RTH, caused by mutation of the TRβ gene, is the consequence of abnormal transcription of thyroid-hormone-responsive genes in T3 target tissues. Using a mouse cDNA microarray containing 11,500 gene probes (representing over 10,000 unique named and unnamed expressed sequence tags (ESTs)), we profiled gene-expression patterns in mouse models of hyperthyroidism and RTH to reveal T3-responsive (hyperthyroid-associated) and mutant PV-dysregulated genes. The three T3-target tissues profiled in this study were the cerebellum, heart and WAT, all of which are important T3-target tissues in which T3-responsive genes have yet to be comprehensively analyzed. The experimental comparisons in this study included: wild-type mice versus TRβ^PV/PV littermates (the RTH group); wild-type mice injected with T3 versus those receiving saline injection only (the iT3 group); and wild-type mice versus a second set of wild-type animals selected in the same way (the control group).

In the RTH experiments, we identified changes in gene expression resulting from the in vivo action of PV and/or elevated T3 characteristic of clinical RTH. The iT3 experiments revealed genes that are responsive to increased levels of T3 in pharmacologic hyperthyroidism. These experiments allowed us to analyze the expression profiles in RTH that are the effects due to the increased levels of circulating thyroid hormone and thus to understand which TR-subtype-dependent transcriptional mechanisms govern gene expression. The purpose of the control group was to identify genes that naturally fluctuate as the result of normal inter-animal differences 15. Genes identified as differentially expressed in the control group were subsequently censored in a tissue-dependent manner. For each experimental comparison, replicate microarray hybridizations were performed with reciprocal labeling (dye-swapping) to minimize technical noise (see Materials and methods).

Differentially expressed genes (outliers) were selected based on the 2.0r criterion whereby the fold change detected on each of two reciprocally labeled arrays (r designation) was greater than twofold (up or down, but in a reciprocal manner) in at least one experimental comparison. The average expression ratios were calculated as described (see Materials and methods) and are used herein as the expression ratio measurements. To evaluate the probe-level reproducibility of the microarray data, we examined the variability of the expression ratios for all genes represented by multiple probes on the array and identified as outliers in multiple experimental comparisons (see Additional data file 1). A high degree of expression-ratio reproducibility was observed, even for changes less than twofold, lending confidence to our microarray results and outlier selection criteria. Notably, the 2.0r criterion was applied to achieve a high level of stringency in detecting gene outliers. However, this approach precludes the identification of genes that are differentially expressed at levels below 2.0r detection (for example, 1.9-fold change) thus obscuring the distinction of tissue-specific outliers and excluding potentially important distinguishing gene cassettes. In the light of this, we compiled a dataset that lists all genes passing the 2.0r criterion in at least one experimental comparison and identifies all instances (in other experimental comparisons) where these genes pass a less stringent threshold of 1.2r (see Additional data files 2 and 3). Thus, a transcript that shows a minimum change of twofold in one comparison can be subsequently screened for smaller changes, or categorized as having virtually no change in expression (that is, less than 1.2-fold), in other comparisons.

To gauge the accuracy of our methodology in correctly identifying T3-regulated genes, we searched for the 2.0r outliers across the three tissues for genes or gene products previously known to be regulated by T3. Representative genes are listed in Additional data file 4 and include 10 genes whose mRNA expression have been reported previously to be regulated by T3 and five genes that are known to be affected by T3 at the protein level. Two genes categorized as effectors are also listed. These include transthyretin (Ttr), a high-affinity serum thyroid-hormone-binding protein that transports thyroid hormone to target tissues, thus facilitating thyroid hormone distribution 16, and the retinoid X receptor α (Rxra), a member of the nuclear receptor superfamily that heterodimerizes with TRs and modulates the transcriptional activity of TRs 1.

To confirm the validity of the microarray results, four outliers (two up- and two downregulated genes) each were selected from the cerebellum and WAT for the determination of mRNA expression by reverse transcription PCR (RT-PCR) analysis. The patterns and fold changes in the expression of adenomatous polyposis coli (APC), somatostatin, carbonic anhydrase 4 and tenascin C in the cerebellum (Figure 1a) and those of lysyl oxidase, carbonic anhydrase 4, enolase 3β and creatine kinase in the WAT (Figure 1b) determined by RT-PCR were all consistent with those obtained by the arrays. Even though there was a slight variability in the magnitude of the response between the two methods, the patterns of response were in concordance.

The number of genes that are differentially expressed in a tissue following T3 treatment can be viewed as an indirect measure of pharmacodynamic effect. For example, tissues that show a large number of gene outliers in response to T3 treatment would suggest higher sensitivity to T3 than tissues responding with a few outliers. This has been found to be true for chemotherapeutic effects, and for estrogen responsiveness 17. This principle can therefore be used as a metric for assessing tissue sensitivity to T3 treatment or the PV mutant. A breakdown of the number of 2.0r outliers identified in each tissue of each comparison is shown in Table 1. The number of outliers varied markedly between tissue types. For example, in the WAT, 58 and 172 outliers were detected in each of the comparisons (that is, with versus without T3 treatment (iT3 mice) and wild-type versus TRβ^PV/PV mice (RTH group), respectively), whereas in the cerebellum only 21 were found in the iT3 group and none was found in the RTH group. In the heart, 98 outliers were found in the iT3 group and 17 in the RTH group. Therefore, adult cerebellum appears to be the tissue least sensitive to T3 treatment and the action of PV, whereas the heart is a particularly responsive tissue in our model of hyperthyroidism, and white adipose is the most sensitive tissue in our model of RTH.

Intriguingly, the distribution patterns of up- and downregulated genes showed a consistent trend in two tissues that distinguished the hyperthyroid and RTH conditions (Table 1). In both the heart and WAT, the predominant response to T3 administration (that is, iT3 group) was mostly downregulation of expression (71% and 76% of the outliers, respectively). By contrast, in the TRβ^PV/PV mice we observed higher proportions of transcriptionally induced genes. In the heart and WAT of RTH animals, 65% and 53% of outliers, respectively, were upregulated genes. Together, these data support the intriguing view that the hyperthyroid phenotype is largely mediated through T3-induced suppression of gene expression, whereas RTH is underscored by an inappropriate increase in transcriptional activity in T3 target tissues. These data strongly suggest that target-tissue response to T3 administration and the RTH genotype are distinctly different.

To further characterize the target-tissue transcriptional response, we analyzed each tissue for its degree of transcriptional uniqueness. We found that in the T3-treated mice, 24% (5/21), 24% (24/98), and 22% (13/58) of outliers in the cerebellum, heart and WAT, respectively, were tissue specific (that is, differentially expressed by twofold or more in only one tissue, and not more than 1.2-fold in any other tissue). In the RTH mice, 0% (0/0), 53% (9/17), and 47% (81/172) of outliers in the cerebellum, heart and WAT, respectively, were detected in only one tissue type (data not shown). These observations reveal a considerable degree of transcriptional specificity that is dependent on tissue type, consistent with the notion that the transcriptional behavior of TRs depends on the molecular composition of the cell type with respect to the availability and levels of cofactors that modulate TR transcriptional activity 1418.

The pathological hallmark of RTH is elevated thyroid hormone associated with nonsuppressible TSH. Even though the phenotypic consequences of elevated thyroid hormone in RTH patients are known, it is not clear what genes mediate the phenotypes. Furthermore, RTH may manifest as hyperthyroidism in one tissue, and, simultaneously, hypothyroidism in another 9. Dissecting this variable phenotype into the hyperthyroid and non-hyperthyroid components at the molecular level requires a comprehensive knowledge of the hyperthyroid-related genes in each tissue, and the expression patterns of these genes in the context of RTH. Therefore, we performed a tissue-by-tissue analysis cross-comparing the genomic effects of iT3 and RTH to segregate genes according to cognate response patterns.

Four categories of response patterns were discerned and labeled A, B, C and D, as shown in Figure 2. The outliers in category A (brown bars in Figure 2) are the genes that showed the same directionality of change in both T3-treated and TRβ^PV/PV mice (that is, up in both or down in both). That the mutant TRβ does not alter the transcriptional response of genes responsive to elevated T3 suggests their potential involvement in the hyperthyroid phenotype. As shown in Figure 2, two genes were identified in the cerebellum, five in the heart, and 39 in the WAT.

Category B (orange bars) includes the outlier genes that showed reversed patterns of change between iT3 and RTH mice (that is, up in iT3 but down in RTH, or vice versa). Presumably, these genes, which are responsive to increased T3 levels (in iT3 mice), are expressed in the opposite direction in TRβ^PV/PV mice as a consequence of mutant TRβ activity. One such outlier was identified in the cerebellum, 25 in the heart, and 33 in the WAT. Interestingly, 84% (21/25) and 79% (26/33) of these genes in the heart and WAT, respectively, were downregulated in response to T3 treatment and concomitantly upregulated in RTH, further supporting the view that an important molecular aspect of RTH is the abnormal expression of otherwise T3-repressed genes. A mechanistic explanation is that these category B genes are regulated by T3 mostly (or exclusively) through the TRβ receptor.

The majority of the outliers identified in this study, however, were found exclusively in either iT3 (category C, blue bars) or RTH (category D, pink bars) mice. Genes in category C showed a greater than twofold change in iT3 mice and less than 1.2-fold change in RTH mice. Therefore, these genes were responsive to increased T3 in the iT3 group but insensitive to elevated T3 in the presence of the mutant TRβ receptor. In this category, we identified 18, 72 and 27 such genes in the cerebellum, heart and WAT, respectively. Conversely, the outliers in category D were detected in RTH but not in the iT3 group. Compared with category C, a relatively smaller number of such genes were identified in the cerebellum (0 genes) and the heart (3 genes), but a significantly larger number of outliers (116 genes) were detected in the WAT. Taken together, these data show that the expression patterns of iT3 and RTH responsive genes are not only tissue-dependent, but also can differ markedly between the hyperthyroid and RTH states within the same tissue.

Hierarchical clustering of gene-expression patterns can provide a robust, composite view of cellular pathways operative in a biological system as evidenced by the coordinate expression of functionally related genes. To gain insight into the pathways most perturbed in our models, we analyzed the expression patterns of genes differentially expressed in one or more tissues of hyperthyroid (iT3) or RTH mice. In Figure 3, the expression patterns of these genes were hierarchically clustered and the functionality of genes within clusters was analyzed using Gene Ontology (GO) terms. We identified three gene clusters with ostensible biological implications: A, an immunity cluster: B, a lipogenesis cluster; and C, a cell cycle/growth inhibitory cluster. The immunity cluster is largely characterized by a trans-tissue downregulation of genes involved in immune response pathways or immune-cell biology (A, black bar, Figure 3). These genes are repressed in one or more tissues of T3-treated and/or TRβ^PV/PV mice and are all consistently downregulated in the WAT of TRβ^PV/PV mice. These include several HLA class II antigen genes (H2-Ab1, H2-Aa, and H2-Eb1), chemotactic factor genes (Ccl22 and Ccl19), and genes involved in lymphocyte activation (Lcp2, Ptprcap and Ms4a1) and adherence (Sell), strongly suggesting an immunomodulatory response in tissues of both hyperthyroid and TRβ^PV/PV mice.

In the lipogenesis cluster (B, blue bar, Figure 2), six of eight genes have clearly defined roles in fatty-acid and lipid metabolism and include Fasn, Acly, Gpd1, Elovl6, Slc25a1, and a gene named 'similar to acetyl-coenzyme A carboxylase, clone IMAGE: 5151139' which has 99% identity at the protein level with rat acetyl-coenzyme A carboxylase. These genes are consistently repressed by T3 in the heart and WAT, and expressed at reduced levels in the WAT of TRβ^PV/PV mice, with half of the genes being induced in the heart of TRβ^PV/PV mice (blue bar, Figure 3). These patterns are consistent with the negative regulation of lipogenesis in the hyperthyroid heart and WAT. In RTH mice this suggests suppression of lipogenesis in the WAT with simultaneous increase in lipogenesis in the heart.

The smaller cell-cycle/growth-inhibitory cluster is composed of several genes known to inhibit growth or cell-cycle progression or promote apoptosis (C, pink bars, Figure 3). These genes include Bnip3, Gadd45a and Cdkn1a (p21). Shown adjacent to this gene cluster is the similarly expressed gene, APC (adenomatous polyposis coli), whose expression is also associated with growth inhibition (that is, via suppression of the Wnt signaling pathway which we recently showed to be negatively regulated by T3 19). A hallmark of these genes (with the exception of Gadd45a) is increased expression in all three tissues of the hyperthyroid mouse. Notably, p21 is also upregulated in the heart and WAT of TRβ^PV/PV mice, suggesting its induction in these tissues in response to high circulating T3. These findings raise the possibility that induction of growth-inhibitory mechanisms in multiple T3 target tissues may contribute to the pathogenesis of RTH.

The heart is highly sensitive to thyroid hormone and the primary mode of T3 action is a direct influence on cardiac gene expression 20. Increased heart rate (tachycardia) and myocardial contractility are clinical features common to both hyperthyroidism and RTH 2122. As the predominant TR isoform in the heart is TRα1, and RTH patients have normal expression levels of TRα1 together with elevated thyroid hormone, it has been postulated that the cardiac effects common to both hyperthyroidism and RTH are mediated by TRα1 signaling 2023. To examine this hypothesis at the molecular level, we compared the genomic effects of induced hyperthyroidism and RTH in the mouse heart. In total, we identified 105 gene outliers that showed twofold change or more in the iT3 or RTH groups (see Additional data file 5). Notably, this list does not include several genes previously implicated in T3-mediated cardiac effects such as those for the α- and β-myosin heavy chains, SERCA2 and HCN2, as they were not represented on our microarray. To focus on the set of most relevant genes in our list, we used the GO database of biological and functional gene classifications 24 to define genes known or expected to have a role in cardiac muscle contraction.

Hence, we classified 23 genes into one or more of the following categories (Table 2): muscle-related; calcium-ion binding (calcium-ion release and retrieval is central to muscle contraction and relaxation); ATP binding and electron transport (ATP is the immediate source of energy that powers muscle contraction); mitochondrion related (cardiac muscle has the richest supply of mitochondria and T3 directly boosts energy metabolism in mitochondria via gene transcription).

We hypothesized that the expression patterns of these genes would be similar in the iT3 and RTH hearts as they are presumably regulated by TRα1 and responsive in the context of elevated T3. Surprisingly, however, this analysis indicated that the iT3 and RTH hearts are molecularly distinct. Of the 23 genes identified, none showed the same or similar response patterns. Eighteen genes, which ranged from fourfold induction to 20-fold suppression in the iT3 heart, showed less than a 1.2-fold change in RTH (that is, category C genes, Figure 2). Moreover, the remaining five genes were expressed in opposite directions (that is, category B genes, Figure 2). These include the gene for myosin light chain, regulatory A (> 20-fold repressed in iT3 and > 6.6-fold increased in RTH) and the sarcolipin gene (> 4.8-fold repressed in iT3 and > 3.6-fold increased in RTH).

This expression pattern trend was also apparent in the remaining 82 non-muscle-contraction-related outliers (see Additional data file 5). Of these genes, 54 showed from 2- to 11-fold change in response to T3 injection and concurrently no change in RTH mice, while only three showed a greater then twofold change in RTH with no concurrent change in iT3 mice. Of the 25 genes in this group that showed change in both iT3 and RTH mice, only five showed change in the same direction (that is, candidates for TRα1-dependent transcription, category A genes, Figure 2), whereas the remaining 20 genes showed opposing expression patterns. Such marked differences between iT3 and RTH hearts suggest a greater role for TRβ in the heart than previously envisioned, or interference of TRα1-mediated transcription by TRβPV or its downstream effects.

Thyroid hormone is an important regulator of adipose tissue development and metabolism. T3 is thought to exert its effects on fat cells through transcriptional regulation by both α1 and β receptors 25. Although a number of T3 target genes in adipose tissues have been described 2627, little is known of the transcriptional response of adipocyte T3 target genes in RTH. Accordingly, we compared WAT expression patterns in the iT3 and RTH groups. In total, we identified 215 genes that showed twofold change or more in either iT3 or RTH (see complete list in Additional data file 6). Assignment of the named genes to GO biological processes (tier 3) revealed four predominant biological classes: metabolism (n = 45), cell growth and maintenance (n = 24), cell communication (n = 24) and response to external stimuli (n = 16). The expression patterns of genes in these categories provide a molecular framework for interpreting the biological properties of WAT in RTH. The length limitation of this paper permits only description of a selected subset of genes. For example, within the cell communication class we identified a number of genes upregulated in RTH that are involved in cell-cell adhesion via desmosomal junctions (Table 3). Desmosomes are major intercellular adhesive junctions that provide strong mechanical attachments between adjacent cells and are thought to have a role in tissue morphogenesis and differentiation 28. The adhesive core of the desmosome is comprised of proteins of the cadherin family, namely desmogleins and desmocollins. In the WAT of RTH animals, we observed a marked upregulation of transcript levels of cadherin 1, desmoglein 2, desmocollin 2 and the CEA-related cell adhesion molecule 1. Desmosomes are anchored to intermediate filaments of keratin in the cytoplasm. Here, we also observed increased transcript levels of the keratins Krt1-13, Krt1-19 and Krt2-6a. In addition, the most highly upregulated transcript in the WAT of RTH mice was a gene called 5730453H04Rik, which shares 95% identity (over 480 amino acids) with human desmoplakin, a major protein of desmosomes involved in the anchoring of keratin intermediate filaments to desmosomes.

Within the metabolism class, we identified a number of genes negatively regulated in the WAT of iT3 and RTH mice that are directly involved in fatty-acid and lipid metabolism (Table 3). These genes include those previously discovered in the lipogenesis cluster of Figure 3 (that is, Fasn, Acly, Gpd1, Elovl6, Slc25a1 and IMAGE: 5151139). Fatty-acid biosynthesis is known to use large amounts of NADPH. As a consequence, adipose tissue is known to express high levels of the enzymes of the pentose phosphate pathway that generate NADPH. Hexose-6-phosphate dehydrogenase and phosphogluconate dehydrogenase are two of the three major enzymes in the pentose phosphate pathway, and both were negatively regulated in the WAT of RTH animals.

The genes assigned to the class response to external stimuli were, with only one exception, all downregulated in RTH animals, and in some cases, in iT3 mice (Table 3). All the genes identified here are involved in immune response or immune-cell biology and overlap with the previously identified immunity-gene cluster of Figure 3. They include a number of chemokines, HLA antigens and other markers of lymphoid lineages.

Taken together, these data suggest that the WAT of RTH mice is characterized by increased cell-cell adhesion via desmosomal structures, and they recapitulate our initial findings of a possible hyperthyroid-associated inhibition of lipogenesis and a repression of immunomodulatory signaling.

Thyroid hormones are essential for normal brain development and regulate neuronal proliferation, migration and synaptogenesis in the cerebellum 2930. In our study, the cerebellum was the least responsive tissue in both iT3 and RTH mice, with only 13 genes and no genes, respectively showing twofold or greater changes. The complete list of genes with all GO classifications is shown in Additional data file 7. The most highly induced gene (> threefold) in the cerebellum of T3-treated mice was the T3/T4-binding protein, transthyretin, which is synthesized in large amounts in the choroid plexus and is essential for the transport of thyroid hormones from the blood to the brain 31. Increased expression of transthyretin may represent a positive feedback mechanism for enhancing the effect of thyroid hormone in the brain. mRNA for the retinoid X receptor (RXR) was also induced more than threefold in response to T3. RXR is a member of the nuclear receptor superfamily that heterodimerizes with TRs and modulates the transcriptional activity of TRs 1. T3-induced upregulation of RXR may reflect the relative importance of TR-RXR heterodimerization for T3 action in the cerebellum. Also induced by T3 treatment were the genes APC and dickkopf homolog 3, which are known or suspected antagonists of the Wnt signaling pathway. Both T3 and Wnt pathways are known to have important roles in neuronal growth and synaptogenesis 32333435. We recently presented genetic and biochemical evidence that thyroid hormone can inhibit the Wnt signaling pathway 19. Thus, T3-mediated inhibition of Wnt signaling may be an important aspect of cerebellar development and function. The physiologic impact of the activation of these genes in hyperthyroidism warrants further investigation.

The animal protocols used in the present study have been approved by the National Cancer Institute Animal Care and Use Committee. The TRβPV mice contain a cytosine insertion in exon 10 of the TRβ1 gene at nucleotide position 1,642 of the TRβ1 cDNA. This mutation leads to a frameshift of the carboxy-terminal 14 amino acids of TRβ1 1112. TRβPV mice were prepared via homologous recombination and maintained on a mixed background of 129/Sv × C57BL/6J 11. TRβ^+/+ and TRβ^PV/PV male siblings at 8-10 weeks of age were used in this study. Another group of wild-type (TRβ^+/+) male littermates of TRβ^PV/PV received daily intraperitoneal injections of T3 (5 μg/mouse) or the same volume of vehicle (phosphate-buffered saline) for 7 days. On the eighth day the animals were sacrificed by CO₂ exposure and tissues (cerebellum, heart and white adipose) were harvested.

Total RNAs were extracted from liquid nitrogen-frozen tissue specimens using Trizol reagent according to the manufacturer's instructions (Gibco-BRL). To minimize differences in gene expression owing to inter-animal variability, for each tissue and animal category, 5 μg of total RNA from four animals were pooled before RNA amplification. RNA amplification and quality assessment was carried out as previously described 64.

The mouse cDNA microarrays contained 11,520 sequences derived from various sequence-verified clone collections including the Incyte GEM1 mouse clone set (8,700; Incyte), the Research Genetics mouse sequence-verified set (2,600; Research Genetics,), and a private collection of 200 mouse ESTs (a gift from Lothar Heninghausen, NCI, Bethesda, MD). All arrays were manufactured at the NCI Microarray Facility at the Advanced Technology Center, Gaithersburg, MD. Arrays were printed using an OmniGrid microarrayer (GeneMachines) on poly-L-lysine-coated-glass according to Eisen and Brown 63.

For each experimental comparison, duplicate microarray hybridizations were performed according to a reciprocal strategy termed dye-swapping. In this approach, the Cy3 and Cy5 labeling scheme for the two arrays are reversed such that a true differentially expressed gene having a ratio of 2.0 (that is, a 2-fold red (Cy5-biased) spot) on one array, for example, would have a reciprocal ratio of approximately 0.5 (that is, a 2-fold green (Cy3-biased) spot) on the other array. Notably, this form of experimental replication is more rigorous than conventional duplication (that is, true replicates using the same labeling scheme) as it takes into account a recognized form of experimental noise in microarray technology characterized by the reproducible bias of one fluorophore over the other at low signal intensity features (termed dye bias). Three micrograms amplified RNA were used to generate Cy3- or Cy5-labeled cDNA for each array hybridization. cDNA labeling reaction and array hybridization were carried out essentially as described 63. Microarrays were scanned on a GenePix 4000A microarray scanner (Axon Instruments) to generate 16-bit TIFF images of Cy3 and Cy5 signal intensities. The images were analyzed using GenePix Pro 3.0 microarray analysis software to measure fluorescence signals and format data for database deposition.

All array data were deposited in the NCI-CIT microarray database 65 where Cy3 and Cy5 signals were median normalized and expression ratios (Cy5/Cy3) were calculated. The following selection criteria were applied. All spots having a mean signal (after background subtraction) less than twice that of background in both Cy3 and Cy5 channels were systematically excluded from the dataset to eliminate measurements based on poorly detected features. The data were also filtered to exclude spots flagged as missing or corrupt in one or both arrays of a dye-swap pair. Remaining outliers were then selected by the 2.0r criterion, whereby an expression ratio must demonstrate a twofold or greater change on each array of a pair of reciprocal dye-swap hybridizations. We also demanded that the ratios must be in opposite expression directions (that is, > 2.0 on one array, and < 0.5 on the other) to exclude 'false-positive outliers' that showed a reproducible bias towards either red or green fluorescence and thus failed to reciprocate in the dye-swap array pairs. We next calculated the average expression ratio from the dye-swap pairs (after taking the reciprocal ratio from one of the two arrays) and report the average expression ratios in all analyses herein. For reporting genes by name, IMAGE clone IDs corresponding to the microarray probe sequences were used to extract UniGene Cluster IDs and names (UniGene Build 129). For genes represented by multiple probes on the array, a single probe was used (that is, the probe that showed the greatest transcriptional variation) to report the expression ratio so as to prevent gene number biasing owing to redundant probes.

The expression of selected genes was also analyzed by quantitative RT-PCR using the following pairs of primers:

APC forward primer 5'-GCTGACATCTGTGCTGTGGATGG-3';

APC reverse primer 5'-TCCTTAAAGCTGCTGCACTTCCC-3';

somatostatin forward primer 5'-TGCATCGTCCTGGCTTTGG-3';

somatostatin reverse primer 5'-GAGTTAAGGAAGAGATATGGG-3';

tenascin C forward primer 5'-TTCGTGTGTTCGCCATCTTG-3';

tenascin C reverse primer 5'-GTGTGAGGTCGATGGTGGT-3';

carbonic anhydrase 4 forward primer 5'-TCACTGCTAGGACAAAGGTG-3';

carbonic anhydrase 4 reverse primer 5'-AGAGGTTGAATGGGGTTTGG-3';

lysyl oxidase forward primer 5'-CTACATCCAGGCTTCCACG-3';

lysyl oxidase reverse primer 5'-TCTCCTCTGTGTGTTGGCAT-3';

enolase 3β, muscle forward primer 5'-TCAAGGGTTACACTCTGCCT-3';

enolase 3β, muscle reverse primer 5'-TCGTTCACCCACACAAGGAA-3';

creatine kinase, muscle forward primer 5'-TTCCTTGTGTGGGTGAACGA-3';

creatine kinase, muscle reverse primer 5'-TTTTCCAGCTTCTTCTCCATC-3';

GAPDH forward primer 5'-ACATCATCCCTGCATCCACT-3';

GAPDH reverse primer 5'-GTCCTCAGTGAGCCCAAG-3'.

Total RNA (5 μg) prepared as described above was used in the quantitative RT-PCR reaction. cDNA was prepared by SuperScript II reverse transcriptase I (Invitrogen) in the presence of poly(dT) primer. Preliminary experiments were carried out to define the conditions under which the amplification by PCR was in the linear range. The final conditions used for each were: 94°C for 30 sec, 60°C for 30 sec, 72°C for 30 sec followed by 72°C for 2 min. A total of 26 cycles were used for the amplification of APC, carbonic anhydrase 4, lysyl oxidase and somatostatin cDNA. A total of 32 cycles were used for the amplification of tenascin C, enolase 3β (muscle) and creatine kinase (muscle) cDNA. The PCR products were analyzed by 2% agarose gel electrophoresis and detected by ethidium bromide staining. The intensities of the bands were quantified by Eagle Eye (Stratagene). The expression of GAPDH was determined similarly (22 cycles) as an internal control. The band intensities of the quest genes were normalized against the control. The data are expressed as mean value ± SEM (n = 3 independent determinations). Differences between groups were examined for statistical significance using Student's t-test. p < 0.05 is considered statistically significant.