Document

The FASEB Journal express article 10.1096/fj.03-0101fje. Published online January 8, 2004.

Genes differentially expressed in thyroid carcinoma identified by comparison of SAGE expression profiles Erwin Pauws,* Geertruda J. M. Veenboer,* Jan W. A. Smit,� Jan J. M. de Vijlder,* Hans Morreau,� and Carrie Ris-Stalpers* *Laboratory of Pediatric Endocrinology, Academic Medical Center, Amsterdam, The Netherlands; �Department of Pathology; and �Department of Endocrinology, Leiden University Medical Center, Leiden, The Netherlands

Corresponding author: C. Ris-Stalpers, Laboratory of Pediatric Endocrinology, AMC, Room G2-136, PO Box 22700, 1100 DE Amsterdam, The Netherlands. E-mail: [email protected]

ABSTRACT

To identify transcripts that distinguish malignant from benign thyroid disease serial analysis of gene expression (SAGE) profiles of papillary thyroid carcinoma and of normal thyroid are compared. Of the 21,000 tags analyzed, 204 tags are differentially expressed with statistical significance in the tumor. Thyroid tumor specificity of these transcripts is determined in silico using the tissue preferential expression (TPE) algorithm. TPE values demonstrate that 42 tags of the 204 are thyroid tumor specific. BC013035, a cDNA encoding a novel protein, is up-regulated from 0 to 24 tags in the thyroid tumor SAGE library. In a tissue panel of 30 thyroid tumors and 12 controls, it has an expression pattern similar to thyroid peroxidase, indicating possible involvement of BC013035 in thyroid differentiation. A tag coding for extracellular matrix protein 1 (ECM1) is absent in the normal thyroid SAGE library and present 55 times in the tumor. ECM1, a protein recently associated with angiogenesis and expressed in metastatic breast carcinoma, is up-regulated in 50% of all thyroid carcinoma and absent in normal controls and follicular adenoma. In conclusion, SAGE analysis and subsequent determination of TPE values facilitates the rapid distinction of genes specifically expressed in cancer tissues.

Key words: tissue preferential expression ● serial analysis of gene expression ● extracellular matrix protein 1

espite the low incidence of thyroid carcinoma (0.5-10 cases per 100,000), thyroid nodules are relatively frequent in the population (1, 2). In the case of a cold nodule, the diagnostic aim is to exclude a thyroid carcinoma. Lack of powerful diagnostic markers

that distinguish between benign and malignant thyroid diseases causes many patients to undergo surgery. By improving diagnosis, the number of thyroid surgeries that are performed for thyroid nodules with a suspicion for thyroid carcinoma can be markedly reduced. Although the general prognosis of thyroid carcinoma is favorable, subgroups are at risk for recurrent disease, metastasis, or death. Because the identification of these groups is difficult, almost all patient will undergo intensive initial therapy, which may not be necessary in all patients when suitable risk factors are available (3). The most common thyroid cancer subtype is papillary thyroid carcinoma (PTC), which accounts for ~60% of cases. Follicular (FTC) and anaplastic (undifferentiated) carcinoma (ATC) are less frequent (4). Thyroid cancers metastasize to lymph

D

nodes (PTC) and lungs and bone (FTC, ATC; ref 1). The fact that thyroid epithelial cells can dedifferentiate to malignant tumors with entirely different histological and clinical behavior makes the molecular pathogenesis of thyroid carcinoma an event with relevance for general processes in dedifferentiation. The identification of genetic variations in thyroid tumors has increased the understanding of the molecular basis of thyroid carcinoma (5). Differentially expressed genes can be used as target for molecular-based diagnosis and therapy. In general, thyroid-specific gene expression is down-regulated in thyroid carcinoma and an anaplastic carcinoma has lost all thyroid-specific expression (6, 7). Loss of heterozygosity, gene rearrangements, and point mutations are linked to the development of thyroid cancer. Comparative genomic hybridization (CGH) demonstrated specific loss of chromosome 8 in ATC (8). Other studies identified chromosomal regions 7q and 10q as specific cytogenetic events in FTC (9, 10). Recently, FTC is associated with a PPARγ and PAX8 rearrangement (11). PTC tumors are often associated with genomic rearrangements of the ret proto-oncogene (12), especially in case of external radiation as in the Chernobyl incident (13, 14). Activating point mutations in the TSH-receptor gene and the ras and gsp oncogenes are implicated in several studies (15, 16). Mutations in mitochondrial DNA are linked to thyroid tumor progression (17). Several differentially expressed genes are described in thyroid carcinoma (for review see ref 18). Galectin-3 was reported as a candidate marker suitable to distinguish benign from malignant thyroid neoplasms (19, 20). However, recently other reports have raised important questions about the accurateness of galectin-3 as a diagnostic marker, especially on the RNA level (21, 22). To further elucidate the molecular pathogenesis behind thyroid cancer and to identify novel markers for differentiated thyroid carcinoma, we used serial analysis of gene expression (SAGE), a high-throughput technique suitable for this purpose. SAGE is a sequence-based approach to identify a cell-specific gene expression profile (23). Analysis and comparison of SAGE expression profiles can identify novel genes involved in the molecular pathology of disease (24). In molecular oncology, SAGE studies have advanced the understanding of tumor-specific changes in gene expression (25, 26). In this study, we constructed and analyzed a SAGE library from a follicular variant of PTC with widespread but indolent metastatic behavior (27). PTC is the most common thyroid carcinoma, and by analyzing this particular tumor sample we expect to identify genetic factors associated with aggressive and/or metastatic behavior. In a previous study, our laboratory analyzed the SAGE expression profile of a normal thyroid to identify novel thyroid-specific genes (28). Comparison of both expression profiles identifies differentially expressed genes in thyroid carcinoma. Subsequent in silico analysis using the TPE algorithm (29), which was refined for this specific purpose, allows the selection of candidate tumor-specific transcripts. Finally, expression of two candidate genes was studied by semiquantitative RT-PCR (sqRT-PCR) on a panel of normal and tumor thyroid tissues.

METHODS

Tumor tissue and RNA isolations

Frozen tissue was collected of 30 patients who underwent thyroidectomy for suspicion of thyroid carcinoma in the Leiden University Medical Center. Pathological classification of the samples was made as follows: 5 follicular adenomas (FA); 20 papillary carcinomas (PTC) from which 6 are follicular variants; and 5 follicular carcinoma (FTC). From eight PTC samples, tumor tissues and adjacent normal thyroid tissue were collected after microdissection. Data on vascular invasion, lymph node metastasis, and/or distant metastasis were recorded. For SAGE analysis

one tumor (Thy_T), a follicular variant of PTC was selected because of its aggressive metastatic behavior and abundant tissue availability. Cytogenetic analysis of this tumor showed a t(3;5)(q12;p15) chromosomal translocation and LOH of chromosome 22 (27). Normal thyroid tissue was obtained from four individuals without thyroid pathology after resection at routine autopsy. After homogenization of tissue samples, total RNA was extracted using TRIzol (Gibco-BRL).

SAGE library construction and analysis

SAGE libraries were made using SAGE protocol 2.0 and as described previously (23, 28). SAGE clones were sequenced with the Dyenamic Direct cycle sequencing kit (Perkin Elmer) using the T7 priming site. Samples were run on an ABI377XL Automatic Sequencer (Perkin Elmer) and analyzed with Sequence Analysis 3.0 software. Tag extraction was performed using the analysis program USAGE 2.01 (30). Comparison of SAGE libraries was done using USAGE and its intrinsic statistical package calculating P values of a comparative Z-test (31). Tag identification and expression were done using NCBI/CGAP�s SAGEmap program (32).

Tissue preferential expression analysis

Tissue preferential expression (TPE) analysis was performed as described previously (29) with some modifications. The following algorithm was used:

TPEN = √[(Ratio(tagN))2+(%Libraries)2];

for every tag with expression level NREF in the reference library the log-ratio with respect to the expression level in all other libraries Na-x is calculated.

Ratio(tagN)=∑a-x{log[(0.001+NREF(tagN))/(0.001+Na(tagN))]}.

The sum of all ratios is Ratio tagN. The percentage of libraries where a given tag is expressed is taken as the second component in the calculation of the TPE value. The resulting TPE values are normalized and range between 0-100. Generally expressed transcripts typically have TPE values <25. Tissue-specific transcripts have TPE values ranging from 75-100. The expression of a selection of SAGE tags was scored in 49 human SAGE libraries available at NCBI/CGAP�s SAGE website at URL:(www.ncbi.nlm.nih.gov/SAGE). Libraries constructed from in vitro cell lines were omitted. The expression in the reference library Thy_T (PTC) was compared with respectively a group of 18 libraries from different normal tissues [Thyroid_N/Chen Normal Pr/PR317 normal prostate/normal prostate/mammary epithelium/Duke 40N/Duke 48N/Br N/normal lung/Duke leukocyte/Duke Kidney/NC1/NC2/Duke thalamus/Duke BB542 normal cerebellum/BB542 white matter/normal pool(6th)/normal cerebellum] and a group of 31 libraries from different tumor tissues (Thyroid_T/gastric cancer xenograft X101/gastric cancer-G234/Chen Tumor Pr/PR317 prostate tumor/PrCA-1/LN-1/Panc 91-16113/Panc 96-6252/OC14/OVT-6/OVT-7/OVT-8/95-347/95-259/95-260/95-348/DCIS/Duke 96-349/DCIS 2/Tu102/Tu98/Duke 1273/Duke H1020/Duke GBM H1110/pooled GBM/Ped GBM1062/Duke H1126/H1126/Duke 757/Duke H1043).

sqRT-PCR

First-strand RACE cDNA was synthesized from thyroid RNA samples using the cDNA synthesis system (Gibco-BRL). Primers were designed on GenBank sequences using the primer design program Primer3 at URL:(www.genome.wi.mit.edu/cgi-bin/primer/primer3.cgi). With the use of the following sequences, oligonucleotides were synthesized by Isogen: extracellular matrix protein 1 (ECM1) accession no. NM_004425, For (forward primer) 823-842, Rev (reverse primer) 1094-1113, fragment size 547 bp; thyroid peroxidase (TPO) accession no. M17755, For 2161-2180, Rev 2589-2608, fragment size 450 bp; hypothetical protein BC013035 accession no. NM_138436, For CGACTATCAGCAGCCACAAA, Rev TGCAAATGGCATAAACTCCA, fragment size 386 bp; mitochondrial ATPase 6 (ATP6) accession no. X62996, For CAGTGATTATAGGCTTTCGCTCTAA, Rev CAGGGCTATTGGTTGAATGAGTA, fragment size 190 bp (33). PCR amplification was performed using standard conditions, 2 mM MgCl2 and 30�� 94°C, 1� 55°C, 1� 72°C for an appropriate number of cycles. The amount of cycles used for ECM1 and TPO was 26, 28 for BC013035, and 25 for the control transcript ATP6. In all cases, this was during the exponential phase of the reaction, well before the plateau phase was reached. Ten microliters of every 25 µl PCR reaction were run on a 2% agarose gel and visualized using ethidium bromide. Imaging of the gel was performed using the Eagle Eye II System (Stratagene). Intensity of bands was measured using OneD-scan software from Scanalytics. Density values of every band are corrected for background and normalized for PCR fragment size. Ratios of respectively ECM1, TPO, and BC013035 over control transcript ATP6 are calculated.

RESULTS

Generation and comparison of SAGE libraries

A SAGE library was generated from the selected PTC tissue (Thy_T). Clones were sequenced resulting in 10,495 tags representing 5,913 unique transcripts. Together with the previously constructed normal thyroid library (Thy_N; ref 28), a total of 21,489 tags representing 11,350 unique transcripts are available for analysis. To identify transcripts up- or down-regulated in the thyroid tumor, both expression profiles are compared using USAGE software (30). SAGE libraries are normalized to a total tag number of 50,000. With the use of a cut-off significance P value of 0.05, 270 tags are differentially expressed, corresponding to 2.4% of the total number of transcripts in the two libraries. In this analysis, the significance P value of 0.05 corresponds to a differential expression of approximately fivefold.

Identification of differentially expressed SAGE tags

TAG-to-GENE identification is performed with USAGE software. Tags that cannot be annotated due to their specific sequence (e.g., AAAAAAAAAA or those containing ALU repeats) and tags corresponding to mitochondrial and ribosomal transcripts are excluded from subsequent analysis. The remaining cohort consists of 204 transcripts, of which 50 are up-regulated and 154 are down-regulated in the tumor. Of these tags, 60 cannot be linked to a known human transcript and are denoted NoMatch tags from which 27 tags are up-regulated and 33 tags are down-regulated in Thy_T. NoMatch tags do in most cases link to UniGene clusters consisting of ESTs or full-length cDNAs with unknown function. Of the remaining 144 tags with a positive UniGene match, 23 tags are up-regulated and 121 tags are down-regulated in Thy_T. Several thyroid-specific transcripts [e.g., thyroglobulin (TG) and TPO] are down-regulated. Genes previously

associated with cancer can be identified (e.g., VEGF, CD9, xpC, ECM1) in the group of up-regulated transcripts. Other up-regulated transcripts do not play a role in (thyroid) tumor biology.

TPE analysis

To investigate the expression profile of 204 differentially expressed transcripts in normal and tumor tissues other than thyroid, in silico analysis using TPE was performed. For each of the 204 tags, two TPE values are calculated to score the specificity toward two groups of tissues. The first group consists of 18 SAGE libraries from different normal tissues and analysis results in a TPE value termed TPE_N. The second group consists of 31 SAGE libraries from different tumor tissues. TPE analysis calculates a value TPE_T. Both values are plotted in a graph (Fig. 1). The bottom-left area of the plot represents differentially expressed tags in our SAGE analysis but with a low specificity for thyroid carcinoma when compared with either normal or tumor tissues. These tags typically correspond to widely expressed housekeeping genes. In the top-right area of the plot, tags with a differential expression specific for the Thy_T library when compared with either tumor or normal tissues are located. These tags are potential thyroid tumor markers because their corresponding genes are specifically over- or underexpressed in thyroid carcinoma when compared with carcinoma from different origin. Tags representing TG and TPO are located in this area. Table 1 lists 42 putative thyroid tumor markers from tags with at least one TPE value ≥75. Positive linkage with a UniGene cluster can be made for 30 tags (groups Match and NoMatch(+hit)) of which ~50% correspond to a UniGene cluster with a defined protein function. Twelve tags cannot be positively linked to any UniGene cluster (group NoMatch(-hit)).

Validation by RT-PCR

To validate the SAGE results in sqRT-PCR, a random selection was made of six transcripts. For this the original tissues from which the SAGE libraries were constructed were used (Fig. 2). The mitochondrial house-keeping gene ATPase 6 (33), which shows similar and relatively high expression levels in both SAGE libraries (131 tags in Thy_N and 157 tags in Thy_T), is used as a control transcript. The intensity of the amplified products corroborates the differential tag counts from the SAGE libraries, taking into account that a tag count of zero in a library of 10,000 tags does not necessarily mean that there is zero expression. The higher level of sensitivity of the RT-PCR experiment shows this as a weak band in lanes where the observed tag count was zero.

sqRT-PCR on thyroid tumor cDNA panel

The expression of three genes was studied in a panel of benign and malignant thyroid neoplasms. ECM1 for its association with metastasized breast carcinoma, TPO for its previously established relation with (de)differentiation of thyroid cells, and a novel transcript coding for a hypothetical protein (BC013035). Figure 3 shows the sqRT-PCR of ECM1, TPO, and BC013035 in a panel of 30 thyroid tumors and 12 normal thyroid controls. Figure 3A shows an agarose gel with RT-PCR products of ECM1, TPO, BC013035, and ATP6. PCR amplification products of ATP6 reflect the amount of cDNA input. Normalized expression ratios are depicted graphically in Fig. 3B. Normal thyroid tissue shows high expression of TPO, no ECM1 expression, and intermediate BC013035 expression. Normal thyroid tissue surrounding PTC tumors (samples A-H) shows slightly decreased expression of TPO, no ECM1 expression, and increased BC013035 expression. There is no ECM1 expression in follicular adenoma while the expression levels of TPO and BC013035 are decreased compared with normal. Papillary and follicular carcinoma

show decreased or absent expression levels of TPO. BC013035 is expressed in 10 out of 25 carcinoma with levels comparable to normal thyroid control. ECM1 is expressed in 11 out of 25 carcinoma showing no preference for PTC, fPTC, or FTC. There is no correlation between expression levels of ECM1, TPO, or BC013035, nor with the invasive characteristics of thyroid carcinoma, such as vasoinvasiveness, lymph node metastasis, and/or distant metastasis.

DISCUSSION

The comparison of SAGE-generated gene expression profiles of a normal thyroid and a follicular variant of PTC identified differentially expressed transcripts. After exclusion of tags corresponding to repetitive sequences, mitochondrial or ribosomal transcripts, 204 tags are differentially expressed, from a total of 11,350 different transcripts present in both libraries (1.8%); 154 of these are statistically significantly down-regulated and 50 are up-regulated (P<0.05) in the thyroid tumor. Similar numbers were reported in a SAGE study of lung cancer (26).

A general problem of high throughput techniques is the extensive output of data. To expedite the identification of SAGE tags of interest, we developed the in silico TPE (29) analysis that can be easily tailored to fit a specific research question, in casu the identification of transcripts differentially expressed in PTC. This �virtual Northern� uses an algorithm to calculate TPE values for every tag, based on the number of SAGE libraries the tag is expressed in and the relative level of expression in other libraries. TPE resembles techniques like Digital Differential Display (34) but since the algorithm assigns a standardized value to each tag, it is easier to sort tags based on their specificity making it possible to distinguish a subgroup of transcripts, tailored to a specific query.

The TPE analysis applied in this study was graphically plotted to select thyroid tumor candidate genes (Fig. 1). Contemplating functions of proteins corresponding to the TPE values shows that most housekeeping genes widely expressed in all tissues have low TPE values (<25). Proteins with a tissue-specific function from genes expressed in only one tissue have high TPE values (>75). The best example in this study is that TG, a gene expressed solely in thyroid follicular cells, has TPE values around 90 and is located in the most upper-right quadrant of Fig. 1. Moreover, all three tags corresponding to TG transcripts because of alternative splicing in the 3′UTR (35) show very high TPE values. TPO, another thyroid-specific transcript, down-regulated in thyroid tumors, also has TPE values around 90. The result of the TPE analysis shows that the largest part of tags differentially expressed in PTC is expressed in a wide range of tissues, normal as well as neoplastic. From 204 differentially expressed annotated tags, 42 tags have TPE values over 75 and are considered candidate thyroid tumor markers.

The most abundant thyroid-specific transcripts, TG and TPO, are significantly down-regulated in the tumor SAGE library. The expression of the low abundant TSH-R and NIS is lost in the tumor library; due to the relatively low levels of expression of these transcripts, statistical significance is not reached. It is generally accepted that the down-regulation of thyroid-specific transcripts in thyroid carcinoma is indicative for the dedifferentiation process during the progression of the tumor (7). This phenomenon was again shown in a study where PTC expression profiling using microarrays was performed (36). This study also identifies two genes (CITED1 and SFTPB) specifically overexpressed in PTC. Neither of these two genes were present in the SAGE profiles in this study.

Among the 50 transcripts that are up-regulated in PTC, transcripts coding for proteins with many different functions can be distinguished. VEGF is the only transcript in this group previously implicated in thyroid tumor progression (37, 38). The mRNA transcript from this protein is overexpressed 11-fold in our tumor SAGE library. Recent reports suggested a bad prognosis for VEGF-positive thyroid tumors (38, 39). VEGF, however, is not a thyroid tumor specific marker, as it is seen in many tumors of different origin. This is supported by the TPE values of the tag representing VEGF (TPE_N: 72 and TPE_T: 56), indicating that it is expressed in some normal tissues and in several tumor tissues.

In the only other thyroid SAGE study by Takano et al. (40), thyroglobulin is down-regulated in differentiated tumors and absent in anaplastic carcinoma. In our follicular variant of papillary carcinoma, thyroglobulin is also absent. This result may indicate the high level of dedifferentiation of the tumor, which was selected for its aggressive behavior. Osteonectin that was identified as a marker for anaplastic carcinoma by Takano is not present in our SAGE library.

sqRT-PCR on a panel of 30 thyroid tumors and 12 normal controls shows down-regulation of TPO in all tumors indicating dedifferentiation of these tissues. Additionally, normal control samples taken from healthy subjects show higher TPO expression then normal tumor-surrounding tissues. The expression pattern of hypothetical protein BC013035 shows a similar pattern to TPO. It is expressed in all normal thyroid samples, and its expression is down-regulated in most tumors, irrespective of the subtype. On the basis of the expression pattern of BC013035 in thyroid tumors, it does not seem a good tumor marker. Since it is expressed in normal thyroid tissue, it is more likely to play a role in general thyroid physiology.

ECM1 shows overexpression in 50% of PTC and 40% of FTC tumor samples. The lack of ECM1 in normal controls and follicular adenomas and its inverse correlation with the expression levels of TPO indicate that ECM1 expression correlates with tumor progression. It is not possible to distinguish between PTC and FTC on the basis of ECM1 expression, but it is possible to distinguish follicular adenoma from thyroid carcinoma.

The 1.9 kb ECM1 mRNA encodes a protein of 85 kDa that was initially isolated from an osteogenic stromal cell line (41, 42). Although ECM1 is implicated in differentiation of keratinocytes (43) and is expressed in surrounding connective tissues of developing bones (44), the function of the protein is still unknown. Recently, it has been associated with angiogenesis and it is expressed in breast carcinoma cells (45). In the TPE analysis, ECM1 expression was not observed in three available breast tumor SAGE libraries. The fact that in the former study expression of ECM1 was seen preferentially in breast tumors of the more malignant type could explain this difference. The link that was suggested between ECM1 expression and tumor progression through angiogenesis is in our case supported by the fact that the transcript for VEGF is up-regulated in our papillary thyroid tumor SAGE library to the same extent as ECM1. VEGF is a recognized angiogenic factor already described as a marker with prognostic value in thyroid carcinoma (37�39), although there are contradictory reports (46).

In conclusion, from a large-scale in silico analysis of gene expression profiles of normal thyroid and follicular variant of PTC, a cohort of 42 putative thyroid tumor markers is defined. Many novel transcripts are among them. ECM1 is identified as a gene up-regulated in a significant

proportion of differentiated thyroid carcinoma and is not expressed in follicular adenoma. ECM1 is associated with malignant breast carcinoma and angiogenesis.

ACKNOWLEDGMENTS

We thank Dr. A. H. C. van Kampen of the Bioinformatics Laboratory, Academic Medical Center, Amsterdam, The Netherlands for his assistance with the TPE analysis. This study was financially supported by the Stichting Kindergeneeskundig Kankeronderzoek (SKK). Dr. J. W. A. Smit is a recipient of a research grant from the Dutch foundation for scientific research (NWO) and the Dutch Cancer Foundation (KWF).

REFERENCES

1. Schlumberger, M. J. (1998) Papillary and follicular thyroid carcinoma. N. Engl. J. Med. 338, 297�306

2. Krohn, K., and Paschke, R. (2002) Somatic mutations in thyroid nodular disease. Mol. Genet. Metab. 75, 202�208

3. Dean, D. S., and Hay, I. D. (2000) Prognostic factors in differentiated thyroid carcinoma. Cancer Control 7, 229�239

4. Correa, P., and Chen, V. W. (1995) Endocrine gland cancer. Cancer Suppl. 75, 338�352

5. Gimm, O. (2001) Thyroid cancer. Cancer Lett. 163, 143�156

6. Elisei, R., Pinchera, A., Romei, C., Gryczynska, M., Pohl, V., Maenhaut, C., Fugazzola, L., and Pacini, F. (1994) Expression of thyrotropin receptor (TSH-R), thyroglobulin, thyroperoxidase and calcitonin messenger ribonucleic acids in thyroid carcinomas: evidence of TSH-R gene transcript in medullary histotype. J. Clin. Endocrinol. Metab. 78, 867�871

7. Ros, P., Rossi, D., Acebron, A., and Santisteban, P. (1999) Thyroid-specific gene expression in the multi-step process of thyroid carcinogenesis. Biochimie 81, 389�396

8. Wilkens, L., Benten, D., Tchinda, J., Brabant, G., Pötter, E., Dralle, H., and von Wasielewski, R. (2000) Aberrations of chromosomes 5 and 8 as recurrent cytogenetic events in anaplastic carcinoma of the thyroid as detected by fluorescence in situ hybridisations and comparative genomic hybridisation. Virchows Arch. 436, 312�318

9. Trovato, M., Fragetta, F., Villari, D., Batolo, D., Mackey, K., Trimarchi, F., and Benvenga, S. (1999) Loss of heterozygosity of the long arm of chromosome 7 in follicular and anaplastic thyroid cancer, but not in papillary thyroid cancer. J. Clin. Endocrinol. Metab. 84, 3235�3240

10. Zedenius, J., Wallin, G., Svensson, A., Grimelius, L., Höög, A., Lundell, G., Bäckdahl, M., and Larsson, C. (1995) Allelotyping of follicular thyroid tumors. Hum. Genet. 96, 27�32

11. Kroll, T. G., Sarraf, P., Pecciarini, L., Chen, C. J., Mueller, E., Spiegelman, B. M., and Fletcher, J. A. (2000) PAX8-PPARgamma1 fusion oncogene in human thyroid carcinoma. Science 289, 1357�1360

12. Pierotti, M. A. (2001) Chromosomal rearrangments in thyroid carcinomas: a recombination or death dilemma. Cancer Lett. 166, 1�7

13. Jhiang, S. M. (2000) The RET proto-oncogene in human cancers. Oncogene 19, 5590�5597

14. Rabes, H. M. (2001) Gene rearrangments in radiation-induced thyroid carcinogenesis. Med. Pediatr. Oncol. 36, 574�582

15. Russo, D., Arturi, F., Schlumberger, M., Caillou, B., Monier, R., Filetti, S., and Suarez, H. G. (1995) Activating mutations of the TSH receptor in differentiated thyroid carcinomas. Oncogene 11, 1907�1911

16. Suarez, H. G. (1998) Genetic alterations in human epithelial thyroid tumours. Clin. Endocrinol. (Oxf.) 48, 531�546

17. Jen Yeh, J., Lunetta, K. L., van Orsouw, N. J., Moore Jr, F. D., Mutter, G. L., Vijg, J., Dahlia, P. L. M., and Eng, C. (2000) Somatic mitochondrial DNA (mtDNA) mutations in papillary thyroid carcinomas and differential mtDNA sequence variants in cases with thyroid tumours. Oncogene 19, 2060�2066

18. Takano, T., and Amino, N. (2002) Cancer-specific mRNAs in thyroid carcinomas: detection, use, and their implication in thyroid carcinogenesis. Endocr. J. 49, 97�107

19. Nascimento, M. C., Bisi, H., Alves, V. A., Longatto-Filho, A., Kanamura, C. T., and Medeiros-Neto, G. (2001) Differential reactivity for galectin-3 in Hurthle cell adenomas and carcinomas. Endocr. Pathol. 12, 275�279

20. Herrmann, M. E., LiVolsi, V. A., Pasha, T. L., Roberts, S. A., Wojcik, E. M., and Baloch, Z. W. (2002) Immunohistochemical expression of galectin-3 in benign and malignant thyroid lesions. Arch. Pathol. Lab. Med. 126, 710-713

21. Niedziela, M., Maceluch, J., and Korman, E. (2002) Galectin-3 is not an universal marker of malignancy in thyroid nodular disease in children and adolescents. J. Clin. Endocrinol. Metab. 87, 4411�4415

22. Martins, L., Matsuo, S. E., Ebina, K. N., Kulcsar, M. A., Friguglietti, C. U., and Kimura, E. T. (2002) Galectin-3 messenger ribonucleic acid and protein are expressed in benign thyroid tumors. J. Clin. Endocrinol. Metab. 87, 4806�4810

23. Velculescu, V. E., Zhang, L., Vogelstein, B., and Kinzler, K. W. (1995) Serial analysis of gene expression. Science 270, 484�487

24. Madden, S. L., Wang, C. J., and Landes, G. (2000) Serial analysis of gene expression: from gene discovery to target identification. Drug Discov. Today 5, 415�425

25. Zhang, L., Zhou, W., Velculescu, V. E., Kern, S. E., Hruban, R. H., Hamilton, S. R., Vogelstein, B., and Kinzler, K. W. (1997) Gene expression profiles in normal and cancer cells. Science 276, 1268�1272

26. Hibi, K., Liu, Q., Beaudry, G. A., Madden, S. L., Westra, W. H., Wehage, W. H., Yang, S. C., Heitmiller, R. F., Bertelsen, A. H., Sidransky, D., et al. (1998) Serial analysis of gene expression in non-small cell lung cancer. Cancer Res. 58, 5690�5694

27. Smit, J. W., Van Zelderen-Bhola, S., Merx, R., De Leeuw, W., Wessels, H., Vink, R., and Morreau, H. (2001) A novel chromosomal translocation t(3;5)(q12;p15.3) and loss of heterozygosity on chromosome 22 in a multifocal follicular variant of papillary thyroid carcinoma presenting with skin metastases. Clin. Endocrinol. (Oxf.) 55, 543�548

28. Pauws, E., Moreno, J. C., Tijssen, M., Baas, F., deVijlder, J. J. M., and Ris-Stalpers, C. (2000) Serial analysis of gene expression (SAGE) as a tool to assess the expression profile of the human thyroid and to identify novel thyroidal genes. J. Clin. Endocrinol. Metab. 85, 1923�1927

29. Moreno, J. C., Pauws, E., van Kampen, A. H. C., Jedlickova, M., de Vijlder, J. J. M., and Ris-Stalpers, C. (2001) Cloning of tissue-specific genes using serial analysis of gene expression and a novel computational subtraction approach. Genomics 75, 70�76

30. van Kampen, A. H., van Schaik, B. D., Pauws, E., Michiels, E. M., Ruijter, J. M., Caron, H. N., Versteeg, R., Heisterkamp, S. H., Leunissen, J. A., Baas, F., et al. (2000) USA.GE: a web-based approach towards the analysis of SAGE data. Bioinformatics 16, 899�905

31. Kal, A. J., van Zonneveld, A. J., Benes, V., van den Berg, M., Groot Koerkamp, M., Albermann, K., Strack, N., Ruijter, J. M., Richter, A., Dujon, B., et al. (1999) Dynamics of gene expression revealed by comparison of serial analysis of gene expression profiles from yeast grown on two different carbon sources. Mol. Biol. Cell 10, 1859�1872

32. Lash, A. E., Tolstoshev, C. M., Wagner, L., Schuler, G. D., Strausberg, R. L., Riggins, G. J., and Altschul, S. F. (2000) SAGEmap: a public gene expression resource. Genome Res. 10, 1051�1060

33. Gerard, C. J., Andrejka, L. M., and Macina, R. A. (2000) Mitochondrial ATP synthase 6 as an endogenous control in the quantitative RT-PCR analysis of clinical cancer samples. Mol. Diagn. 5, 39�46

34. Scheurle, D., DeYoung, M. P., Binninger, D. M., Page, H., Jahanzeb, M., and Narayanan, R. (2000) Cancer gene discovery using digital differential display. Cancer Res. 60, 4037�4043

35. Pauws, E., van Kampen, A. H. C., van de Graaf, S. A. R., de Vijlder, J. J. M., and Ris-Stalpers, C. (2001) Heterogeneity in polyadenylation cleavage sites in mammalian mRNA sequences: implications for SAGE analysis. Nucleic Acids Res. 29, 1690�1694

36. Huang, Y., Prasad, M., Lemon, W. J., Hampel, H., Wright, F. A., Kornacker, K., LiVolsi, V., Frankel, W., Kloos, R. T., Eng, C., Pellegata, N. S., and de la Chapelle, A. (2001) Gene

expression in papillary thyroid carcinoma reveals highly consistent profiles. Proc. Natl. Acad. Sci. USA 98, 15044-15049

37. Katoh, R., Miyagi, E., Kawaoi, A., Hemmi, A., Komiyama, A., Oyama, T., and Shibuya, M. (1999) Expression of vascular endothelial growth factor (VEGF) in human thyroid neoplasms. Hum. Pathol. 30, 891�897

38. Klein, M., Vignaud, J. M., Hennequin, V., Toussaint, B., Bresler, L., Plenat, F., Leclere, J., Duprez, A., and Weryha, G. (2001) Increased expression of the vascular endothelial growth factor is a pejorative prognosis marker in papillary thyroid carcinoma. J. Clin. Endocrinol. Metab. 86, 656�658

39. Lennard, C. M., Patel, A., Wilson, J., Reinhardt, B., Tuman, C., Fenton, C., Blair, E., Francis, G. L., and Tuttle, R. M. (2001) Intensity of vascular endothelial growth factor expression is associated with increased risk of recurrence and decreased disease-free survival in papillary thyroid cancer. Surgery 129, 552�558

40. Takano, T., Hasegawa, Y., Matsuzuka, F., Miyauchi, A., Yoshida, H., Higashiyama, T., Kuma, K., and Amino, N. (2000) Gene expression profiles in thyroid carcinomas. Br. J. Cancer 83, 1495�1502

41. Mathieu, E., Meheus, L., Raymakers, J., and Merregeart, J. (1994) Characterization of the osteogenic stromal cell line MN7: identification of secreted MN7 proteins using two-dimensional polyacrylamide gel electrophoresis, western blotting and microsequencing. J. Bone Miner. Res. 9, 903�913

42. Smits, P., Ni, J., Feng, P., Wauters, J., van Hul, W., El Boutabi, M., Dillon, P. J., and Merregaert, J. (1997) The human extracellular matrix gene 1 (ECM1): genomic structure, cDNA cloning, expression pattern and chromosomal localization. Genomics 45, 487�495

43. Smits, P., Poumay, Y., Karperien, M., Tylzanowski, P., Wauters, J., Huylebroek, D., Ponec, M., and Merregaert, J. (2000) Differentiation-dependent alternative splicing and expression of the extracellular matrix protein 1 gene in human keratinocytes. J. Invest. Dermatol. 114, 718�724

44. Deckers, M. M. L., Smits, P., Karperien, M., Ni, N., Tylzanowski, P., Feng, P., Parmelee, D., Zhang, J., Bouffard, E., Gentz, R., et al. (2001) Recombinant human extracellular matrix protein 1 inhibits alkaline phosphatase activity and mineralization of mouse embryonic metatarsals in vitro. Bone 28, 14�20

45. Han, Z., Ni, J., Smits, P., Underhill, C. B., Xie, B., Chen, Y., Liu, N., Tylzanowski, P., Parmelee, D., Feng, P., et al. (2001) Extracellular matrix protein 1 (ECM1) has angiogenic properties and is expressed by breast tumor cells. FASEB J. 15, 988�994

46. Akslen, L. A., and Livolsi, V. A. (2000) Increased angiogenesis in papillary thyroid carcinoma but lack of prognostic importance. Hum. Pathol. 31, 439�442

Received March 3, 2003; accepted November 20, 2003.

Table 1

Putative thyroid tumour markers identified by SAGE and TPE analysis

TAG TAG-to-GENE Identification Thy_NThy_T P value TPE-N TPE-T UniGene_ID Locus

Match

CTGTGCTCTA ADP-ribosylation factor-like 5 0 37 0.0059 88 77 Hs.342849 2q23.3

AGCCTGCTCA CD73 (5'nucleotidase) 0 24 0.0245 82 88 Hs.153952 6q14-q21

ACCGTCCACT cyclophilin C 0 24 0.0245 82 76 Hs.110364 5q23.2

ACTGCCCGCT extracellular matrix protein 1 (ECM1) 0 55 0.0007 93 71 Hs.81071 1q21

ATCCTTTTAT golgin-67 0 91 <0.0001 95 98 Hs.182982 15q11.2

AGGACAAATA iduronate 2-sulfatase 0 24 0.0245 82 85 Hs.172458 Xq28

ACTGTATTGG quinone reductase (crystallin Z) 0 24 0.0245 91 96 Hs.83114 1p31-p22

ATGACAGATG SMAP31/homeodomain only protein (HOP) 10 164 <0.0001 75 76 Hs.13775 4q11-q12

TAAGAATTAA myozenin 2 0 43 0.0029 93 91 Hs.381047 4q26-q27

CACGGAGGCG xeroderma pigmentosum, group C 0 24 0.0245 95 85 Hs.320 3p25

ACCGCAGTGC insulin-like growth factor binding protein 4 0 24 0.0245 95 93 Hs.1516 17q12-q21.1

AGGCCCACAA mannosidase, alpha2C 0 43 0.0029 79 82 Hs.26232 15q11-q13

TTTTCCAAAA ring finger protein 13 87 12 0.0021 80 90 Hs.6900 3q25.1

CGGTGAAGCA thyroglobulin (TG) 217 0 <0.0001 88* 96* Hs.305916 8q24.2-q24.3

GATGAATAAA thyroid peroxidase (TPO) 106 0 <0.0001 82 96 Hs.2041 2p25

NoMatch +hit

ATAGCTGGTG KIAA0220 protein 0 30 0.0119 96 96 Hs.110613 16p12.1

AAGCAAGAAT KIAA1376 protein 0 24 0.0245 82 85 Hs.24684 5q14.3

AACAGCTTTA hypothetical protein BC013035 0 24 0.0245 91 85 Hs.10018 8p11.1

GTCATAGCTG hypothetical protein MOT8 0 414 <0.0001 100 100 Hs.25924 1p36.31

AGGAACTCAT cDNA FLJ31203 fis 0 43 0.0029 97 97 Hs.350840 X

CCAGCCCTAG

cDNA FLJ13793 fis, highly similar to PAIRED BOX

PROTEIN PAX-8 0 110 <0.0001 100 92 Hs.150539 2

TTGGCTGGGC EST 0 49 0.0015 79 86 Hs.274511 -

ACTGCTATTT ESTs 0 24 0.0245 95 96 Hs.220971 -

TAAAAAGTGG ESTs 0 24 0.0245 95 96 Hs.137928 -

TTCCGAGGTT ESTs 0 24 0.0245 95 96 Hs.40840 -

TATGTACTCC EST 0 79 0.0001 94 75 Hs.240443 -

CTTCCTGGCC EST 5 73 0.0005 85 83 Hs.83623 -

CCAGCTGCCT hypothetical protein MGC45438 34 0 0.0186 64 92 Hs.11782 16p13.13

GCAAGCCCCA hypothetical protein MGC40397 34 0 0.0186 68 89 Hs.295563 12q23.3

CAGTGAAAAA ESTs 43 0 0.0076 58 83 Hs.203763 -

NoMatch -hit

ATACAGTTCC NoMatch B07 0 61 0.0004 98 97 - -

GCATAGCTGT NoMatch B08 0 61 0.0004 98 97 - -

GTAACTTGGG NoMatch B10 0 49 0.0015 97 97 - -

ATGTATTGAT NoMatch B12 0 43 0.0029 97 97 - -

ATTGATGTGT NoMatch B28 0 37 0.0059 91 97 - -

GCCGTGCCGC NoMatch B56 0 24 0.0245 95 93 - -

TAGTTGCACA NoMatch B59 0 24 0.0245 69 85 - -

TGTGTTGAGG NoMatch B64 0 24 0.0245 86 79 - -

GTACAGGACT NoMatch B67 0 24 0.0245 95 96 - -

TGGTGAAAAA NoMatch B44 72 0 0.0006 82 78 - -

TAGATACTTT NoMatch B47 43 0 0.0076 79 95 - -

TACCTAATTG NoMatch B65 24 0 0.0467 56 83 - -

Tags with a preferential differential expression in thyroid papillary carcinoma SAGE library Thy_T (see also Fig. 1). Tags are divided in 3 groups: Match, NoMatch (+hit), and NoMatch (-hit). Corresponding transcript, tag count in libraries Thy_N and Thy_T (normalized per 50.000 tags), P value (Z-test), TPE values, UniGene cluster accession number, and cytogenetic locus are shown. Transcripts in bold are used in RT-PCR validation (Fig. 2). *Average TPE values from 3 tags corresponding to alternatively spliced TG transcripts.

Fig. 1

Figure 1. TPE Anaylsis. TPE plot of 204 differentially expressed SAGE tags in papillary thyroid carcinoma. TPE values were calculated as described in Material and Methods. The circle contains tags with at least one TPE value >75, which are selected as potential thyroid tumor markers. White squares are upregulated Match tags, and black squares are downregulated Match tags. White circles are upregulated NoMatch tags, and black circles are downregulated NoMatch tags. TPE-N (x-axis) value is calculated with data from 18 normal tissue SAGE libraries, and TPE-T (y-axis) is from 31 tumor tissue SAGE libraries.

Fig. 2

Figure 2. RT-PCR validation of SAGE results. Validation by RT-PCR on RNA from tissue used in the SAGE analysis of selected transcripts showing differential expression and high TPE values after subtraction of normal (Thy_N) and papillary thyroid carcinoma (Thy_T) SAGE libraries. Transcripts in bold are used in subsequent analysis. PCR fragments are analyzed on a 1.5% agarose gel. Lanes are marked N: normal thyroid Thy_N; T: tumor Thy_T, with their SAGE expression level in brackets. M = DNA marker.

Fig. 3

Figure 3. SqRT-PCR on thyroid tumor cDNA panel. Semi-quantitative RT-PCR analysis of ECM1, TPO, and cDNA BC013035 on a panel of 30 thyroid tumors and 12 controls. A) gel with PCR products; ECM1 547bp, TPO 450bp, BC013035 465bp, ATP6 190bp. Tumors samples as used in SAGE analysis are depicted as Thy_N and Thy_T. B) Relative expression levels of, respectively, ECM1, TPO, and BC013035 expressed as their ratio divided by ATP6, corrected for background and size of the fragment. Samples are divided into 4 groups: N (normal thyroid); FA (follicular adenoma); PTC (papillary carcinoma), with follicular variants marked fPTC; and FTC (follicular carcinoma). Pairs of tissues (tumor and normal tissue from the same patient) are indicated by A-H.