.21814 3.9 NPY1R: Neuropeptide Y receptor Y1 Hs.169266 3.5 Cell to cell communication Pool: *CLDN 4: Claudin4 Hs.5372 2.5 Signal transduction proteins Pool: WISP2: WNT1 inducible signaling pathway protein 2 Hs.194679 3.2 ARL4A: ADP-ribosylation factor-like 4A Hs.380784 5.6 DUSP2: Dual-speciﬁcity phosphatase 2 Hs.1183 5.7 RGS2: Regulator of G-protein signaling 2, 24 kDa Hs.78944 3.3 *DUSP5: Dual speciﬁcity phosphatase 5 Hs.2128 2.2 *DUSP1: CL 100 protein tyrosine phosphatase Hs.171695 2.8 PPP1R7: Protein phosphatase 1, regulatory subunit 7 Hs.10784 3.6 PPP1R15A: Protein phosphatase 1, regulatory (inhibitor) subunit 15A Hs.76556 3.2 KLRK1: Killer cell lectin-like receptor subfamily K, member 1 Hs.387787 3.1 RELN: Reelin Hs.12246 3.8 Separatedsamples: *DUSP1: CL 100 protein tyrosine phosphatase Hs.171695 1.9

Oncogene Gene expression in thyroid autonomous adenomas S Wattel et al 6908 Table 2 (continued ) GO at level 3 andbiological Genes (pool: ratio>3 andint>25% or confirmed;separatedsamples: ratio>2 andin most UniGene Ratio process samples; Colorflip verified)

Metabolism Pool: APOA4: Apolipoprotein A-IV Hs.1247 6.2 UBD: Ubiquitin D Hs.44532 4.2 AKR1C2: Aldo-keto reductase family 1, member C2 (dihydrodiol dehydrogenase 2; bile Hs.201967 6.2 acid-binding protein; 3-alpha hydroxysteroid dehydrogenase, type III) RNASE4: Ribonuclease, RNase A family, 4 Hs.283749 3.4 ADH1C: Alcohol dehydrogenase 1C (class I), gamma polypeptide Hs.2523 4.1 ADH1A: Alcohol dehydrogenase 1A (class I), alpha polypeptide Hs.73843 7.7 APOD: Apolipoprotein D Hs.75736 3.4 BF: B-factor, properdin Hs.69771 3.2 PGLS: 6-phosphogluconolactonase Hs.100071 4.0 DNAJB1: DnaJ (Hsp40) homolog, subfamily B, member 1 Hs.82646 3.1 DPYSL3: Dihydropyrimidinase-like 3 Hs.74566 4.0 ALDO B: Aldolase B Hs.234234 3.0 ACOX1: Acyl CoA oxidase Hs.11135 3.3 Separatedsamples: GLUL: Glutamate-ammonia ligase (glutamine synthase) Hs.442669 2.5 *HSPCA: Heat shock 90 kDa protein 1, (hsp90) Hs.446579 1.8 DnaJ: hsp40 protein homolog (DNAJA1) Hs.445203 2.1 Morphogenesis Pool: cytoskeleton SERPINF1: Serine (or cysteine) proteinase inhibitor, clade F (alpha-2 antiplasmin, pigment Hs.173594 3.7 epithelium-derived factor), member 1 MYH11: Myosin, heavy polypeptide 11, smooth muscle Hs.78344 4.2 *GSN: Gelsolin (amyloidosis, Finnish type) Hs.290070 3.2 FBLN2: Fibulin 2 Hs.198862 4.3 MFAP4: Microﬁbrillar-associated protein 4 Hs.296049 3.1 Death Pool: *GADD45A: Growth arrest and DNA-damage-inducible, alpha Hs.80409 3.5 *TNFSF10: Tumor necrosis factor (ligand) superfamily, member 10 Hs.83429 2.9 *NFKBia: Nuclear factor of kappa light polypeptide gene enhancer in B-cell inhibitor, Hs.81328 2.1 alpha *TP53BP2: Tumor protein p53-binding protein, 2 Hs.44585 2.1 *Clusterin (trpm-2, apolipoprotein J) Hs.75106 2.8 Separatedsamples: *Clusterin (trpm-2, apolipoprotein J) Hs.75106 4.3 *Lipocortin (anxa1) Hs.287558 4.3 Extracellular structure Pool: organization and biogenesis *LUM: Lumican Hs.79914 16.9 ADAMTS1: A disintegrin-like and metalloprotease (reprolysin type) with thrombospondin Hs.8230 3.1 type 1 motif, 1 CHI3L1: Chitinase 3-like 1 (cartilage glycoprotein-39) Hs.75184 5.2 SPON2: Spondin 2, extracellular matrix protein Hs.288126 4.5 *THBS2: Thrombospondin 2 Hs.458354 3.6 FBLN5: Fibulin5 Hs.11494 6.7 LAMA2: Laminin, alpha 2 (merosin, congenital muscular dystrophy) Hs.75279 4.4 Separatedsamples: PG40: Chondroitin/dermatan sulfate proteoglycan core protein 8.3 Fibulin-1D Hs.445240 2.4 col1A2: Prepro-alpha2(I) collagen Hs.232115 2.4 Gene expression Pool: transcription factor *ELK4: ELK4, ETS-domain protein (SRF accessory protein 1) Hs.169241 5.2 SERTAD1: SERTA domain containing 1, CDK4-binding protein p34SEI1 Hs.44281 4.1 FOS: V-fos FBJ murine osteosarcoma viral oncogene homolog Hs.25647 4.5 *SRF: Serum response factor (c-fos serum response element-binding transcription factor) Hs.155321 3.4 NR4A3: Nuclear receptor subfamily 4, group A, member 3 Hs.80561 3.7 CITED1: Cbp/p300-interacting transactivator, with Glu/Asp-rich carboxy-terminal do- Hs.40403 3.6 main, 1 *EGR2: Early growth response 2 (Krox-20 homolog, Drosophila) Hs.1395 8.9 MAFF: V-maf musculoaponeurotic ﬁbrosarcoma oncogene homolog F (avian) Hs.51305 4.2 KLF4: Kruppel-like factor 4 (gut) Hs.356370 5.0 NKX3-1: NK3 transcription factor related, locus 1 (Drosophila) Hs.55999 3.2 KLF2: Kruppel-like factor 2 (lung) Hs.107740 3.0 *EGR1: Early growth response 1 Hs.326035 4.2 FOSB: FBJ murine osteosarcoma viral oncogene homolog B Hs.75678 4.4 MYF6: Myogenic factor 6 (herculin) Hs.35937 3.1 RARRES 3: Retinoid receptor responder Hs.17466 3.9 *TR3 orphan receptor (NR4A1) Hs.1119 2.5 *ATF3: Activating transcription factor 3 Hs.460 2.3 EGR3: Early growth response 3 Hs.74088 3.1

Oncogene Gene expression in thyroid autonomous adenomas S Wattel et al 6909 Table 2 (continued ) GO at level 3 andbiological Genes (pool: ratio>3 andint>25% or confirmed;separatedsamples: ratio>2 andin most UniGene Ratio process samples; Colorflip verified)

*COPEB: Proto-oncogene Bcd orf1 and orf2 Hs.285313 2.8 Separatedsamples: *TR3 orphan receptor (NR4A1) Hs.1119 2.4 JUN-B protein Hs.25292 2.1 ATF3: Activating transcription factor 3 Hs.460 4.5 *COPEB: Proto-oncogene Bcd orf1 and orf2 Hs.285313 6.2 Cellular physiological Pool: process SCNN1B: Sodium channel, nonvoltage-gated 1, beta (Liddle syndrome) Hs.37129 3.9 IGLL1: Immunoglobulin lambda-like polypeptide 1 Hs.348935 25.6 IGHG1: Immunoglobulin heavy constant gamma 1 (G1m marker) Hs.525641 26.9 PROM1: Prominin 1 (PROML1) Hs.112360 3.1 DSCR1: Down’s syndrome critical region gene 1 Hs.184222 3.3 TRB@: T-cell receptor beta locus Hs.303157 3.1 *NELL2: NEL-like 2 (chicken) Hs.79389 3.7 *CAV1: Caveolin 1, caveolae protein, 22 kDa Hs.74034 4.7 Separatedsamples: Ig rearranged light-chain mRNA V region 4.3 Immunoglobulin kappa chain (VNJ) Hs.525893 4.0 Immunoglobulin heavy chain, V region Hs.406489 2.0 Immunoglobulin lambda heavy chain Hs.413826 1.9 *CAV1: Caveolin 1, caveolae protein, 22 kDa Hs.74034 2.2 CAV2: Caveolin 2, caveolae protein, 20 kDa Hs.139851 1.9 Cellular physiological Pool: process HBB: Hemoglobin, beta Hs.155376 3.4 HBA2: Hemoglobin, alpha 2 Hs.347939 3.6 Varia Pool: OGN: Osteoglycin (osteoinductive factor, mimecan) Hs.109439 8.8 PODLX2: Endoglycan Hs.352420 38.3 ZNF185: Zinc-ﬁnger protein 185 (LIM domain) Hs.16622 3.2 FLJ10178: Hypothetical protein FLJ10178 Hs.274267 3.0 C1S: Complement component 1, s subcomponent Hs.169756 7.6 C3: Complement component 3 Hs.284394 6.1 C7: Complement component 7 Hs.78065 6.0 LOC51134 : NY-REN-58 antigen Hs.56148 5.6 KIAA1324 : KIAA1324 protein, maba1 Hs.104696 5.3 ETAA16: ETAA16 protein Hs.82664 4.8 PCOLCE2: Procollagen C-endopeptidase enhancer 2 Hs.8944 4.1 FGL2: Fibrinogen-like 2 Hs.351808 4.0 DKFZP564O0423 : ODZ4 Hs.5028 3.8 C6orf37: Chromosome 6 open reading frame 37 (FAM46A) Hs.36587 3.6 HMGB3: High-mobility group box 3 Hs.19114 3.4 FLJ13181: Hypothetical protein FLJ13181, (TRIM45) Hs.301526 3.4 HNOEL-iso: HNOEL-iso protein (OLFML3) Hs.9315 3.4 FLJ11151: Hypothetical protein FLJ11151 (CSTP1) Hs.14992 3.0 Separatedsamples: C1S: Complement component 1, s subcomponent Hs.169756 2.4 DFFA: DNA fragmentation factor-45 (DFF1, ICAD, DFF-45) Hs.484782 2.3 SFPQ: PTB-associated splicing factor (splicing factor proline/glutamine rich (polypyrimi- Hs.180610 2.0 dine tract-binding protein associated), PSF, POMP100)

The genes, with expression ratio (tumor/normal) >3 and intensity >25% for the pooled samples analysis, or with ratio >2 for the separated samples analysis, were classiﬁed based on their gene ontology function. Genes with * were conﬁrmed by real-time RT–PCR

sequence homology could be the same gene, is upregu- hyperactivity was first verified by evaluating NIS lated in both series. mRNA overexpression (Figure 2). We normalized expression of all the transcripts to porphobilinogen deaminase (PBGD), which was shown to have similar Validation of microarray data by real-time RT–PCR expression in thyroid adenoma and normal tissue, and To validate the microarray data, we evaluated the to be not regulated by TSH in cultured thyrocytes expression of the transcripts in some of the samples (unpublished observations). The real-time RT–PCR initially used for microarray analyses, and in 10 data were also confirmed using CNAP1 (a gene additional pairs of tumor and control adjacent tissue identified as nonregulated from the microarray results, specimens, using real-time RT–PCR. For each sample, analysis I) for normalization (not shown). Results are

Oncogene Gene expression in thyroid autonomous adenomas S Wattel et al 6910 shown in Figure 3. A moderate (factor 2) under- Discussion expression of IGF1 in the pooled samples (analysis II) was confirmed by real-time PCR on separated samples. In this study, we have analysed by two different cDNA Whereas CDH2 and TPO were similarly upregulated microarray technologies gene expression in a well- in each autonomous adenoma tested, FGFR3, S100a1 defined benign thyroid tumor, the hyperfunctioning and CKB were only upregulated in six, seven or nine out autonomous adenoma. While several groups have of 10 samples, respectively. Among the downregulated genes, egr1, egr2 and lumican were downregulated in all independent samples, as evaluated by RT–PCR (10/10), a 0.6 but the others were downregulated only in the majority of the samples (in at least in seven out of 10 samples). 0.5 There was generally a good agreement between the 0.4 microarray and the real-time RT–PCR data, when 0.3 considering the expression ratio adenoma/adjacent 0.2 0.1 tissue, as obtained from different methods (Figure 3c fold modulation and d): microarray data on pooled RNA samples 0 (analysis II) or averaged from separated samples ctgf cldn4 dusp5 gsn gadd45 trail nfkbia tp53bp lum thbs2 elk4 srf egr2 egr1 nell2 dusp1 anxa1 hsp90 clu tr3 atf3 bcd cav1 igf1 (analysis I), and real-time RT–PCR data on pooled genes RNA samples or averaged from individual samples. b 18 16 Western blotting and immunohistochemistry data confirm 14 the differential expression of N-cadherin (CDH2) and 12 caveolin 1 (cav1) 10 8 We then focused on the upregulation of CDH2,andon 6 the downregulation of cav1. Protein expression was fold modulation 4 assessed by Western blotting in five pairs of adenoma/ 2 adjacent tissues, and the results obtained confirmed the 0 fgfr3 cdh2 tpo ckb s100a1 RNA expression (Figure 4). genes The overexpression of N-cadherin and decreased expression of cav1 in follicular thyroid cells were c 0.7 average taqman subsequently confirmed by immunohistochemistry on pool taqman 0.6 tissue microarrays (Figure 5). N-cadherin was present as microarray a weak or stronger lining at the cellular border in the 0.5 hyperfunctioning autonomous adenoma, and in one case also as a brown cytoplasmic staining, whereas no 0.4 staining was observed in the control adjacent tissues (Figure 5a). For cav1, detection underlines more 0.3

intensely the endothelial cells of the capillaries in the fold modulation adenomas. However, there was a granular scattered 0.2 brown immunostaining of the follicular cells in the 0.1 adenoma contrasting with a stronger cytoplasmic staining of the follicular cells of the adjacent quiescent 0

1 1

1 tissue (Figure 5b). l

r l

5 m

u f

s a

e u

t s

p u

c d g g

t n t l

t e s e

e n d a

h c t a

b c genes

d 26 24 average taqman Figure 3 Real-time RT–PCR analyses on a selection of 23 22 pool taqman downregulated genes (a) and ﬁve upregulated genes (b) from the 20 microarray microarray data. Expression of these genes, as well as IGF1, was 18 analysed by real-time RT–PCR and depicted as standardized ratios 16 7 ( s.e.m.) derived from 10 different patients. The standard 14 deviation of the logarithms of the ratio of the 10 different patients 12 was calculated for every gene. The error was subtracted or added to 10

the mean of these logarithms and the exponential was taken to fold modulation make the graph. (c, d) Comparison of the differential gene 8 expression data determined by microarray analysis and real-time 6 RT–PCR. The RT–PCR data are either the average of the 4 expression data obtained individually from the different tissue 2 samples (‘average taqman’), or the expression data obtained on the 0 pool of ﬁve different tissues tested in microarray analysis II (‘pool fgfr3 cdh2 tpo ckb s100a1 taqman’) genes

Oncogene Gene expression in thyroid autonomous adenomas S Wattel et al 6911 J3 V4 S7 B8 B7 CDH2 135 kDa Adaptin β 99 kDa

Caveolin1 22 kDa

HN N HN NHN N HN N HN N Figure 4 Western blot analysis of cdh2, cav1 and adaptin b, used as loading control, in ﬁve pairs of adenoma/adjacent tissues. In all, 5 or 20 mg of proteins for caveolin1 or cdh2, respectively, was loaded in each lane and separated by SDS–PAGE (HN: hyperfunctioning autonomous adenoma, N: normal adjacent tissue)

a normal adenoma Several lines of evidence support the validity of our analyses (see results). First, where the two microarray analyses overlap, their results are similar, even though different samples and different arrays and protocols for CDH2 samples processing were used. Second, we confirmed several regulations already identified previously (Deleu et al., 2000; Eszlinger et al., 2001, 2004; O’Donovan et al., 2002; Finley et al., 2004; Mazzanti et al., 2004; Detours et al., 2005). Moreover, several findings are in line with data obtained in different but related systems such as thyroid cells or other cells chronically stimulated by the cAMP pathway (Yamazaki et al., 2003). Third, a significant proportion of the microarray data have been confirmed by real-time RT–PCR. Fourth, as will be further elaborated in the discussion, the pattern of our b normal adenoma findings provides a comprehensive physiological frame- work, which demonstrates for the first time, or confirms and explains, diverse and sometimes apparently un- related data. Unsupervised hierarchical clustering of the samples caveolin1 shows that autonomous adenomas are relatively homogeneous tumors, despite tissue microscopic heterogeneity reflecting the presence of follicles of different sizes and apparent activity and of fibrotic elements in the capsula or within the adenoma. This is consistent with the clonality of these lesions and their common physiopathological mechanism: the constitutive activation of the cAMP-dependent mitogenic pathway. Interestingly, the tumors which clustered the closest, that is, which are molecularly the most similar, are also Figure 5 Immunohistochemical detection of CDH2 (a) and cav1 the ones which were characterized by the highest rate of (b) on tissue arrays from thyroid adenoma samples and their iodide uptake. This hyperactivity (or higher expression normal adjacent counterparts of NIS mRNA) is considered as the best criterion in the identification of autonomous adenomas (Van Sande et al., 1988). Depending on the size of the lesion and on reported gene expression profiles in follicular adenomas the level of iodine dietary supply, these adenomas may and in follicular and papillary carcinomas (Huang et al., cause hyperthyroidism or not, explaining why some 2001; Aldred et al., 2003; Barden et al., 2003; Yano patients of this group have normal thyroid hormone et al., 2004), only one similar study has been performed levels despite hyperactivity of their tumor. so far (Eszlinger et al., 2004). Our data set provides Differences in the pattern of gene expressed in tumors considerable new information about proteins and path- and in normal tissue are often interpreted as differences ways involved in this disease, which will now be the between tumor cells and normal cells. Although this is subject of more detailed investigations. Among these valid for leukemias and other liquid tumors, in which genes, five, ANXA1, clusterin, creatine kinase B, MT1X almost pure tumor cells are compared to their normal and FcGBP, have already been described as part of a six- counterparts, it is not for solid tumors and tissues in genes expression signature allowing to discriminate which the proportion of different cell types may be very between autonomous adenomas and papillary thyroid different (Venet et al., 2001). Several major findings on carcinomas (Detours et al., 2005). gene expression reflect such differences. The most

Oncogene Gene expression in thyroid autonomous adenomas S Wattel et al 6912 striking finding is the important downregulation of autonomous adenomas are compatible with such a lymphocyte specific genes. It is well known by pathol- picture (Dralle and Bocker, 1977; Becker et al., 1997). ogists that follicular adenomas contain few lymphocytes The increased expression of N-cadherin in autonomous and this decrease has been previously reported (Aust adenomas which do not metastasize and which rarely et al., 2001). A similar reduction has been described in evolve to carcinomas is a strong argument against the the thyroids of Tg-A2aR mice (Goffard et al., 2004) and concept of N-cadherin as a marker for tumor cell in the corresponding human adrenal disease, adrenal invasion and metastasis (Rieger-Christ et al., 2004). It hyperplasia (Bourdeau et al., 2004). Decreased comple- would rather facilitate direct interaction of tumor and ment expression (C1s) extends this conclusion to endothelial cells, which would lead to the capillarization macrophages. In fact, unless corroborated otherwise, of the autonomous adenomas, but could also be a all such decreases should be considered with caution. In primary step in the penetration of cancer cells in the particular, the decreased expression of mRNAs corre- vascular lumen. Decreased expression of genes coding sponding to proteins involved in the TGFb signaling for antiangiogenic proteins (Serpins, ADAMTS1) could pathway in autonomous adenomas could perhaps be also play a role in the endothelial expansion. explained in this way (Eszlinger et al., 2004). If this Caveolin I and II are considered as tumor suppressor decrease in immune cell population reflects the in vivo genes and their decrease linked to metastasizing proper- situation rather than preferential cell loss from surgical ties (Carver and Schnitzer, 2003). Decreased caveolin samples, it might indicate that the growth of autono- expression has previously been observed in Tg-A2aR mous adenomas is not, like that of the thyroid in Graves mice (Goffard et al., 2004) and has also been shown to disease, hampered by immunological reactions. In- be a characteristic of thyroid follicular carcinomas creased expression of genes of putative inhibitors of (Aldred et al., 2003, 2004). The decreased expression immune responses such as the chemokine CCL18 and in autonomous adenomas, which are benign, noninvad- FcGBP supports this hypothesis. Decreased blood cell ing tumors, certainly shows that, as for N-cadherin markers (CD2 antigen, hemoglobin b and a2) also overexpression, underexpression of caveolins is not indicate lesser blood cells presence in the adenoma sufficient to confer a malignant phenotype and is not samples. The decreased population of lymphocytes and even a marker of malignancy. On the other hand, blood cells also possibly explains decreased expressions caveolins play a major antiproliferative role in endothe- of quite a few genes: CXCL9, IL1RL1, BF and IL20RA. lial cells, and, in Tg-A2aR mice, their decrease mostly A genuine increase in the population of endothelial bears on endothelial cells (Goffard et al., 2004). This cells is suggested by the increased expression of would be in agreement with the increased endothelial endothelial specific genes (endothelial cell-specific mole- cell presence in the adenomas. cule 1, connexin 37, von Willebrandt factor, NPR3)and Our gene expression data suggest both a relative by the decreased expression of inhibitors of angiogen- decrease of intracellular cytoskeletal proteins (myosin, esis, such as Thbs2, ADAMTS1, EDNRA and SerpinF1, gelsolin, MFAP4) and of proteins allowing adhesion to indicating a higher vascularization of the tumor. The the extracellular matrix (laminin, fibulins, spondins, increased expression of N-cadherin could be interpreted lumican, NELL2). This would entail a previously in the same way, but we have shown by immunohis- unrecognized loosening of tissue and cell structure with tochemistry that this higher expression level is also unknown consequences. found in the follicular cells themselves (Figure 5b), Our data enlighten and explain several major suggesting that the increase affects the tumor cells. It is physiological characteristics of the autonomous adeno- also found in human thyroid cancer cell lines (Husmark mas. A first important finding is the low proportion of et al., 1999; unpublished data). An increased capillary genes with modified expression in the adenomas. The network is a correlate of follicle activity in human expression of the RNAs of most proteins corresponding thyroids (Gerard et al., 2002), of constitutive activation to the basic machinery of the cell was not found to be of the cAMP cascade in the thyrocytes of Tg-A2aR mice significantly different in normal and adenomatous (Ledent et al., 1992) and of chronic stimulation of rat tissues, for example, proteins involved in protein thyroids in vivo (Wollman et al., 1978). It certainly synthesis, ribosomal proteins, etc. Signal transduction explains the clinical observation of increased blood flow proteins were regulated in both directions with more in autonomous adenomas. The mechanism of this effect genes downregulated than upregulated. The greater of thyrocytes on endothelial cells remains to be defined, number of downregulated genes has been reported but some clues appear in our data. An increased before (Eszlinger et al., 2001). It can be related to the expression of N-cadherin on thyrocytes would certainly decrease in lymphocytic population and/or to a simpli- enhance the contact between these cells and endothelial fication of signal transduction in the tumor cell. cells. Similarly, sialyltransferase1, whose expression is The low frequency of altered regulations already also increased, is proposed to increase cell adhesion to observed in autonomous adenomas (Eszlinger et al., endothelial cells (Hanasaki et al., 1995). Indeed, chronic 2001) and in Tg-A2aR mice (Goffard et al., 2004) stimulation of rat thyroids leads to a sponge-like aspect reflects the physiology of such tumors. The metabolism of the glands with follicles completely surrounded by of the adenomas is normal if stimulated. All this is capillaries and coated by a continuous sheet of consistent with the concept of a minimal deviation endothelial cells (Wollman et al., 1978, 1982). Electron benign tumor: the cells are intrinsically stimulated, but microscopies or image-directed Doppler sonography of their functional behavior is normal (Van Sande et al.,

Oncogene Gene expression in thyroid autonomous adenomas S Wattel et al 6913 1988; Weber, 2002). This has been further demonstrated some previous data obtained by Northern blotting in by proteomic methodology (2D gels), by the virtual our group (Deleu et al., 2000) and in thyroid oncocy- identity of both populations of proteins synthesized and tomas (Baris et al., 2004). This result is not surprising, phosphorylated in adenomatous and quiescent tissues considering the low proportion of cells in the cell cycle (Van Sande et al., 1988). as well as the necessary biphasic nature of the expression The known and clinically important increased func- of these genes in the G1 phase of the cycle (Reuse et al., tional activity can be related to increased expressions of 1990; Pirson et al., 1994, 1996). One candidate growth thyroid-specific functional genes, like thyroperoxidase, factor suggested to play a role in the growth of deiodinase 1, rab3, metallothionein or sialyltransferase autonomous adenomas, IGF1, is contrary to expecta- (Eszlinger et al., 2004). Similarly, we found significant tions downregulated, while its inhibitor IGFBP1 is increased expression of syntaxin 7 (Wang et al., 1997), upregulated. This rather suggests, as in the case of the and its cooperating partner SNAP25. Both proteins, as cAMP cascade, a negative feedback. well as upregulated S100a1, may play a role in the Autonomous adenomas are characterized by a low activation of the thyroglobulin endocytic pathways, proliferation rate, but also by a virtual absence of leading to thyroid hormone release. Increased thyroglo- apoptosis, as studied by the Tunel method (Deleu et al., bulin endocytosis and hormone secretion are observed 2000). This can be related to the general downregulation in slices of autonomous adenomas (van den Hove- of proapoptotic genes (JNK, clusterin, Gadd45a, Vandenbroucke et al., 1976; Deleu et al., 2000). NFKBIA, TRAIL, TP53BP2, ANXA1) also observed Conversely, ANXA1, which is reported to inhibit in the thyroids of Tg-A2aR mice and to the upregulation secretion, is downregulated. Also, as expected in of antiapoptotic genes such as BIRC5. hyperactive cells, transporters (SLC25, SLC27) are Beside these general conclusions, the data offer upregulated. Upregulation of mitochondrial proteins additional clues to pursue on the physiopathology of (CKMT2, SLC25, MRPL1, CYP27B1) with no down- the autonomous adenomas. FGF stimulates the pro- regulation can also be related to increased functional liferation of human thyroid cells (Roger and Dumont, activity. An increase in the number of mitochondria in 1984; Eggo et al., 1996). The upregulation of FGFR3 these lesions has been demonstrated by electron micro- could indicate a role of FGF in this disease. What is the scopy (Hamburger et al., 1987). No gene coding for a role of upregulated FcGBP, of downregulated serpins specific function protein is downregulated in our study. and fibulins? These findings will be confirmed at the The constitutive activation of the cAMP-dependent protein level and their biological meaning further pathway in the adenomas stands in contrast with the investigated. barely increased cAMP levels in these cells (Van Sande et al., 1988). The upregulation of phosphodiesterase Materials and methods PDE7B shown here as well as the weaker upregulations of PDE1B and PDE4D suggest a negative feedback on Tissue samples cAMP levels. An upregulation of phosphodiesterases by cAMP has been observed previously in thyroid cells in In all, 21 human thyroid samples were obtained after surgical vitro as well as in other systems (Nilson et al., 1980; resection from patients diagnosed for thyroid adenomas or multinodular goiter. Diagnosis was based on TSH serum Persani et al., 2000). Negative feedback in the adenoma levels, on the demonstration by scintigraphy of a highly cells is further provided by an upregulation of GRK5, radioactive nodule with poorly radioactive surrounding and which desensitizes the TSH receptor (Voigt et al., 2004), contralateral tissue, and by post-surgical histological analysis and by a minor downregulation of adenylylcyclase demonstrating encapsulation. Tissues were immediately dis- (Eszlinger et al., 2004). Cell heterogeneity and the great sected, placed on ice, snap-frozen in liquid nitrogen and then sensitivity of the cells to small cAMP rises (Neve and stored at À801C until processing. The protocol received Dumont, 1970; Ketelbant-Balasse et al., 1976) might approval from the ethics committees of the institutions. explain how the slightly increased averaged cAMP levels To validate the diagnosed identity of autonomous adenoma can sustain the increased activity and growth of and their contralateral quiescent tissues, in a previous work thyrocytes. (Van Sande et al., 1980), slices of both had been incubated with radioiodide and methimazole to measure the iodide trapping The increased but still very low (0.8%) number of Ki- (tissue to medium (T/M) ratio). This gave values of 0.7–1 in 67-positive cells (Deleu et al., 2000; Eszlinger et al., untreated patients, for the normal quiescent tissue, and largely 2004) correlates with the increased expression of above 1 for the adenomas. As some tissues were obtained from mitogenic genes (Ki-67, FGFR3, RbBP9), as well as the other institutions where no facility was available for these decreased expression of antimitogenic genes (Thbs2, measurements, we relied in this study on the ratio of NIS rarres3). These results provide clues about the mechan- mRNA expression in both tissues. The ratio of NIS mRNA ism involved. It is striking that different DUSP that expression was measured by real-time RT–PCR. In a series of inactivate MAP kinases (DUSP1, DUSP2, DUSP5) are Brussels samples, the correlation between the ratios of T/M downregulated, while another DUSP is upregulated and NIS mRNA expression between seven adenomas and (DUSP9). This might provide the cell with an increased quiescent tissue was Rs ¼ 0.929 (Po0.003). background ERK activity that could account for the low but increased proliferation (Vandeput et al., 2003). RNA purification The general downregulation of immediate early genes For RNA purification, frozen tissues (100–200 mg) were (fos, fosB, egr-1, egr-2, junB, NGFIB, SRF) confirms reduced to powder in liquid nitrogen, homogenized, and total

Oncogene Gene expression in thyroid autonomous adenomas S Wattel et al 6914 RNA was extracted using a Trizol reagent kit (Life Technol- Real-time RT–PCR ogies Inc.), followed by purification on Qiagen RNeasy Real-time RT–PCR was performed as described in Detours columns. Concentration was assessed spectrophotometrically. et al. (2005). Oligonucleotide sequences corresponding to the RNA integrity was verified by visualization of intact 18S and selected transcripts were designed using Primer Express 28S ribosomal RNA bands after gel migration. Typically, 15– software (Applied Biosystems, CA, USA). The nucleotide 90 mg total RNA was recovered by this procedure. sequences are available upon request. Relative quantifications of gene expression were performed with the ABI Prism 7700 Sequence Detection System (AB/Perkin-Elmer). For calculat- cDNA synthesis and labeling, microarray hybridization and ing the expression level, we determined the relative quantifica- detection tion of a target gene in comparison to a reference gene (PBGD Two different kinds of cDNA microarray slides were used. or CNAP1) based on the PCR efficiencies of each gene and the Analysis of the separated samples was performed on the Ct (threshold cycle) deviation of an unknown sample versus a Micromax Human cDNA microarray system (Perkin-Elmer control (tumor sample versus adjacent quiescent tissue Life Science, Boston, MA, USA), containing 2400 known sample). human cDNA. Tumor RNA were hybridized together with patient-matched nontumoral RNA as described (Detours ðEtargetÞDCPtargetðcontrol À sampleÞ Ratio ¼ et al., 2005). The second microarray analyses (pooled samples, ðEpbgdÞDCPpbgdðcontrol À sampleÞ analysis II) were performed on dual-channel human 5k microarrays (Microarray Facility, Flanders Interuniversity with E ¼ 10(À1/slope) (Pfaffl, 2001). All the real-time RT–PCR Institute for Biotechnology, Leuven, Belgium; http://www.mi- experiments were performed in duplicate or triplicate. croarrays.be). The human 5k slide set includes four slides onto which 18 442 ESTs representing 13 000 human genes are Gel electrophoresis and immunodetection of proteins spotted in duplicate. The protocol used differs in several ways from the previous one. First, the RNA quality was addition- Frozen tissues were reduced to powder and homogenized in ally verified with a bioanalyzer (Agilent Technologies, Palo 200–500 ml of lysis buffer composed of Laemmli buffer 2 Â , Alto, CA, USA). Second, antisense RNA amplification was 1 mg/ml aprotinin, 1 mg/ml leupeptin and 60 mg/ml pefabloc, performed before hybridization (using a modified protocol of boiled for 5 min and frozen in liquid nitrogen. The total in vitro transcription as published by Barry and Eberwine cellular proteins obtained were separated according to their (Puskas et al., 2002)). A pool of 10 mg of total RNA was molecular mass by SDS–PAGE (10% gel) and immunode- amplified (about 200 Â ), purified with the RNeasy purification tected as described (Baptist et al., 1995) with rabbit polyclonal kit (Qiagen) and labeled with Cy3 or Cy5. The cDNA probes antibodies against cav1 (dilution 1/5000) (BD Biosciences, CA, were directly labeled during the reverse transcription using USA), mouse monoclonal anti-N-cadherin antibodies (dilution 200 U Superscript II (Invitrogen) and anchored-oligodT 1/500) (Zymed Laboratories Inc., CA, USA) or mouse primers. monoclonal anti-adaptin b antibodies (dilution 1/2000) (BD Biosciences, CA, USA). Anti-mouse and anti-rabbit immunoglobulin secondary antibodies (Amersham), both coupled to Microarray scanning and data analysis horseradish peroxidase, were used for detection by enhanced chemiluminescence (Western Lightning, Perkin-Elmer Life Micromax human cDNA microarrays (analysis I) were Sciences). scanned on a GMS 418 (Affymetrix) array scanner and analysed using the ImaGene (BioDiscovery, Inc.) software and Immunohistochemistry several in-house-developed software’s. To increase the specificity of the results and bypass the inherent variability of the A total of 10 formalin-fixed, paraffin-embedded thyroid technique, each experiment was performed at least twice with adenoma samples and their normal counterparts (five adeno- dye switching and averaging of ratios. Prior to quantitative mas, five normal thyroid counterparts) were used for tissue analysis, normalization was performed using the MatArray microarray construction, as described (Kononen et al., 1998). toolbox (Venet, 2003). In short, the intensity dependence of the Each case comprised a triplet of tumor tissue in addition to a ratio was removed using a loess regression technique, as triplet of the normal counterpart. A total of 30 spots of 0.6 mm described in Yang et al. (2002). Then the dependence of the diameter were available for the study. Immunostaining was ratio on the spot position was removed using a localization- performed using the standard avidin–biotin peroxidase tech- dependent normalization. A cut-off value of 2 and statistical nique with antigen retrieval, in accordance with the studied significance (Po0.05) was used to select up- or downregulated antibody. The primary antibodies were: anti-N-cadherin, genes. We considered the median of the expression ratios dilution 1 : 500 (Zymed Laboratories Inc., San Francisco, among all the investigated samples. USA), anti-caveolin 1 (N-20) : sc-894, dilution 1 : 400 (Santa The arrays from the Microarray Facility (analysis II) were Cruz Biotechnology). scanned on a Generation III scanner (Amersham Pharmacia Diaminobenzidin was used as the chromogen and hematox- Biotech) and analysed using Array Vision (Imaging Research, ylin as the nuclear counterstain. For negative controls, the Inc., St. Catharines, Ontario, Canada). Spot intensities were primary antibody was either omitted or replaced by a suitable background corrected and filtered based on 2 s.d. above concentration of normal IgG of the same species. background. Ratios were normalized by a linear regression between log 10 ratio (Cy5/Cy3) and log 10 total intensity of Acknowledgements Cy5 Â Cy3. From duplicate spots, average ratios Cy5/Cy3 We thank Chantal Degraef for excellent technical assis- were used for further analysis. Only the spots with an average tance. This work was supported by the Ministe` re de la ratio greater than 3 in both hybridizations were retained for Politique Scientifique (PAI), the Fonds de la Recherche further analysis. Some potentially interesting genes with a Scientifique Me´dicale, Te´le´vie, Fe´de´ration Belge contre le smaller average ratio were also retained because their Cancer, Fondation Wiener-Anspach, Fondation Van Buuren, differential expression was confirmed by RT–PCR. Fortis Banque Assurances and UCB-Re´gion Wallonne.

Oncogene Gene expression in thyroid autonomous adenomas S Wattel et al 6915 Hortensia Mircescu was supported by a fellowship grant by a grant from the Philippe Wiener-Maurice Anspach from the Fonds de la recherche en sante´du Que´bec and Foundation.

References

Aldred MA, Ginn-Pease ME, Morrison CD, Popkie Goffard JC, Jin L, Mircescu H, Van Hummelen P, Ledent C, AP, Gimm O, Hoang-Vu C, Krause U, Dralle H, Dumont JE and Corvilain B. (2004). Mol. Endocrinol., 18, Jhiang SM, Plass C and Eng C. (2003). Cancer Res., 63, 194–213. 2864–2871. Hamburger JI. (1987). Endocr. Rev., 8, 439–447. Aldred MA, Huang Y, Liyanarachchi S, Pellegata NS, Gimm Hanasaki K, Varki A and Powell LD. (1995). J. Biol. Chem., O, Jhiang S, Davuluri RV, de la CA and Eng C. (2004). 270, 7533–7542. J. Clin. Oncol., 22, 3531–3539. Haugen DR, Fluge O, Reigstad LJ, Varhaug JE and Lillehaug Aust G, Steinert M, Kiessling S, Kamprad M and Simchen C. JR. (2003). Thyroid, 13, 613–620. (2001). J. Clin. Endocrinol. Metab., 86, 3368–3376. Huang Y, Prasad M, Lemon WJ, Hampel H, Wright FA, Baptist M, Dumont JE and Roger PP. (1995). Exp. Cell Res., Kornacker K, LiVolsi V, Frankel W, Kloos RT, Eng C, 221, 160–171. Pellegata NS and de la CA. (2001). Proc. Natl. Acad. Sci. Barden CB, Shister KW, Zhu B, Guiter G, Greenblatt DY, USA, 98, 15044–15049. Zeiger MA and Fahey III TJ. (2003). Clin. Cancer Res., 9, Husmark J, Heldin NE and Nilsson M. (1999). Int. J. Cancer, 1792–1800. 83, 692–699. Baris O, Savagner F, Nasser V, Loriod B, Granjeaud S, Karsten SL, Van Deerlin VM, Sabatti C, Gill LH and Guyetant S, Franc B, Rodien P, Rohmer V, Bertucci F, Geschwind DH. (2002). Nucleic Acids Res., 30, E4. Birnbaum D, Malthiery Y, Reynier P and Houlgatte R. Ketelbant-Balasse P, Van Sande J, Neve P and Dumont JE. (2004). J. Clin. Endocrinol. Metab., 89, 994–1005. (1976). Horm. Metab. Res., 8, 212–215. Bauch K. (1998). Exp. Clin. Endocrinol. Diabetes, 106 (Suppl Knudsen N, Bulow I, Jorgensen T, Laurberg P, Ovesen L and 4), S16–S22. Perrild H. (2000). Eur. J. Endocrinol., 143, 485–491. Becker D, Bair HJ, Becker W, Gunter E, Lohner W, Lerch S Kononen J, Bubendorf L, Kallioniemi A, Barlund M, Schraml and Hahn EG. (1997). J. Clin. Ultrasound, 25, 63–69. P, Leighton S, Torhorst J, Mihatsch MJ, Sauter G and Bourdeau I, Antonini SR, Lacroix A, Kirschner LS, Matya- Kallioniemi OP. (1998). Nat. Med., 4, 844–847. khina L, Lorang D, Libutti SK and Stratakis CA. (2004). Krohn K, Fuhrer D, Holzapfel HP and Paschke R. (1998). Oncogene, 23, 1575–1585. J. Clin. Endocrinol. Metab., 83, 130–134. Carver LA and Schnitzer JE. (2003). Nat. Rev. Cancer, 3, Krohn K, Wohlgemuth S, Gerber H and Paschke R. (2000). 571–581. J. Pathol., 192, 37–42. Cho-Chung YS. (2004). Biochim. Biophys. Acta, 1697, Lania A, Persani L, Ballare E, Mantovani S, Losa M 71–79. and Spada A. (1998). J. Clin. Endocrinol. Metab., 83, Corvilain B, Van Sande J, Dumont JE and Vassart G. (2001). 1624–1628. Clin. Endocrinol. (Oxford), 55, 143–158. Ledent C, Dumont JE, Vassart G and Parmentier M. (1992). Deleu S, Allory Y, Radulescu A, Pirson I, Carrasco N, EMBO J., 11, 537–542. Corvilain B, Salmon I, Franc B, Dumont JE, Van Sande J Mazzanti C, Zeiger MA, Costouros NG, Umbricht C, and Maenhaut C. (2000). Thyroid, 10, 131–140. Westra WH, Smith D, Somervell H, Bevilacqua G, Detours V, Wattel S, Venet D, Hutsebaut N, Bogdanova T, Alexander HR and Libutti SK. (2004). Cancer Res., 64, Tronko MD, Dumont JE, Franc B, Thomas G and 2898–2903. Maenhaut C. (2005). Br. J. Cancer, 92, 1545–1552. Michiels FM, Caillou B, Talbot M, Dessarps-Freichey F, Dralle H and Bocker W. (1977). Virchows Arch. A Pathol. Maunoury MT, Schlumberger M, Mercken L, Monier R Anat. Histol., 374, 285–301. and Feunteun J. (1994). Proc. Natl. Acad. Sci. USA, 91, Dremier S, Coulonval K, Perpete S, Vandeput F, Fortemaison 10488–10492. N, Van Keymeulen A, Deleu S, Ledent C, Clement S, Namba H, Matsuo K and Fagin JA. (1990). J. Clin. Invest., 86, Schurmans S, Dumont JE, Lamy F, Roger PP and 120–125. Maenhaut C. (2002). Ann. NY. Acad. Sci., 968, 106–121. Neve P and Dumont JE. (1970). Exp. Cell Res., 63, Dumont JE, Jauniaux JC and Roger PP. (1989). Trends 285–292. Biochem. Sci., 14, 67–71. Nilson JH, Orlicky DJ and Kerkof PR. (1980). Endocrinology, Eggo MC, King WJ, Black EG and Sheppard MC. (1996). 107, 262–269. J. Clin. Endocrinol. Metab., 81, 3056–3062. O’Donovan N, Fischer A, Abdo EM, Simon F, Peter HJ, Ermans AM and Camus M. (1972). Acta Endocrinol. Gerber H, Buergi U and Marti U. (2002). J. Endocrinol., (Copenhagen), 70, 463–475. 174, 517–524. Eszlinger M, Krohn K, Frenzel R, Kropf S, Tonjes A and Persani L, Lania A, Alberti L, Romoli R, Mantovani G, Filetti Paschke R. (2004). Oncogene, 23, 795–804. S, Spada A and Conti M. (2000). J. Clin. Endocrinol. Metab., Eszlinger M, Krohn K and Paschke R. (2001). J. Clin. 85, 2872–2878. Endocrinol. Metab., 86, 4834–4842. Pfafﬂ MW. (2001). Nucleic Acids Res., 29, e45. Finley DJ, Arora N, Zhu B, Gallagher L and Fahey III TJ. Pirson I, Coulonval K, Lamy F and Dumont JE. (1996). (2004). J. Clin. Endocrinol. Metab., 89, 3214–3223. J. Cell. Physiol., 168, 59–70. Fuhrer D, Holzapfel HP, Wonerow P, Scherbaum WA Pirson I, Reuse S and Dumont JE. (1994). Exp. Cell Res., 210, and Paschke R. (1997). J. Clin. Endocrinol. Metab., 82, 33–38. 3885–3891. Puskas LG, Zvara A, Hackler Jr L and Van Hummelen P. Gerard AC, Many MC, Daumerie C, Costagliola S, Miot F, (2002). Biotechniques, 32, 1330–1334, 1336, 1338, 1340. DeVijlder JJ, Colin IM and Denef JF. (2002). J. Clin. Reuse S, Maenhaut C and Dumont JE. (1990). Exp. Cell Res., Endocrinol. Metab., 87, 1291–1299. 189, 33–40.

Oncogene Gene expression in thyroid autonomous adenomas S Wattel et al 6916 Rieger-Christ KM, Lee P, Zagha R, Kosakowski M, Van Sande J, Parma J, Tonacchera M, Swillens S, Dumont J Moinzadeh A, Stoffel J, Ben Ze’ev A, Libertino JA and and Vassart G. (1995). J. Clin. Endocrinol. Metab., 80, Summerhayes IC. (2004). Oncogene, 23, 4745–4753. 2577–2585. Roger PP and Dumont JE. (1984). Mol. Cell. Endocrinol., 36, Vandeput F, Perpete S, Coulonval K, Lamy F and Dumont 79–93. JE. (2003). Endocrinology, 144, 1341–1349. Russo D, Arturi F, Wicker R, Chazenbalk GD, Schlumberger Vanvooren V, Uchino S, Duprez L, Costa MJ, Vandekerc- M, DuVillard JA, Caillou B, Monier R, Rapoport B khove J, Parma J, Vassart G, Dumont JE, Van Sande J and and Filetti S. (1995). J. Clin. Endocrinol. Metab., 80, Noguchi S. (2002). Eur. J. Endocrinol., 147, 287–291. 1347–1351. Venet D. (2003). Bioinformatics, 19, 659–660. Savonet V, Maenhaut C, Miot F and Pirson I. (1997). Anal. Venet D, Pecasse F, Maenhaut C and Bersini H. (2001). Biochem., 247, 165–167. Bioinformatics, 17 (Suppl 1), S279–S287. Taniguchi M, Miura K, Iwao H and Yamanaka S. (2001). Voigt C, Holzapfel HP, Meyer S and Paschke R. (2004). Genomics, 71, 34–39. J. Endocrinol., 182, 173–182. Tonacchera M, Agretti P, Chiovato L, Rosellini V, Ceccarini Wang H, Frelin L and Pevsner J. (1997). Gene, 199, 39–48. G, Perri A, Viacava P, Naccarato AG, Miccoli P, Pinchera Weber G. (2002). Adv. Enzyme Regul., 42, 83–100. A and Vitti P. (2000). J. Clin. Endocrinol. Metab., 85, Wollman SH, Herveg JP and Smeds S. (1982). Endocrinology, 2270–2274. 111, 1867–1873. Tonacchera M, Vitti P, Agretti P, Ceccarini G, Perri A, Wollman SH, Herveg JP, Zeligs JD and Ericson LE. (1978). Cavaliere R, Mazzi B, Naccarato AG, Viacava P, Miccoli P, Endocrinology, 103, 2306–2314. Pinchera A and Chiovato L. (1999). J. Clin. Endocrinol. Yamazaki K, Yamada E, Kanaji Y, Yanagisawa T, Kato Y, Metab., 84, 4155–4158. Takano K, Obara T and Sato K. (2003). Thyroid, 13, van den Hove-Vandenbroucke MF, De Visscher M and 149–158. Couvreur-Eppe M. (1976). J. Clin. Endocrinol. Metab., 43, Yang YH, Dudoit S, Luu P, Lin DM, Peng V, Ngai J and 178–181. Speed TP. (2002). Nucleic Acids Res., 30, e15. Van Sande J, Lamy F, LecocqR, Mirkine N, Rocmans P, Yano Y, Uematsu N, Yashiro T, Hara H, Ueno E, Miwa M, Cochaux P, Mockel J and Dumont JE. (1988). J. Clin. Tsujimoto G, Aiyoshi Y and Uchida K. (2004). Clin. Cancer Endocrinol. Metab., 66, 570–579. Res., 10, 2035–2043. Van Sande J, Mockel J, Boeynaems JM, Dor P, Andry G Zeiger MA, Saji M, Gusev Y, Westra WH, Takiyama Y, and Dumont JE. (1980). J. Clin. Endocrinol. Metab., 50, Dooley WC, Kohn LD and Levine MA. (1997). Endocrino- 776–785. logy, 138, 3133–3140.

Oncogene