RNA from snap-frozen tissues was isolated using the RNeasy Mini Kit and RNeasy Lipid Tissue Mini Kit for the adipose tissue samples, respectively (QIAGEN, Hilden, Germany).

For the formalin-fixed paraffin-embedded (FFPE) tissues of thyroid tumors, histopathologic diagnoses were performed according to the World Health Organization Classification of Tumours 21 (table 1). As to RNA isolation, FFPE blocks were cut into six sections of 5 μm for each sample using a microtome. Total RNA isolations were performed using the Roche High Pure RNA Paraffin Kit (Roche, Mannheim, Germany) for the THADA expression investigation and the RNeasy FFPE Kit (QIAGEN, Hilden, Germany) for the NIS expression analysis. Three samples were cytogenetically characterized by 2p21 translocations. In all three cases, two of which published previously 22 23 , the breakpoints were narrowed down to the THADA locus. One of the anaplastic thyroid samples served as the source of a newly established cell line. Cytogenetical analysis revealed a highly complex karyotype with a range of 80 to 117 chromosomes (100.8 on average). Several marker chromosomes, telomeric associations, and double minutes were detected.

RNAs were reverse-transcribed into cDNA by M-MLV Reverse Transcriptase (Invitrogen, Karlsruhe, Germany). Real-time PCR was performed using the Applied Biosystems 7300 sequence detection system according to TaqMan Gene Expression Assay Protocol (Applied Biosystems, Darmstadt, Germany) in 96-well microtiter plates with a total volume of 20 μl. In case of TaqMan gene expression assay of THADA (assay number Hs00152982, Applied Biosystems, Foster City, USA), targeting exons 31-32, and of NIS (assay number Hs00166567_m1), each reaction consisted of 2 μl of cDNA reverse transcribed from 25 ng of total RNA, 10 μl of TaqMan Universal PCR Master Mix (Applied Biosystems), 1 μl of TaqMan assay and 7 μl of ddH₂O. For the 18S rRNA assay, using 18S forward and 18S rev_1 primers 24 , each reaction consisted of 2 μl of cDNA (1:10 diluted, with regard to higher expression of 18S rRNA) reverse transcribed from 25 ng of total RNA, 10 μl of TaqMan Universal PCR Master Mix, 600 nM of forward and reverse primers, 200 nM of 18S probe 24 and 5.4 μl of ddH₂O.

Thermal cycling conditions were 2 min at 50°C followed by 10 min at 95°C, 50 cycles at 95°C for 15 s and 60°C for 1 min. A non-template control of amplification and two previous negative controls of cDNA synthesis (one without RNA and one missing Reverse Transcriptase) were included in each plate. Software Sequence Detection Software 1.2.3 (Applied Biosystems) was programmed with the reaction condition. All testing reactions were performed in triplicate.

Serial dilutions were made using cDNA derived from 25, 5, 1, 0.2, and 0.04 ng of total RNA from FFPE tissue of one thyroid adenoma for THADA and 18S rRNA, and from fresh frozen tissue of one normal thyroid sample for NIS. In each dilution, THADA, NIS, and 18S rRNA gene expression assays were performed using absolute quantification. Afterwards, the standard curves for both assays were plotted with the log ng of input cDNA for each dilution on the x-axis, and the matched C_T value on the y-axis. Furthermore, in order to evaluate the differences of amplification efficiencies, the difference of two curve slopes was calculated. If the absolute difference of the slopes is less than 0.1, the amplification efficiencies of two assays are considered to be equal and the comparative C_T method is valid (User Bulletin No. 2, ABI PRISM 7700 Sequence Detection System, Applied Biosystems). 18S rRNA was used as endogenous control as suggested previously 25 26 27 28 . The 18S rRNA assay showed an amplification efficiency of 92.6% (slope = -3.514, R² = 0.995). The THADA assay had an amplification efficiency of 92.0% (slope = -3.531) and an R²-value of 0.96. For NIS, the amplification efficiency was 93.4% (slope = -3.4917), the coefficient of determination amounted to 0.997). As recommended for FFPE samples 24 29 30 31 the fragment sizes amplified by all three assays were small, ranging between 60 and 78 bp, a validation of these values was performed via gelelectrophoresis of the PCR-products (data not shown). When applying the comparative C_T method, one histological normal thyroid tissue was used as calibrator sample. Afterwards, data were compared with results from conventional histology.

For statistical analysis, the Wilcoxon signed rank test was used to compare average values (two-sided, exact version for at most 40 cases involved, otherwise using the t approximation); relationships were quantified by linear regression and Spearman's rank correlation coefficient. Sensitivity, specificity and decision limits were calculated from non-parametric density estimations. Therefore, sensitivity and specificity may differ from raw empirical values and decision limits need not coincide with measured values. A p-value of less than 0.05 was considered significant.

The use of human thyroid samples for this study was approved by the local medical ethics committee (Ethikkommission bei der Ärztekammer Bremen) and followed the guidelines of the declaration of Helsinki. Only samples that were initially taken for diagnostic purposes were secondarily used for the present study. During pathological examination, a sample of the tissue was snap-frozen. The procedure was approved by the local ethics committee. Because the samples were deidentified and were considered as samples normally discarded, the committee felt that there was no specific patient consent necessary.

As for the normal tissue samples, these were anonymously collected for earlier studies, each following the guidelines of the declaration of Helsinki.

Thirty-four snap-frozen samples from eight different tissues were tested for the level of THADA expression.

The mean level per tissue type ranged from 1 (blood) to 6.14 (thyroid), and the lowest single value for a thyroid sample (4.04) was above the highest one (3.39, myometrium) from any of the other tissues (Figure 1). Accordingly, statistical analysis using Wilcoxon's exact signed rank showed significant differences between normal thyroid tissues and the group of all other tissues (p < 0.0001). Using the THADA expression to discriminate between thyroid and non-thyroid tissue, a sensitivity of 82.5%, a specificity of 97.4% and an efficiency of 95.2% with a decision limit value of 4.23 were achieved.

Ninety-three formalin-fixed-paraffin-embedded thyroid samples, including eight normal tissues (from four patients), 18 goiters, 35 benign, and 32 malignant tumors were measured. For single tumor samples the expression ranged between 0.065 (anaplastic carcinoma) and 2.986 (follicular adenoma) in relation to normal tissue, i.e. a ratio of 1 : 45.94. Samples with a 2p21 translocation showed a level of expression of 1.123, 1.624, and 0.662 fold, respectively. The mean values for the different tumor entities ranged from 0.423 (anaplastic carcinoma) to 1.156 (adenoma) (Figure 2 and table 2).

Significant differences of THADA expression were noted between benign and malignant thyroid lesions. Wilcoxon's signed rank test showed a highly significant difference comparing the joint group of goiters and benign tumors with malignant tumors (p = 0.0009).

Using the exact Wilcoxon test, no significant differences were detected comparing the level of THADA expression between normal tissue and benign lesions (p = 0.2802) and papillary carcinomas (p = 0.2170). In contrast, significant differences were found between anaplastic carcinomas (ATCs), the most dedifferentiated type of thyroid tumors, and all other samples (p = 0.0107) and ATCs and all other malignant tumors (p = 0.0234). Comparing anaplastic carcinomas with each single group, the difference in expression between ATCs and goiters (p = 0.0049) and adenomas (p = 0.0058) were marked as significant. As this finding was a result of systematically comparing anaplastic carcinomas with the other lesions, a Bonferroni correction for multiple testing was used (corrected α = 0.0083). Without the need of correcting for multiple testing also normal tissue and papillary carcinoma would have been assessed as significantly different from anaplastic carcinoma (p = 0.0485 and p = 0.0350, respectively). Overall, significant results were mostly seen with the group of anaplastic carcinomas, indicating a relative stable level of expression in comparatively differentiated tissues with a significant reduction only in dedifferentiated tissues.

Recently HMGA2 expression has been shown to indicate thyroid malignancy and can thus be considered marking the dedifferentiation of thyroid epithelium 32 33 34 . As to the study by Belge et al. 32 and the present one 48 samples were identical in both studies (seven normal tissues, one goiter, 15 adenomas and 25 carcinomas, including three anaplastic carcinomas). For these, RNA was isolated from adjacent cuts of the same FFPE block and, except for the different qRT-PCR assays, all samples were treated identical in both investigations. Thus, it was feasible to check these samples for a possible correlation between THADA and HMGA2. Using Spearman's rank correlation, there was a highly significant inverse correlation between THADA and HMGA2 expression (correlation coefficient = -0.452; p = 0.0015), further underlining a possible role of THADA in thyroid differentiation.

NIS (sodium-iodide symporter), the transmembrane glycoprotein accountable for the uptake of iodine in thyroid cells, was found to be a marker of thyroid differentiation 35 36 37 38 . To validate our findings NIS expression was measured in 41 samples, including seven normal tissue samples, six nodular goiters, five adenomas, and 23 carcinomas (15 papillary, four follicular, and all four anaplastic thyroid carcinomas). Using Spearman's rank correlation, no significant correlation (p = 0.1288) was detected comparing THADA and NIS expression from all samples. By contrast, a significant correlation was found constraining the analysis to the follicular and papillary carcinoma samples (p = 0.0497, r = 0.456, n = 19), an even stronger correlation between the expression of THADA and NIS was found in normal and all malignant samples (p = 0.0021, r = 0.540, n = 30), and in normal tissue and anaplastic carcinomas (p = 0.0128, r = 0.718, n = 11)

Using the SABiosciene DECODE Transcription Factor Search, no THADA-promotor binding sites for thyroid-specific transcription factors paired box gene 8 (pax8), thyroid transcription factor 1 (TTF1), also known as NK2 homeobox 1 (NKX2-1), and thyroid transcription factor (TTF-2), sometimes referred to as forkhead box protein E1 (FOXE1), were found. Amongst others cAMP response element-binding protein (CREB), activating transcription factor (ATF-2), c-Jun, hepatic leukemia factor (Hlf), and germ cell nuclear factor (GCNF) were marked as relevant, FOXC1, Nkx2-2, Nkx2-5, and Nkx6-1 were displayed with low relevance (data not shown). HHEX (hematopoietically expressed homeobox) has been found to be expressed in the adult thyroid gland and in differentiated thyroid cell lines and to be correlated with thyroid differentiation 39 40 41 , but is not included in the SABiosciene DECODE Transcription Factor Search. A manual search for this transcription factor revealed no assured binding sites in the THADA promoter.