This Article Abstract Full Text (PDF) Alert me when this article is cited Alert me if a correction is posted Services Similar articles in this journal Similar articles in PubMed Alert me to new issues of the journal Download to citation manager Citing Articles Citing Articles via HighWire Citing Articles via Google Scholar Google Scholar Articles by Chevillard, S. Articles by El-Naggar, A. K. Search for Related Content PubMed PubMed Citation Articles by Chevillard, S. Articles by El-Naggar, A. K.

Gene Expression Profiling of Differentiated Thyroid Neoplasms

Diagnostic and Clinical Implications

¹ Laboratoire de Cancérologie Expérimentale, Commissariat á L’Energie Atomique, Direction des Sciences du Vivant, Département du Radiobiologie et Radiopathologie, Fontenay-aux-Roses Cedex France; ² Département de Pathologie, Institut Gustave Roussy, Villejuif Cedex, France; Departments of ³ Head and Neck Surgery and ⁴ Pathology, The University of Texas M.D. Anderson Cancer Center, Houston, Texas

	ABSTRACT

Top ABSTRACT INTRODUCTION MATERIALS AND METHODS RESULTS DISCUSSION REFERENCES

 
Purpose: The purpose of this research was to identify novel genes that can be targeted as diagnostic and clinical markers of differentiated thyroid tumors.

Experimental Design: Gene expression analysis using microarray platform was performed on 6 pathologically normal thyroid samples and 12 primary follicular and papillary thyroid neoplasms. Microarrays containing probes for 5,760 human full-length cDNAs were used for hybridization with total RNA from normal and tumor thyroid samples labeled with Cy3-dUTP and Cy5-dUTP, respectively. Scanned array images were recorded, and data analysis was performed. Selected sets of differentially expressed genes were analyzed using quantitative real-time reverse transcription-PCR for verification.

Results: We identified 155 genes that differentiate histologically normal thyroid tissues from benign and malignant thyroid neoplasms. Of these 75 genes were differentiated between follicular neoplasms (adenoma and carcinoma) and the follicular variant of papillary carcinoma. Purely follicular neoplasms (adenomas and carcinomas) shared many genetic profiles, and only 43 genes were distinctly different between these tumors. Hierarchical cluster analysis also differentiated conventional papillary carcinoma from its follicular variant and follicular tumors. The differentially expressed genes were composed of members of cell differentiation, adhesion, immune response, and proliferation associated pathways. Quantitative real-time reverse transcription-PCR analysis of selected genes corroborated the microarray expression results.

Conclusions: Our study show the following: (1) differences in gene expression between tumor and nontumor bearing normal thyroid tissue can be identified, (2) a set of genes differentiate follicular neoplasm from follicular variant of papillary carcinoma, (3) follicular adenoma and carcinoma share many of the differentiated genes, and (4) gene expression differences identify conventional papillary carcinoma from the follicular variant.

	INTRODUCTION

Top ABSTRACT INTRODUCTION MATERIALS AND METHODS RESULTS DISCUSSION REFERENCES

 
Differentiated thyroid epithelial tumors represent a spectrum of morphologically and biologically diverse lesions. The follicular-derived neoplasms (adenoma, carcinoma, and the follicular variant of papillary carcinoma) manifest overlapping cytomorphologic features and not infrequently pose diagnostic and treatment difficulties (1, 2, 3, 4) . Previous molecular studies of thyroid tumors have been limited to individual or multiple targeted markers and have failed to define any diagnostic or prognostic markers (5, 6, 7, 8, 9, 10, 11, 12, 13, 14) . Novel approaches are needed to identify reliable markers for pathological classification and to predict disease progression.

Parallel analysis of gene expression by microarray techniques offers a large-scale platform for screening tumors for novel markers for potential clinical applications. Using this approach, the biological and the molecular characteristics of several neoplastic entities, including a few thyroid neoplasms, have been defined (15, 16, 17) . To characterize the genetic events underlying the morphologic and biological heterogeneity of differentiated thyroid neoplasms, we performed gene expression analysis of histologically normal thyroid tissues from tumor- and nontumor-bearing resections and benign and malignant follicular and papillary neoplasms.

	MATERIALS AND METHODS

Top ABSTRACT INTRODUCTION MATERIALS AND METHODS RESULTS DISCUSSION REFERENCES

 
Tumor Samples.
The samples for this study were taken from primary thyroid neoplasms and 6 nontumorous thyroid tissues (4 with matched tumor specimens) accessioned in the Department of Pathology at The University of Texas M.D. Anderson Cancer Center from 1996 to 2000. The study protocol was approved by the Institutional Review Board committee. The tumors included 4 follicular adenoma, 3 follicular carcinoma, 3 follicular variants of papillary carcinomas, and 2 conventional papillary carcinomas. The tissues were harvested immediately on arrival at the pathology suite, placed in liquid nitrogen, and stored at –80°C until used. Histopathologic diagnosis was performed according to the WHO guidelines.

cDNA Collection and Probes.
Our platform consisted of 5,760 full-length cDNA clones from the Soares human infant brain 1NIB library (18) kindly provided by Genethon (Evry, France). Plasmids containing clones were grown in Escherichia coli by standard microbiologic methods. Each insert was (PCR) amplified (30 cycles) with vector primers derived by sampling 1 µL of cell culture. PCR products were purified by EtOH precipitation, washed in 70% EtOH, dried, dissolved in 1 mmol/L EDTA/10 mmol/L Tris-HCl (pH 8; TE)/DMSO at 50:50 concentration and stored at –80°C. The quality, size, and concentration of PCR products were determined after agarose gel electrophoresis using Genetools software (Syngene; Merck Eurolab, Fontenay-sous-Bois, France) and were automatically annotated in the final differential expression data file.

Arraying.
PCR products were arrayed on poly-L-lysine-coated slides (Menzel Glaser: CML, Nemours, France) using the Microgrid II pro arrayer (Biorobotics Ltd., Cambridgeshire, United Kingdom). Slides were packed and stored in a dark, dry place at room temperature until use.

Array Hybridizations.
Before hybridization, the slides were hydrated over boiled water, dried for 3 seconds at 80°C, and exposed to UV irradiation (300 mJ, 254 nm) for DNA cross-linking. Slides were immersed in a freshly prepared blocking solution consisting of succinic anhydride (0.02 mmol/L) dissolved in 150 mL of 1-methyl-2-pyrrolidinone and 17 mL of sodium borate (0.2 mol/L; pH 8) and placed in an orbital shaker for 20 minutes. Slides were washed in H₂O and were immersed in 100% EtOH at –20°C before a final 5-minute centrifugation at 500 rpm at room temperature. Slides were then prehybridized at 42°C in a 3x SSC, 0.1% SDS, 0.1% bovine albumin for 30 minutes at 42°C, washed in H₂O, and immersed and in isopropanol and absolute EtOH, successively. The slides were then centrifuged at 500 rpm for 5 minutes at room temperature.

RNA, cDNA, and Labeling.
Total RNA was prepared using RNAplus according to the manufacturer’s protocol (Q-Biogene, Illkirsch, France). For each competitive hybridization, total RNA from a clinical sample and the normal thyroid reference from Clontech (Palo Alto, CA) were labeled with Cy3-dUTP and Cy5-dUTP (Amersham Pharmacia Biotech, Saclay, France) by reverse transcription. For each reverse transcription reaction, 20 µg of RNA was mixed with random hexamer (pdN6, Amersham Pharmacia Biotech; 2.5 µg/mL) in a total volume of 15.5 µL, heated at 65°C for 10 minutes, and placed on ice for at least 5 minutes. Unlabeled nucleotide pool (final concentration 500 µmol/L each dATP, dCTP, dGTP, and 200 µmol/L dTTP), either Cy3 or Cy5 conjugated dUTP (final concentration 66 µmol/L; NEN, Saclay, France), 1x first-strand Superscript II buffer, 10 mmol/L DTT, and reverse transcription (400 U; Superscript II; Life Technologies, Inc., Cergy-Pontoise, France) were added to a final volume of 30 µL. After incubation at 42°C for 2 hours in a dark room, RNA was hydrolyzed by adding EDTA (45 mmol/L) and NaOH (180 mmol/L) and incubated at 65°C for 5 minutes. The mixture was neutralized by Tris-HCl (450 mmol/L; pH 7.5) and by adjusting the volume to 500 µL with Tris-EDTA. Labeled cDNA was purified by centrifugation in a Microcon YM-30 (Amicon; Millipore, Bedford, MA) and eluted twice with 40 µL of TE.

Hybridization.
The Cy3 and Cy5-labeled cDNAs were mixed with 20 µg human Cot-1 DNA, 20 µg yeast tRNA (Life Technologies, Inc.) and 10 µg poly(A; Sigma-Aldrich, Saint-Quentin Fallavier, France), then precipitated with 0.5 volumes of 7.5 mol/L sodium acetate (pH 5.2) in EtOH. The pellet was dissolved in 60 µL of hybridization solution (50% formamide, 2.5x Denhardt’s solution, 0.5% SDS, and 6x SSPE. The probe was heated for 2 minutes at 100°C, incubated at 37°C for 20 to 30 minutes, and placed between slide and coverslip. The arrays were incubated overnight at 42°C in a custom humidified slide chamber. The slides were then washed in a first bath of 0.1x SSC with 0.1% SDS for 5 minutes and then twice in 0.1x SSC for 5 minutes at room temperature. The arrays were dried by centrifugation at 600 rpm for 10 minutes. As recommended, each hybridization was performed twice with reverse labeling (dye-swap; refs. 19, 20, 21 ).

Image Analysis and Array Quantification.
Arrays were scanned using a two-color laser confocal scanner (GMS 418, Genetic Microsystems, Woburn, MA). Independent images were acquired for Cy3 (532 nm) and Cy5 (635 nm). Each spot was defined by automatic positioning of a grid using image-analysis spot-tracking software (patent US 10/173,672 June, 19, 2002; CA 2,389,901 June, 20, 2002). Each signal intensity was integrated >15 to 20 µm square pixels and recorded in 16-bit format. For each spot and fluorochrome, the background and the signal pixels were segmented with a specific Expectation-Maximization algorithm, and finally the net signal intensity of a spot was obtained by subtracting calculated local background intensity from the signal intensity. The dye-swap hybridization allows us to estimate the reproducibility between the two independent measures obtained for either the reference thyroid or the clinical sample (21) . A reliability factor was calculated to compare the two measurements of a given spot and to score the reproducibility of signal intensities obtained for ^Cy3C and ^Cy5C. The reliability factor (i) factor of the i^th spot is equal to the ratio of the two relative intensities measured for the i^th spot (or inverse ratio if >1).

For additional intermicroarray comparisons and for calculating the expression ratio "clinical sample/thyroid reference," the series of paired ^Cy3C and ^Cy5C measures (each being already scored with its reliability factor) were corrected on the line of slope 1, and normalization was performed according to a set of constantly expressed genes (22) . A cutoff of 0.6 for reliability factor was used, considering that the two measures are reproducible.

The discrimination between two categories of samples, for example, thyroid adenomas and normal thyroid samples, is performed in two steps. The first step involves finding a set of genes that are differentially expressed within each category. The second step involves clusterizing samples according to this set of genes. The expression data are organized in a matrix N.P, where the N lines represent the genes and the P columns denote the samples. Each line of the matrix (vector line) is centered according to the following formula:

and normalized according to:

where i and j represent respectively the indexes of lines and columns of the N.P matrix. And x the expression value at the (i,j) position of the matrix. The centering allowed forthe analysis of a given gene, as a function of the mean of expressionof this gene, in all samples of the analyzed categories. This permits the determinations of each gene expression, sample per sample, in relation to the average of expression. The normalization step minimizes the differences in the fold expression level between samples and genes. Finally, we are able to compare each vector line according to the relative value of the difference in the level of expression without dealing with the level of expression itself.

Gene Clustering.
To identify these genes that differentiate between two categories, we constructed two templates that represent the two theoretical vector lines. The first vector line has a value of 1 in all of the samples belonging to the first category and a value of –1 in samples belonging to the second category. The second vector line is the inverse of the first one with values of –1 in samples belonging to the first category and 1 for those belonging to the second category.

After centering and normalization of the data, the two templates and the experimental vector lines were clustered using locally developed software (ClusterIt, Paris, France)

All of the initiator vectors located at a weighted Euclidian distance inferior to a threshold s were grouped in the same cluster. If a vector also belonged to another cluster, there are two possible solutions. First, if this common vector is the initiator vector of one of the two clusters of interest then for each cluster, the mean of the vectors exclusively belonging to it is calculated. If the weighted Euclidian distance between the two means is inferior to the threshold S, the two clusters are grouped. If the distance is not inferior to the threshold, the two clusters are maintained as different clusters, and the common vectors are included in the nearest cluster (except if it is the initiator cluster). Secondly, if the common vector corresponds to a noninitiator vector, the vector is directly included in the closest cluster.

The threshold s is steadily decreased (s/100), and with each reduction, the same procedure is applied on each new cluster derived from the previous reduction. The initial threshold s is chosen such that

corresponds to the highest distance existing between two centered and normalized vectors, where p is the number of columns of the matrix. The weighted Euclidian distance is defined as followed:

where

represents a vector line of the matrix,

the expression of the gene i in condition j (centered and normalized value), and

a weighting factor between 0 and 1 proportional to the reproducibility of each expression value (dye-swap experiments). The values of template vector lines are arbitrarily fixed at a very high level to force the experimental vector lines to clusterize with the templates.

The final step of the cluster analysis corresponds to s = 0, then each cluster is composed of a unique vector line. By following step by step the evolution of the cluster composition, it is possible to determine the lineage of each cluster and to define a hierarchical tree as a function of the cluster composition. In that context, attention was given to the two clusters, including the two template vector lines. To define at which s value (i.e., at what step), the composition of the cluster is homogeneous (see below), we developed two parameters Cd and RSC that permit the evaluation of cluster stability of a given:

where:

represents the real or the theoretical densities of probability of the distribution of the n vectors within the considered class,

the SE of the real distribution of the vector in the considered class,

the covariance computed between the real and the theoretical distributions of the vectors within the considered class, and

represents the average of the weighted values of a vector line of the N.P matrix.

The definition of Cd and RSC factors are based on the following hypothesis: using DNA microarray, the error in the measure follows a normal distribution. Thus, as soon as a class of genes is homogeneous, a normal distribution of the vectors around the mean of all vectors within the class is expected. At the first step the clustering, defined with a threshold

, all of the gene vectors are grouped within the same cluster, and the mean vector of the entire cluster was null. The measured Cd could be >0.8 if the number of different vectors within the cluster is large enough to ensure, by hazard, a normal distribution. The observed distribution resulted in a mixture of different classes of genes. Along with the steps of clustering and as s decreased, affiliated clusters segment and the number of vectors per cluster decreased, thus minimizing the probability of normal distribution due to hazard. At this point, the Cd coefficient becomes small (<0.8), and the vectors do not follow a normal distribution, because the number of mixed classes is not large enough to ensure a normal distribution due to the hazard. As soon as s is small enough to ensure that vectors belong to the same cluster and are homogeneous, Cd increases >0.8, the cluster is considered stable, and the vectors follow a normal distribution. At this step, the clustering could be improved by searching for the minimal RSC and checking vector by vector for the loss or gain of cluster stability.

Specimen Clustering.
Gene clusters, from the template vector line and the experimental vector lines, represent the set of genes that discriminate between two categories. For tumor classification, a new matrix excluding the previous template vector lines and adding two sample vector columns representing each of the two experimental categories were developed. The lines for the new columns were coded 1 for the first cluster of genes and –1 for the second cluster of genes. These two additional columns represent an experimental sample template for each category, whereas other columns correspond to the gene expression of the selected genes in each experimental sample. The vector lines of the transposed matrix organized around the vector template allowed for the construction of the hierarchical sample tree. For calculating the mean value of gene expression within each category, a matrix N'.Q is defined for a given category, where the N' lines represent the genes and the Q columns the different samples within the chosen category. For each gene, the weighted mean of expression was calculated according to the following equation:

where i and j represent, respectively, the indexes of lines and columns of the N'.Q matrix. And x the expression value at the (i,j) position of the matrix.

The mean reproducibility factor is calculated according to the following equation:

with a sign "+" if

>0 and "–" if

< 0, where w is a weighting value between 0 and 1 proportional to the reproducibility of the expression value x.

We also investigated whether the follicular variant of papillary carcinoma can be distinguished from conventional papillary carcinoma by constructing 2^p vectors (p being the number of tumors) with the coordinates randomly equal to 1 or –1. The se vectors were organized in amatrix N.P, where the N lines represent the genes and the P columns the tumors. The five papillary tumors were randomly organized in the matrix, where each gene represents a vector. Within the homogeneous clusters of genes, we looked for clusters presenting two distinct levels of gene expression within these groups of tumors. We identified all of the clusters that permit the discrimination of one tumor against four or two tumors against three. We then grouped clusters that identify the same tumor, and, finally, we retained the group of clusters occurring at the highest frequency. With a such procedure, we constructed a reduced matrix N'. P and transposed the matrix to classify the five papillary tumors.

Quantitative Real-Time Reverse Transcription-PCR.
For several genes, mRNA expression was analyzed by real-time reverse transcription-PCR (ABI PRISM 7700 Sequence Detector apparatus, Applied Biosystems (Courtaboeuf, France). First-strand cDNA synthesis was described previously, reactions were performed using Syber green incorporation (SYBR Green PCR Core Reagent, Perkin-Elmer), and quantification was performed using the dedicated software (Gene Amp software, Perkin-Elmer). Quantification of each gene expression was calibrated using a reference standard curve obtained by serial dilutions of PCR product prepared from a mixture of cDNAs from normal and tumorous thyroid samples handled separately but concomitantly with clinical samples. Gene expression was normalized as a function of the expression of the 18S rRNA, as described previously (23 , 24) .

	RESULTS

Top ABSTRACT INTRODUCTION MATERIALS AND METHODS RESULTS DISCUSSION REFERENCES

 
The patients and tumor characteristics are presented in Table 1 . As criteria for inclusion in the analysis, each array spot must be well measured in all of the samples. Accordingly, 4,287 spots satisfied such requirement in all of the specimens and formed the basis for the analysis. Thresholds of 2 and 0.5 were used as the mean expression ratio, for over- or underexpressed genes compared with the normal thyroid tissue, respectively. In the first phase of the analysis, the centered mean ratio of expression (tumor/reference) in different samples for each of the normal, adenoma, and follicular and papillary carcinomas was calculated for each spot.

View larger version (30K): [in this window] [in a new window]   Fig. 1. Classification of differentially expressed genes according to their biological functions in normal thyroid samples (A), follicular adenomas (B), follicular carcinomas (C), and papillary carcinomas (D; expressed sequence tags or genes with unknown functions are not included).

The gene differences between two tumors within a category were used to develop a strategy for the expression within a category independent of the expression ratio. In that strategy, a given gene can be used to discriminate between two categories without applying a strict cutoff if the expression ratio of this gene is as follows: (1) the same for all samples within a category, and (2) different between the two categories. As shown in Fig. 2, A and B , and Fig. 3, A and B , 23, 52, and 80 genes define adenoma from papillary carcinoma, follicular carcinoma from follicular variant of papillary carcinoma, and adenomas from follicular carcinomas, respectively. The differentially expressed genes classifying follicular adenomas from follicular carcinomas are listed in Table 2 . Restricting gene clustering to the five papillary carcinomas, a set of genes (Table 3) differentiated the follicular variant from conventional papillary carcinomas (Fig. 3B) .

View larger version (48K): [in this window] [in a new window]   Fig. 2. Hierarchical clustering based on differentially expressed genes that distinguished adenomas (AD) from papillary carcinomas (PC) (A) and follicular carcinoma (FC) from papillary carcinomas (B), respectively.

View larger version (32K): [in this window] [in a new window]   Fig. 3. Hierarchial clustering based on differentially expressed genes that segregate follicular carcinomas (FC) from adenomas (AD) and papillary carcinoma (PC) from follicular variant of papillary carcinomas (FV PC; B), respectively. Details of genes with known functions are given on Tables 3 and 4 .

	DISCUSSION

Top ABSTRACT INTRODUCTION MATERIALS AND METHODS RESULTS DISCUSSION REFERENCES

The data also show that adenomas can be differentiated from follicular carcinomas by a set of 43 genes of known functions. Of these, 10 were reported previously (17) to be highly expressed in follicular carcinomas (PEG10, PLCB4, FBN1, ID4, DUSP1, NID, BSG, CBS, HXB, RPP1, and TRAP240) and 3 genes (SORD, X123, MCG14797) in follicular adenomas (25 , 26) . Granulin, a putative growth factor, was noted to be significantly overexpressed in carcinomas compared with adenomas. Previous studies of glioblastoma and gastric and ovarian cancers have also shown an up-regulation of this gene (27, 28, 29) . Of the down-regulated genes, DUSP 1, which has known proliferation and differentiation functions, has been shown previously to be down-regulated in follicular carcinomas (17) and in advanced epithelial ovarian cancer (30) . We also noted the PKIA gene, a potent inhibitor of protein kinase B, to be down-regulated in follicular carcinomas compared with adenomas, supporting previous molecular studies of sporadic thyroid cancer (31) . As in earlier studies, the VEGF gene was characteristically noted in follicular carcinomas (32 , 33) .

Our study identified a set of 23 and 52 genes that differentiated papillary carcinoma from both follicular adenomas and carcinomas, respectively (Fig. 2, A and B) . These genes were composed of members of the protein metabolism/catabolism (15%), intracellular trafficking (12%), ion/metal binding and transport proteins (10%), and transcription related (12%) gene families. The Cyr61 gene was underexpressed in both papillary and follicular carcinomas (Fig. 2A) compared with adenomas, and it was shown previously to be down-regulated in papillary carcinomas (16) . Also MCM7, a DNA replication gene, was overexpressed in papillary carcinomas compared with adenomas. Although, the overall gene expressions of papillary and follicular carcinoma (Fig. 1) showed common features, we identified individual genes that differentiate between papillary and follicular carcinomas. Among these are the DUSP6 and DUSP5, members of the dual-specificity phosphatase gene family. DUSP5 has been reported previously to be up-regulated in premalignant lesions and underexpressed in invasive pancreatic carcinoma (34) . The MFGE8 gene that plays a key role in activities such as cell motility, activation, contact, and the maintenance of the membrane is also highly expressed in papillary carcinomas compared with adenomas. MFGE8 was shown previously to be highly overexpressed in human breast tumors (35) .

Gene cluster analysis additionally segregated the conventional form from the follicular variant of papillary carcinomas (Fig. 3B) . Fourteen of these genes had transcriptional functions and included 2 of the Id4 and HOXC6, down-regulated and 12 (TRIM28, CUTL1, MSL3L1, SNW1, E64F1, CGBP, JUNB, ASCL1, IFGS3G, HCNGP, HCNP, and SSX1) overexpressed. Only a few of these genes have been reported previously to be associated with thyroid gland development. The Id4 expression, a dominant-negative transcriptional regulator, is known to be mainly expressed in thyroid (26) and to induce cell proliferation and inhibit differentiation induced by thyroid hormone (36) . The NR0B2 gene, a negative regulator of receptor-dependent signaling pathways, interacts with thyroid hormone and retinoid and thyroid hormone receptors to repress nuclear hormone receptor-mediated transactivation (37) .

Interestingly, the human ACSL1/HASH1 gene, a transcription factor that is overexpressed in the follicular variant of papillary carcinoma, has also been reported to be highly expressed in neuroendocrine tumors, including medullary thyroid and small cell lung cancers (38) . The finding suggests that the follicular variant, in contrast to the conventional form, preferentially manifest neuroendocrine features. The finding of under- and overexpression of the NTRK2 and NTRK3 genes, respectively (Table 4) , and reports of medullary thyroid carcinoma support this hypothesis (39) . Similarly, the 2NF151/MIZ2I, CDK2API, and P12DOC-I genes, which are involved in cell proliferation (40 , 41) , were found to be overexpressed in follicular variants than in conventional papillary carcinomas. The PEG10 gene was distinctly overexpressed in the follicular variant compared with the conventional papillary carcinoma. This gene has been reported to be highly expressed in hepatocellular carcinomas (42) and was found to differentiate follicular thyroid tumors from normal thyroid (17) . Cyclin D1 (CCND1), a regulatory subunit of CDK4 or CDK6, has been found to be overexpressed in papillary thyroid carcinomas (43) and predicts lymph node metastases in papillary thyroid carcinoma (44) .

In addition, the BGLAP, SPARC, and MSN genes, which bind to the Osteopontin (OPN) gene, were found to be distinctly overexpressed in the follicular variant of papillary carcinoma. These genes play a prominent role in bone and calcium metabolism and the formation of psammoma bodies in meningiomas (45) and ovarian serous papillary (46) and thyroid carcinomas (47) . In addition, SPARC and MSN are also associated with tumor invasiveness and metastasis (48 , 49) . Tumor necrosis factor, a proinflammatory cytokine, TNFRSF25, a receptor expressed preferentially in tissues enriched with lymphocytes, NOTCH2, furin (PACE-1), GRB2, LNK, and HLA-DRB1 and HLA-DQA1, MHC-class II, were also overexpressed in follicular variants compared with papillary carcinomas. The expression of MHC class II was reported previously to be altered in papillary carcinomas, including the follicular variant (50) .

Our study shows that gene expression array analysis correlated generally with classification of thyroid tumors. We also identified a set of genes that are differentially expressed between follicular benign and malignant neoplasms and between these tumors and papillary carcinoma. Although interesting, the final verification of the true biological nature must await additional follow-up on these patients.

	FOOTNOTES

 
Grant support: PIC Curie-CEA, Electricité de France, European Commission contract FIS5–2002-00004 (GENRAD-T), The Kenneth Muller Professorship (A. K. El-Naggar), and the Specialized Programs of Research Excellence in Head and Neck Cancer (SPORE).

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.

Requests for reprints: Adel K. El-Naggar, The University of Texas M.D. Anderson Cancer Center, Department of Pathology, 1515 Holcombe Boulevard, Houston, TX 77030. Phone: 713-792-3109; Fax: 713-792-5532; E-mail: anaggar{at}mdanderson.org

Received 1/15/04; revised 5/11/04; accepted 6/22/04.

	REFERENCES

Top ABSTRACT INTRODUCTION MATERIALS AND METHODS RESULTS DISCUSSION REFERENCES

 

1. Fagin JA. Perspective lesions learned from molecular genetic studies of thyroid cancer—insights into pathogenesis and tumor-specific therapeutic targets. Endocrinology 2002;143:2025-8.[Free Full Text]
2. Links TP, van Tol KM, Meerman GJ, deVries EG. Differentiated thyroid carcinoma: a polygenic disease. Thryoid 2001;11:1135-40.
3. Brennan MD, Bergstralh EJ, van Heerden JA, McConahey WM. Follicular thyroid cancer treated at the Mayo Clinic 1946 through 1970: initial manifestations pathologic findings, therapy, and outcome. Mayo Clin Proc 1991;66:11-22.[Medline]
4. Wanebo HJ, Andrews W, Kaiser DL. Thyroid cancer: some basic considerations. CA Cancer J Clin 1983;33:87-97.[Abstract/Free Full Text]
5. Fukushima T, Suzuki S, Mashiko M, et al BRAF mutations in papillary carcinomas of the thyroid. Oncogene 2003;22:6455-7.[CrossRef][Medline]
6. Wiseman SM, Loree TR, Rigual NR, et al Papillary thyroid cancer: high inter-(simple sequence repeat) genomic instability in a typical indolent cancer. Head Neck 2003;25:825-32.[CrossRef][Medline]
7. Nikiforova MN, Biddinger PW, Caudill CM, Kroll TG, Nikiforov YE. PAX8-PPARgamma rearrangement in thyroid tumors: RT-PCR and immunohistochemical analyses. Am J Surg Pathol 2002;26:1016-23.[CrossRef][Medline]
8. Garcia-Rostan G, Camp RK, Herrero A, Carcangiu M, Rimm DL, Tallini G. Beta-catenin dysregulation in thyroid neoplasms down-regulation, aberrant nuclear expression, a CTNNB1 exon 3 mutations are markers for aggressive tumor phenotypes and poor prognosis. Am J Pathol 2001;158:987-96.[Abstract/Free Full Text]
9. Masood S, Auguste LJ, Westerband A, Belluco C, Valderama E, Attie J. Differential oncogenic expression in thyroid follicular and Hurthle cell carcinomas. Am J Surg 1993;166:366-8.[CrossRef][Medline]
10. Hoos A, Stojadinovic A, Singh B, et al Clinical significance of molecular expression profiles of Hurthle cell tumors of the thyroid gland analyzed via tissue microarrays. Am J Pathol 2002;160:175-83.[Abstract/Free Full Text]
11. Frisk T, Kytola S, Wallin G, Zedenius J, Larsson C. Low frequency of numerical chromosomal aberrations in follicular thyroid tumors detected by comparative genomic hybridization. Genes Chromosomes Cancer 1999;25:349-53.[CrossRef][Medline]
12. Takahashi M. Oncogenic activation of the ret protooncogene in thyroid cancer. Crit Rev Oncog 1995;6:35-46.[Medline]
13. Yeh JJ, Marsh DJ, Zedenius J, et al Fine-structure deletion mapping of 10q22–24 identifies regions of loss of heterozygosity and suggests that sporadic follicular thyroid adenoma and follicular thyroid carcinomas develop along distinct neoplastic pathways. Gene chromosomes Cancer 1999;26:322-8.
14. Rodrigues-Serpa A, Catarino A, Soares J. Loss of heterozygosity in follicular and papillary thyroid carcinomas. Cancer Genet Cytogenet 2003;141:26-31.[CrossRef][Medline]
15. Huang Y, Prasad M, Lemon WJ, et al Gene expression in papillary thyroid carcinoma reveals highly consistent profiles. Proc Natl Acad Sci USA 2001;98:15044-9.[Abstract/Free Full Text]
16. Wasenius VM, Hemmer S, Kettunen E, Knuutila S, Franssila K, Joensuu H. Hepatocyte growth factor receptor, matrix metalloproteinase-11, tissue inhibitor of metalloproteinase-1, and fibronectin are up-regulated in papillary thyroid carcinoma: a cDNA and tissue microarray study. Clin Cancer Res 2003;9:68-75.[Abstract/Free Full Text]
17. Barden CB, Shister KW, Zhu B, 3rd, et al Classification of follicular thyroid tumors by molecular signature: results of gene profiling. Clin Cancer Res 2003;9:1792-800.[Abstract/Free Full Text]
18. Soares MB, Bonaldo MF, Jelene P, Su L, Lawton L, Efstratiadis A. Construction and characterization of a normalized cDNA library. Proc Natl Acad Sci USA 1994;91:9228-32.[Abstract/Free Full Text]
19. Lee ML, Kuo FC, Whitmore GA, Sklar J. Importance of replication in microarray gene expression studies: statistical methods and evidence from repetitive cDNA hybridizations. Proc Natl Acad Sci USA 2000;97:9834-9.[Abstract/Free Full Text]
20. Schulze A, Downward J. Analysis of gene expression by microarrays: cell biologist’s gold mine or minefield?. J Cell Sci 2000;113:4151-6.[Abstract]
21. Tu Y, Stolovitzky G, Klein U. Quantitative noise analysis for gene expression microarray experiments. Proc Natl Acad Sci USA 2002;99:14031-6.[Abstract/Free Full Text]
22. Tseng GC, Oh MK, Rohlin L, Liao JC, Wong WH. Issues in cDNA microarray analysis: quality filtering, channel normalization, models of variations and assessment of gene effects. Nucleic Acids Res 2001;29:2549-57.[Abstract/Free Full Text]
23. Ory K, Lebeau J, Levalois C, et al Apoptosis inhibition mediated by medroxyprogesterone acetate treatment of breast cancer cell lines. Breast Cancer Res Treat 2001;68:187-98.[CrossRef][Medline]
24. Bishay K, Ory K, Olivier MF, Lebeau J, Levalois C, Chevillard S. DNA damage-related RNA expression to assess individual sensitivity to ionizing radiation. Carcinogenesis 2001;22:1179-83.[Abstract/Free Full Text]
25. Ishibashi T, Bottaro DP, Michieli P, Kelley CA, Aaronson SA. A novel dual specificity phosphatase induced by serum stimulation and heat shock. J Biol Chem 1994;269:29897-902.[Abstract/Free Full Text]
26. Rigolet M, Rich T, Gross-Morand MS, Molina-Gomes D, Viegas-Pequignot E, Junien C. cDNA cloning, tissue distribution and chromosomal localization of the human ID4 gene. DNA Res 1998;5:309-13.[Abstract]
27. Markert JM, Fuller CM, Gillespie GY, et al Differential gene expression profiling in human brain tumors. Physiol Genomics 2001;5:21-33.[Abstract/Free Full Text]
28. Line A, Stengrevics A, Slucka Z, Li G, Jankevics E, Rees RC. Serological identification and expression analysis of gastric cancer-associated genes. Br J Cancer 2002;86:1824-30.[CrossRef][Medline]
29. Jones MB, Michener CM, Blanchette JO, et al The granulin-epithelin precursor/PC-cell-derived growth factor is a growth factor for epithelial ovarian cancer. Clin Cancer Res 2003;9:44-51.[Abstract/Free Full Text]
30. Manzano RG, Montuenga LM, Dayton M, et al CL100 expression is down-regulated in advanced epithelial ovarian cancer and its re-expression decreases its malignant potential. Oncogene 2002;21:4435-47.[CrossRef][Medline]
31. Sandrini F, Matyakhina L, Sarlis NJ, et al Regulatory subunit type I-alpha of protein kinase A (PRKAR1A): a tumor-suppressor gene for sporadic thyroid cancer. Genes Chromosomes Cancer 2002;35:182-92.[CrossRef][Medline]
32. Huang SM, Lee JC, Wu TJ, Chow NH. Clinical relevance of vascular endothelial growth factor for thyroid neoplasms. World J Surg 2001;25:302-6.[CrossRef][Medline]
33. Lewy-Trenda I, Wierzchniewska-Lawska A. Expression of vascular endothelial growth factor (VEGF) in human thyroid tumors. Pol J Pathol 2002;53:129-32.[Medline]
34. Furukawa T, Sunamura M, Motoi F, Matsuno S, Horii A. Potential tumor suppressive pathway involving DUSP6/MKP-3 in pancreatic cancer. Am J Pathol 2003;162:1807-15.[Abstract/Free Full Text]
35. Carmon L, Bobilev-Priel I, Breener B, et al Characterization of novel breast carcinoma-associated BA46-derived peptides in HLA-A2.1/D(b)-beta2m transgenic mice. J Clin Investig 2002;110:453-62.[CrossRef][Medline]
36. Kondo T, Raff M. The Id4 HLH protein and the timing of oligodendrocyte differentiation. EMBO J 2000;19:1998-2007.[CrossRef][Medline]
37. Seol W, Hanstein B, Brown M, Moore DD. Inhibition of estrogen receptor action by the orphan receptor SHP (short heterodimer partner). Mol Endocrinol 1998;12:1551-7.[Abstract/Free Full Text]
38. Ball DW, Azzoli CG, Baylin SB, et al Identification of a human achaete-scute homolog highly expressed in neuroendocrine tumors. Proc Natl Acad Sci USA 1993;90:5648-52.[Abstract/Free Full Text]
39. Gimm O, Dziema H, Brown J, et al Mutation analysis of NTRK2 and NTRK3, encoding 2 tyrosine kinase receptors, in sporadic human medullary thyroid carcinoma reveals novel sequence variants. Int J Cancer 2001;92:70-4.[CrossRef][Medline]
40. Peukert K, Staller P, Schneider A, Carmichael G, Hanel F, Eilers M. An alternative pathway for gene regulation by Myc. EMBO J 1997;16:5672-86.[CrossRef][Medline]
41. Shintani S, Mihara M, Terakado N, et al Reduction of p12DOC-1 expression is a negative prognostic indicator in patients with surgically resected oral squamous cell carcinoma. Clin Cancer Res 2001;7:2776-82.[Abstract/Free Full Text]
42. Okabe H, Satoh S, Furukawa Y, et al Involvement of PEG10 in human hepatocellular carcinogenesis through interaction with SIAH1. Cancer Res 2003;63:3043-8.[Abstract/Free Full Text]
43. Kim JY, Cho H, Rhee BD, Kim HY. Expression of CD44 and cyclin D1 in fine needle aspiration cytology of papillary thyroid carcinoma. Acta Cytol 2002;46:679-83.[Medline]
44. Khoo ML, Beasley NJ, Ezzat S, Freeman JL, Asa SL. Overexpression of cyclin D1 and underexpression of p27 predict lymph node metastases in papillary thyroid carcinoma. J Clin Endocrinol Metab 2002;87:1814-8.[Abstract/Free Full Text]
45. Hirota S, Nakajima Y, Yoshimine T, et al Expression of bone-related protein messenger RNA in human meningiomas: possible involvement of osteopontin in development of psammoma bodies. J Neuropathol Exp Neurol 1995;54:698-703.[Medline]
46. Maki M, Hirota S, Kaneko Y, Morohoshi T. Expression of osteopontin messenger RNA by macrophages in ovarian serous papillary cystadenocarcinoma: a possible association with calcification of psammoma bodies. Pathol Int 2000;50:531-5.[CrossRef][Medline]
47. Tunio GM, Hirota S, Nomura S, Kitamura Y. Possible relation of osteopontin to development of psammoma bodies in human papillary thyroid cancer. Arch Pathol Lab Med 1998;122:1087-90.[Medline]
48. Carmeci C, Thompson DA, Kuang WW, Lightdale N, Furthmar H, Weigel RJ. Moesin expression is associated with the estrogen receptor-negative breast cancer phenotype. Surgery (St Louis) 1998;124:211-7.
49. Rempel SA, Ge S, Gutierrez JA. SPARC: a potential diagnostic marker of invasive meningiomas. Clin Cancer Res 1999;5:237-41.[Abstract/Free Full Text]
50. Lloyd RV, Johnson TL, Blaivas M, Sisson JC, Wiksibm BS. Detection of HLA-DR antigens in paraffin-embedded thyroid epithelial cells with a monoclonal antibody. Am J Pathol 1985;120:106-11.[Abstract]

This article has been cited by other articles:

				S. Durand, C. Ferraro-Peyret, S. Selmi-Ruby, C. Paulin, M. El Atifi, F. Berger, N. Berger-Dutrieux, M. Decaussin, J.-L. Peix, C. Bournaud, et al. Evaluation of Gene Expression Profiles in Thyroid Nodule Biopsy Material to Diagnose Thyroid Cancer J. Clin. Endocrinol. Metab., April 1, 2008; 93(4): 1195 - 1202. [Abstract] [Full Text] [PDF]

				R. T. Netea-Maier, S. W. Hunsucker, B. M. Hoevenaars, S. M. Helmke, P. J. Slootweg, A. R. Hermus, B. R. Haugen, and M. W. Duncan Discovery and Validation of Protein Abundance Differences between Follicular Thyroid Neoplasms Cancer Res., March 1, 2008; 68(5): 1572 - 1580. [Abstract] [Full Text] [PDF]

				J. P. Klopper, A. Berenz, W. R. Hays, V. Sharma, U. Pugazhenthi, J. Janssen, M. Singh, R. P. Bissonnette, and B. R. Haugen In vivo and Microarray Analysis of Rexinoid-Responsive Anaplastic Thyroid Carcinoma Clin. Cancer Res., January 15, 2008; 14(2): 589 - 596. [Abstract] [Full Text] [PDF]

				G. Riesco-Eizaguirre and P. Santisteban New insights in thyroid follicular cell biology and its impact in thyroid cancer therapy Endocr. Relat. Cancer, December 1, 2007; 14(4): 957 - 977. [Abstract] [Full Text] [PDF]

				K. Fujarewicz, M. Jarzab, M. Eszlinger, K. Krohn, R. Paschke, M. Oczko-Wojciechowska, M. Wiench, A. Kukulska, B. Jarzab, and A. Swierniak A multi-gene approach to differentiate papillary thyroid carcinoma from benign lesions: gene selection using support vector machines with bootstrapping Endocr. Relat. Cancer, September 1, 2007; 14(3): 809 - 826. [Abstract] [Full Text] [PDF]

				M. Eszlinger, K. Krohn, A. Kukulska, B. Jarzab, and R. Paschke Perspectives and Limitations of Microarray-Based Gene Expression Profiling of Thyroid Tumors Endocr. Rev., May 1, 2007; 28(3): 322 - 338. [Abstract] [Full Text] [PDF]

				O. L. Griffith, A. Melck, S. J.M. Jones, and S. M. Wiseman Meta-Analysis and Meta-Review of Thyroid Cancer Gene Expression Profiling Studies Identifies Important Diagnostic Biomarkers J. Clin. Oncol., November 1, 2006; 24(31): 5043 - 5051. [Abstract] [Full Text] [PDF]

				C. C. Lubitz, S. K. Ugras, J. J. Kazam, B. Zhu, T. Scognamiglio, Y.-T. Chen, and T. J. Fahey III Microarray Analysis of Thyroid Nodule Fine-Needle Aspirates Accurately Classifies Benign and Malignant Lesions J. Mol. Diagn., September 1, 2006; 8(4): 490 - 498. [Abstract] [Full Text] [PDF]

				M. Eszlinger, M. Wiench, B. Jarzab, K. Krohn, M. Beck, J. Lauter, E. Gubala, K. Fujarewicz, A. Swierniak, and R. Paschke Meta- and Reanalysis of Gene Expression Profiles of Hot and Cold Thyroid Nodules and Papillary Thyroid Carcinoma for Gene Groups J. Clin. Endocrinol. Metab., May 1, 2006; 91(5): 1934 - 1942. [Abstract] [Full Text] [PDF]

This Article Abstract Full Text (PDF) Alert me when this article is cited Alert me if a correction is posted Services Similar articles in this journal Similar articles in PubMed Alert me to new issues of the journal Download to citation manager Citing Articles Citing Articles via HighWire Citing Articles via Google Scholar Google Scholar Articles by Chevillard, S. Articles by El-Naggar, A. K. Search for Related Content PubMed PubMed Citation Articles by Chevillard, S. Articles by El-Naggar, A. K.